Suitability of SNOMED CT-AU for use in Australian general practice JULIE FRANCES O HALLORAN. Sydney School of Public Health. Sydney Medical School

Transcription

1 Suitability of SNOMED CT-AU for use in Australian general practice JULIE FRANCES O HALLORAN A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy Sydney School of Public Health Sydney Medical School University of Sydney October 2012

2 Abstract An Australian national licence for the Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) was purchased in This thesis aims to assess the suitability of the Australian SNOMED CT release (SNOMED CT-AU) for use in general practice. Methods used were qualitative and quantitative. Workshops held with general practitioners (GPs) and GP clinical software vendors identified issues subsequently investigated through analyses of data from the Bettering the Evaluation and Care of Health (BEACH) study of general practice (coded using the ICPC-2 PLUS clinical terminology) measuring: the extent to which GPs free text terms recorded in BEACH were adequately represented by the ICPC-2 PLUS terms selected in secondary coding; frequency of individual ICPC-2 PLUS term use for reasons for encounter (RFEs) and problems managed; and range of terms used for problems managed. ICPC-2 PLUS symptom and diagnosis terms were mapped to SNOMED CT-AU using a terminology mapping tool (Snapper), and SNOMED CT-AU coverage and quality assessed. At workshops GPs identified deficiencies with current terminologies, including absence of important terms. Coded terms and free text descriptions had equal specificity for 59.3% of 15,447 problems managed. One-quarter of free text descriptions had greater specificity than coded terms. Of 7,580 terms in ICPC-2 PLUS, 6,098 (80.4%) were used by GPs at least once to describe RFEs and/or problems managed. Distribution of term usage was skewed few terms being used to describe a large proportion of both. Of 5,453 ICPC-2 PLUS symptom and diagnosis terms mapped to SNOMED CT-AU, 69.6% were acceptably mapped, but less than half (47.5%) were included in SNOMED CT-AU. Those ICPC-2 PLUS terms acceptably mapped to SNOMED CT-AU usually had high usage in BEACH. The current content of SNOMED CT-AU is not suitable for implementation in Australian general practice. Considerable work on SNOMED CT-AU is required to rectify the deficiencies identified, particularly addition of new content. ii

3 Acknowledgements I would like to express my sincere thanks to the many people who have supported me both professionally and personally throughout this PhD. First and foremost thank you to my supervisors, Prof. Helena Britt and Prof. Graeme Miller, for helping me to conceptualise the research that led to this thesis and for the day-to-day guidance, encouragement and editorial assistance provided. Thank you for believing that I was capable of completing this thesis. Helena, thank you for taking me on as an inexperienced new graduate, and for all the opportunities you have given me over the years. Sincere thanks also to Graeme for his clinical input during the mapping from the ICPC-2 PLUS terms to SNOMED CT-AU. Thank you to all the staff at the Family Medicine Research Centre, in particular Lisa Valenti and Christopher Harrison for their analytic input and the coding staff for the extra work created by the free text data collection. I am especially grateful to Carmen Zhang and Denise Barratt for taking on extra ICPC-2 PLUS tasks when needed. I would also like to thank Clare Bayram for her encouragement as we tackled this journey simultaneously. I know I will look back on our Saturday lunches with fond memories. Sincere thanks to NEHTA. Thank you also to all the BEACH supporting organisations and to the GPs who participated in the BEACH study, without whom BEACH would not exist. I would also like to acknowledge the individuals in organisations of which I am a member the Wonca International Classification Committee, the IHTSDO IFP/GP SIG and IHTSDO GP/FP RefSet and ICPC mapping project group. Their willingness to share their extensive experience in the area of classifications and terminologies has helped to develop my ideas and improve my knowledge in this field. On a personal level I thank my family my parents James (Jim) and Christine O Halloran and my brother Stephen O Halloran. You first taught me the concept of a place for everything, and everything in its place, and, in iii

4 choosing a career in classifications and terminologies, this has inadvertently become one of the principles of my working life. I could not have completed this thesis without your support and encouragement. I would especially like to thank Mum (Christine O Halloran) for editing this thesis. Finally, most sincere thanks to Craig Gordon for the sacrifices you have made to allow me to complete this thesis. I will always be grateful for your constant encouragement and unfailing patience. iv

5 Table of contents Abstract... ii Acknowledgements... iii Table of contents... v List of tables... ix List of figures... x List of boxes... xi List of appendices... xii Glossary of terms used in this thesis... xiii List of abbreviations used... xix 1 Introduction A history of classifications and clinical terminologies until the mid-20 th century Development of general practice classifications International development of clinical terminologies Clinical terminologies in Australian general practice The use of SNOMED CT-AU in Australian general practice Aims of this thesis and the candidate s contribution Aims Candidate s contribution Candidate s involvement in research leading to this thesis Resources used in this thesis Bettering the Evaluation and Care of Health (BEACH) Glossary of terms used in BEACH Governance and funding Ethics GP sampling and recruitment Data elements collected in BEACH Data entry Representativeness Statistical methods used in BEACH Current status of BEACH ICPC-2 PLUS Glossary of terms used for ICPC-2 and ICPC-2 PLUS ICPC ICPC-2 PLUS ICPC-2 PLUS maintenance Relationship between ICPC-2 PLUS and BEACH Current status of ICPC-2 PLUS, April SNOMED CT v

6 3.3.1 SNOMED CT glossary Scope and structure Additional functionality Maintenance SNOMED CT-AU Status of SNOMED CT-AU Application of resources in the methods used in this thesis Development of requirements for an Australian SNOMED CT general practice reference set Background Introduction Methods Method of selecting participants Method of analysis Findings Characteristics of workshops Terminologies or coding systems currently incorporated into GP clinical software Themes emerging from workshops Development of requirements from results of the workshops Content requirements Implementation requirements Governance requirements Limitations Conclusion Current use of medical terminology by GPs in Australia Background Comparison of GP free text descriptions and ICPC-2 PLUS coded terms Background Aims Methods Results Discussion Number of terms used in a general practice clinical terminology Background Aims Methods Results Discussion The content of the problems managed data element Background Aims Method Results Discussion Maintenance of a general practice terminology Background vi

7 5.5.2 Aims Methods Results Discussion Discussion Labelling problems managed Certainty of diagnosis Conclusion Mapping terms from ICPC-2 PLUS to SNOMED CT-AU Introduction Aims Methods Tooling Map result categories Mapping principles Mapping process Analysis Results Overall mapping results Comparison of acceptable and unacceptable matches in the mapping results Relationship between map result category and utilisation of terms Qualitative findings and recommendations Ambiguity in SNOMED CT-AU concepts Inconsistent/incorrect inclusion of synonyms Inconsistent concept labels Precoordination in SNOMED CT-AU Content missing in SNOMED CT-AU Issues when choosing a map of best fit Issues with the terminology used in SNOMED CT-AU descriptions SNOMED CT hierarchies Ambiguity in SNOMED CT-AU Discussion Limitations of this study Mapping terminologies to SNOMED CT-AU Use of SNOMED CT-AU as an interface terminology Editorial consistency in SNOMED CT-AU content Hierarchical issues Education and user satisfaction Maintenance Conclusion Discussion Conclusion Postscript References in citation order vii

8 References in alphabetical order (citation order in parentheses) Appendices viii

9 List of tables Table 3.1: Number of ICPC-2 PLUS terms by ICPC-2 chapter and component, April Table 4.1: Comparison of market share against consultation Table 5.2.1: Assessment of the match type for problems managed between free text and ICPC-2 PLUS coded terms Table 5.2.2: Assessment of terms marked as more specific Table 5.2.3: Distribution of attributes included in free text Table 5.3.1: Use of terms from ICPC-2 PLUS in BEACH, Table 5.3.2: ICPC-2 PLUS terms most frequently used to describe reasons for encounter, Table 5.3.3: ICPC-2 PLUS terms most frequently used to describe problems managed, Table 5.4.1: Distribution of problems managed, by ICPC-2 component, Table 5.4.2: Distribution of problems managed, by ICPC-2 component, BEACH, to Table 5.4.3: Distribution of process codes used to describe problems managed, Table 5.5.1: ICPC-2 PLUS terms added according to ICPC-2 component, Table 5.5.2: Growth in keywords in ICPC-2 PLUS, to Table 5.5.3: Requests for new terms by requestor, Table 5.5.4: Results of new term suggestions, Table 6.1: Overall mapping results, ICPC-2 PLUS to SNOMED CT-AU Table 6.2: Distribution of SNOMED CT-AU concepts included in the map according to the SNOMED CT-AU semantic tag Table 6.3: Distribution of mapping results according to ICPC-2 PLUS terms used at least once as reasons for encounter and/or problems managed in BEACH (a) Table 6.4: Comparison of mapping results and utilisation for RFEs and/or problems managed combined, and problems managed alone Table 6.5: Distribution of map result types according to ICPC-2 PLUS term utilisation in either the RFE and/or problems managed data elements. 154 ix

10 List of figures Figure 3.1: The BEACH relational database Figure 3.2: The structure of the International Classification of Primary Care Version 2 (ICPC-2) Figure 3.3: Structure and meaning of an ICPC-2 PLUS code Figure 3.4: Example of the representation of links between keywords and terms Figure 5.3.1: Number of terms describing reasons for encounter expressed as a proportion of total usage Figure 5.3.2: Number of terms describing problems managed expressed as a proportion of total usage Figure 5.5.1: Size and growth of ICPC-2 PLUS, to (a) Figure 5.5.2: New term additions as a proportion of all terms added, by ICPC-2 component, to Figure 6.1: Distribution of mapping results within each ICPC-2 chapter, by map category Figure 6.2: Distribution of mapping results according to ICPC-2 components and map result category Figure 6.3: Chapter component specific proportion of acceptable mapping results Figure 6.4: Relationship between number of terms used to describe RFEs and/or problems managed and term utilisation x

11 List of boxes Box 3.1: Glossary of terms from the BEACH study used in this thesis Box 3.2: Glossary of terms used in conjunction with ICPC-2 and/or ICPC-2 PLUS Box 3.3: Glossary of terms used in conjunction with SNOMED CT and/or SNOMED CT-AU Box 3.4: Top level hierarchies in SNOMED CT Box 3.5: Example of SNOMED CT concept and its constituent parts Box 5.2.1: Criteria used to assess relationship between free text and ICPC-2 PLUS term Box 5.4.1: Labels and definitions used to describe the concept of problem/diagnosis Box 6.1: Categories used to evaluate mapping results Box 6.2: Hierarchical representation of Postpartum subinvolution of breast (finding) and Breast subinvolution, postpartum (disorder) Box 6.3: Representation of Abscess and its descendents Box 6.4: Representation of Irritable colon and its descendents Box 6.5: Representation of hierarchy below Finding of size of ear canal Box 6.6: Hierarchical representation of concepts related to flatulence Box 6.7: Representation of Tenosynovitis and its descendents Box 6.8: Hierarchical representation of Thyroid disease in pregnancy Box 6.9: Hierarchical representation of Feeling of loss of feeling (finding) xi

12 List of appendices Appendix 1: Project plan and requirements specification: General Practice Reference Set Project Appendix 2: BEACH encounter form, Appendix 3: ICPC-2 pager Appendix 4: GP characteristics questionnaire, Appendix 5: Letter of invitation to GPs attending GPRS workshop xii

13 Glossary of terms used in this thesis A general glossary of terms used in this thesis is found below, in alphabetical order and with sources identified in italics. There are separate glossaries for terms that relate specifically to the different resources used in this thesis: Terms used in relation to the BEACH study of general practice are found in Box 3.1. Terms from ICPC-2 and/or ICPC-2 PLUS are included in Box 3.2. Terms that relate specifically to SNOMED CT and/or SNOMED CT-AU are included in Box 3.3. ATOMIC The least portion of a thing or quality. (Australian Standard AS Language of health concept representation) ATTRIBUTE Represents a characteristic of the meaning of a concept or the nature of a refinement. (SNOMED CT Technical Implementation Guide, January 2011) CLASSIFICATION A hierarchical organisation of terms or ideas that allows aggregation into categories that can be counted and compared without double counting. Also called aggregate terminology. (SNOMED CT Technical Implementation Guide, January 2011) CLINICAL TERMINOLOGY The component of health language used at the point of care for the purpose of clinical management of subject(s) of care. (Australian Standard AS Language of health concept representation) In this thesis, a clinical terminology refers to the group of terminologies that include interface terminologies and reference terminologies, but not classifications. xiii

14 CODE A controlled representation of a concept or group of concepts. Codes may be alphabetic, numeric or combinations of these and are constrained in meaning and use. The code may be structured to indicate a hierarchy. Ideally should not have meaning. (Australian Standard AS Language of health concept representation) CODING SYSTEM A system of code sets, coding standards and code maintenance procedures together with their authorization and governance. (Australian Standard AS Language of health concept representation) COMPOSITE CONCEPT A concept created from two or more concepts, forming a new concept by the combination of other concepts. (Australian Standard AS Language of health concept representation) CONCEPT A concept is an idea that represents a thing. It can have one or more names. (Australian Standard AS Language of health concept representation) DATA Representation of real world facts, concepts or instructions in a formalised manner suitable for communication, interpretation or processing by human beings or by automatic means. (Australian Standard AS Language of health concept representation) DATA ELEMENT The descriptive name of the information object that contains data. A unit of data for which the definition, identification, representation and permissible values are specified by means of a set of attributes. (Australian Standard AS Language of health concept representation) xiv

15 DOMAIN An organisational or functional part of any social construct or field of knowledge (e.g. hospital, enterprise, acute care system, general practice, the health system, etc). A domain has a specified scope and specific characteristics that are relevant to that scope. (Australian Standard AS Language of health concept representation) ELECTRONIC HEALTH RECORD (EHR) A repository of information regarding the health status of a subject of care, in computer processable form. (Australian Standard AS Language of health concept representation) GENERAL PRACTITIONER A medical practitioner who provides primary, comprehensive and continuing care to patients and their families within the community. (The Royal Australian College of General Practitioners) GENERAL PRACTICE The provision of primary, continuing, comprehensive, whole-patient medical care to individuals, families and their communities. (The Royal Australian College of General Practitioners) GRANULARITY See Specificity INFORMATION Data that are interpreted, organised and structured. (Australian Standard AS Language of health concept representation) INFORMATION MODEL Formal representation of a domain using diagrammatic tools, according to a set of modelling rules, showing key concepts and their relationships. (Australian Standard AS Language of health concept representation) xv

16 INTERFACE TERMINOLOGY A maintained set of unique identified terms designed to be compatible with the natural language of the user. (Australian Standard AS Language of health concept representation) MAPPING A relationship between the code or term used to represent a health concept in one system, and the code or term that would be used to represent the same concept in another coding or terminology system. (Australian Standard AS Language of health concept representation) MEDICAL RECORD A collection of paper files or electronic documentation containing the patients demographic data the medical history obtained by a physician or other professionals, opinions and other relevant health information and a problem list. (Wonca Dictionary of General/Family Practice) NATURAL LANGUAGE Words, phrases, images and sentences used by a human community where there is no structure imposed by a system. (Australian Standard AS Language of health concept representation) POSTCOORDINATED TERM A concept which is expressed using two or more concepts, which together, are not represented by a single code. (Australian Standard AS Language of health concept representation) PRECOORDINATED TERM Two or more concepts combined, which together, are considered a single concept with unique identification. (Australian Standard AS Language of health concept representation) xvi

17 REFERENCE TERMINOLOGY A terminology designed to uniquely represent concepts. It does this by listing the concepts and specifying their structure, relationships and, if present, their systematic and formal definitions. It normally contains a unique identifier, a rubric and may contain reference to alternate terms to the preferred term (which may be conceptualised as an interface term) and it may contain maps or pointers to aggregate terminology. (Australian Standard AS Language of health concept representation) SEMANTIC EQUIVALENCE The extent to which two or more terms or concepts have the same meaning. SEMANTIC INTEROPERABILITY The capture, transfer and use of information in a manner in which the meaning of the information is retained between different parties in the health care process. (Semantic Interoperability for Better Health and Safer Healthcare: deployment and research roadmap for Europe) SPECIFICITY The level of detail in a term or concept. TERM A word, phrase or symbol which represents a concept in a special language. (Australian Standard AS Language of health concept representation) TERMINOLOGY MODEL A model which describes the meaning of what is to be stored, typically expressed in hierarchies, frame systems or their more modern successors, description logics. (Australian Standard AS Language of health concept representation) xvii

18 USER INTERFACE The way a software application presents itself to a user including, on its screen appearance, the commands it puts at a users disposal, and the manner in which the user can access and update information by using the application. (SNOMED CT Technical Implementation Guide, January 2011) xviii

19 List of abbreviations used ACRRM Australian College of Rural and Remote Medicine AIHW AMA AMTS APCC AS BEACH CAP CAP STS CATCH CI CORE Australian Institute of Health and Welfare Australian Medical Association Australian Morbidity and Treatment Survey Australian Primary Care Collaboratives Australian Standard Bettering the Evaluation and Care of Health College of American Pathologists College of American Pathologists SNOMED Terminology Solutions Classification And Terminology of Community Health Confidence interval Clinical Observations and Encoding (subset of SNOMED CT) CSIRO Commonwealth Scientific Industrial and Research Organisation CTV-3 Clinical Terms Version 3 CUI Docle DoHA EDRS EHR FMRC FSN Common User Interface Doctor Command Language Department of Health and Ageing (Australian Government) Emergency Department Reference Set Electronic Health Record Family Medicine Research Centre (University of Sydney) Fully Specified Name xix

20 GORD GP GPCG GPRS ICD ICD-9 ICD-10 ICD-10-AM ICHPPC ICHPPC-2 ICPC Gastro-oesophageal reflux disease General practitioner General Practice Computing Group General Practice Reference Set International Statistical Classification of Diseases and Related Health Problems International Statistical Classification of Diseases and Related Health Problems, Ninth Revision International Statistical Classification of Diseases and Related Health Problems, Tenth Revision International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification International Classification of Health Problems in Primary Care International Classification of Health Problems in Primary Care, Version 2 International Classification of Primary Care ICPC-2 International Classification of Primary Care, Version 2 ICPC-Plus The PLUS interface terminology classified to the International Classification of Primary Care, Version 1 ICPC-2 PLUS The PLUS interface terminology classified to the International Classification of Primary Care, Version 2 ID IFP/GP SIG Identifier International Family/General Practice Special Interest Group xx

21 IHTSDO MBS MED MSIA NEC NEHTA NHS NOS PCEHR PIP POMR RefSet RACGP RF1 RF2 RFE RFEC SNOMED SNOMED CT International Health Terminology Standards Development Organisation Medicare Benefits Schedule Medical Entities Dictionary Medical Software Industry Association Not elsewhere classified National e-health Transition Authority National Health Service Not otherwise specified Personally Controlled Electronic Health Record Practice Incentives Program Problem Oriented Medical Record Reference set (of SNOMED CT) Royal Australian College of General Practitioners Release Format 1 (SNOMED CT) Release Format 2 (SNOMED CT) Reason for encounter Reason for encounter Classification Systematized Nomenclature of Medicine Systematized Nomenclature of Medicine Clinical Terms SNOMED CT-AU Systematized Nomenclature of Medicine Clinical Terms, Australian release SNOMED RT Systematized Nomenclature of Medicine Reference Terminology SNOP Systematized Nomenclature of Pathology xxi

22 UK URTI URU US WICC Wonca United Kingdom Upper Respiratory Tract Infection Understandable, Reproducible and Useful United States (of America) Wonca International Classification Committee World Organisation of National Colleges, Academies and Academic Associations (World Organisation of Family Doctors) xxii

23 1 Introduction General practice is the provision of primary, continuing, comprehensive, whole-patient medical care to individuals, families and their communities. 1 General practice sits at the core of the Australian health care system. General practitioners (GPs) are usually the first point of contact in the Australian health care system, and act as a gatekeeper to access other services (e.g. specialist services). In , 83% of Australians visited a GP at least once, as determined by claims for GP services from Australia s national health insurance scheme, known as Medicare (personal communication, Department of Health and Ageing (DoHA), June 2010). Between April 2010 and March 2011, Medicare rebates were paid for approximately million GP services, 2 equating to an average of 5.3 visits per head of population, or 6.3 visits per person who visited a GP at least once. Australian general practice is funded on a fee-for-service basis. Medicare provides subsidised access to hospital and privately provided medical services (including visits to general practitioners) for all Australian citizens. 3 In the overall context of health care, the importance of general practice is best represented by research conducted in the United States (US) by White et al, who suggested in 1973 that nearly three-quarters of a population will visit an ambulatory care physician (equivalent to a GP) at least once in a year, compared with 10% of the population admitted to a hospital. 4 This research demonstrated that clinicians (such as GPs) working in the community provide the majority of health care to the population, and therefore that GPs have an important responsibility to manage the health of the community. Nearly 40 years later, White et al s results are still used to differentiate between the frequency of health service use in terms of hospital and ambulatory care. In Australia, patients can choose which GP they visit. Patients do not register with a practice, and there is no obligation to see the same GP continuously. Despite this, GPs have a significant role in co-ordinating the care of their patients, especially for older patients and those with chronic disease. A concept that has recently become popular (particularly in the United States) is that of the medical home, where a patient can choose a clinician who acts as 1

24 the central point of co-ordinating and managing their health care. 5 This concept is designed to promote continuity of care. The Royal Australian College of General Practitioners (RACGP) supports the introduction of the medical home construct in Australia. This would involve patients voluntarily registering with a GP, who then acts in a medical home management role. 6 Often GPs in Australia informally take on the role of care co-ordinator, making the concept of the medical home a logical extension to the role of GPs. However there are currently no arrangements in place or funding allocated to formalise the medical home construct in Australia. The role of care co-ordinator places vital importance on GPs having accurate and timely information about their patients, in the form of medical records. Until the 1980s medical records in Australia were primarily held on paper. In the 1980s initiatives were introduced to increase the uptake of computers in general practice. 7 This continued throughout the 1990s, with a 1995 study reporting that 14% of GPs in the state of New South Wales used computerised patient records, while a study commissioned in 1997 by the then (Australian) Department of Health and Family Services found that 31% of GP practices had a computer, and 15% of GPs reported using computerised medical records. 8 The uptake of computerised medical records has increased substantially in recent years. In % of Australian GPs stated they used computers for clinical purposes. Nearly two-thirds (64.7%) reported that their practice was paperless. More than one-quarter of GPs (28.8%) used a hybrid system, utilising a combination of electronic and paper-based systems. 9 The RACGP first introduced standards for paper-based health records in based on Dr Lawrence Weed s Problem Oriented Medical Record (POMR). 11 The introduction of computerised medical records (now known as electronic health records or EHRs) meant that clinical information could be stored dynamically, using structured relational databases. Quickly a need developed to standardise the data structures used in computerised medical records. The RACGP developed a proposed minimum standards document for computerised medical records in 1981, and updated this in 1988 to reflect changes in technology. 7 A decade later the General Practice Data Model and 2

25 Core Data Set Project produced a detailed standard for the structure of general practice electronic health records, released in Standards only have an impact if they are accompanied by policies to enforce or encourage their adoption. Although standards for the structure of EHRs were developed during the early computerisation of general practice, policies were not developed in parallel (either in the form of incentives or penalties) to provide an impetus for software developers to include the standards in their EHR software. At present, no standards are enforced in Australia regarding the structure of GP EHRs and general practice software vendors create their own EHR specifications. Therefore each EHR has been developed in an ad hoc manner with no standardisation between products. As a result there are approximately fifteen commercial EHR software products available in Australia (unpublished data from the Bettering the Evaluation and Care of Health (BEACH) study, ), each with a different technical structure. This is a significant hindrance to users, particularly for those who travel between practices on a regular basis. There are two types of standardisation needed in GP EHRs: 1. standards relating to the structure of the EHR (creating technical interoperability 13 ) 2. standards about the content of the EHR (creating semantic interoperability 13 ). The development and implementation of both technical and content standards for EHRs are vital for interoperability. Technical standards for EHR data models need to be agreed upon and implemented, including standardised labels for each piece of data collected (called data elements) and definitions for each data element. Standardisation of the technical aspects of an EHR will promote technical interoperability, by which data containing the same labels and definitions can be transferred between EHRs. Semantic interoperability is the ability to capture, transfer and use information in a manner in which the meaning of the information is retained between 3

26 different parties in the health care process. The Semantic Health project, funded by the European Union, has defined four levels of semantic interoperability, ranging from no interoperability (where the meaning of information cannot be transferred and interpreted) to full interoperability (where the meaning of all available information is seamlessly transferred and accurately interpreted). 14 Technical and semantic interoperability are intrinsically linked so that data are collected using equivalent data element labels and definitions, and the content contained within these data elements is collected in a manner that allows accurate interpretation by third parties. The need for technical and semantic interoperability between EHRs is growing. Clinical information is flowing between different health care providers in the form of problem lists, health summaries, referral letters and discharge summaries. In addition, there is an increasing need to extract data from medical records for epidemiological or statistical purposes, patient recall for chronic conditions, quality assurance, and in some countries to facilitate billing. If semantic interoperability is to be achieved, the content contained in EHRs must be standardised. This can be done through the use of classifications and standardised clinical terminologies. At the most basic level, classifications and clinical terminologies are standardised lists of clinical terms. When inserted into an EHR, they act as tools that allow the clinical terms and language used by clinicians to be represented in a manner that promotes interoperability. Classifications are an ordering of all elements of a domain into groups according to established criteria. 15 In other words, clinical classifications group similar clinical concepts together for the purposes of counting. Classifications are primarily used for statistical and epidemiological purposes. In contrast, clinical terminologies are the component of health language used at the point of care for the purpose of clinical management of subject(s) of care. 16 All classifications and clinical terminologies are made up of codes, terms and/or concepts. According to Australian Standard AS , a code is a controlled representation of a concept or group of concepts. Codes may be 4

27 alphabetic, numeric or combinations of these and are constrained in meaning and use. The code may be structured to indicate a hierarchy. 16 Terms are a word, phrase or symbol which represents a concept in a single language 16 and represent the label of the clinical concept (equivalent to a clinical idea). In most classifications and clinical terminologies, terms and/or concepts are linked to codes in a one-to-one relationship, whereby each term/concept has one linked alphabetic, numeric or alphanumeric code to uniquely represent that individual term/concept. The act of recording a term/concept with its associated code from a classification or clinical terminology in a medical record is called clinical coding. Classifications and clinical terminologies both represent tools to standardise the language used in health care. However, they have distinctly different purposes, which need to be clearly understood. Classifications are primarily used for grouping, allowing clinicians, researchers and epidemiologists to analyse the clinical conditions (morbidity) seen in a population. Labels used in classifications are often expressed in conjunction with the terms other, not elsewhere classified (NEC) and not otherwise specified (NOS). For example, the International Classification of Primary Care, which will be discussed in more detail at a later point, includes a group labelled Other viral diseases. In classifications the label of the concept is usually called a rubric label. It is clear that this rubric label includes a collection of viral diseases; however by itself the label does not provide any details about the clinical conditions contained within the group. Therefore, the recording of this rubric label in a GP EHR, and transfer of the label to another clinician would provide insufficient detail for the receiving clinician to interpret and understand the patient s clinical condition, and provide continuity of care. In contrast, clinical terminologies include labels that contain more specificity than classifications, and are designed for use within EHRs at the point of care. Labels in clinical terminologies do not include the terms NOS or NEC; each concept is intended to be an independent entity that is appropriate for use within a clinical record. As such, semantic interoperability is achieved through clinical terminologies rather than classifications. 5

28 1.1 A history of classifications and clinical terminologies until the mid-20 th century Until the mid 20 th century, health information was primarily needed for statistics about morbidity (disease) and mortality (causes of death). The earliest coding systems from the 1600s and 1700s resembled classifications, and focussed on the epidemiological need to describe diseases for population health and statistical purposes. The state of health among the people differs in different times and in different places; and the principal purpose of the registration of diseases is to determine the degree of their variation in each district, and in each class of the population, as well as the extent to which they are modified by circumstances (William Farr, Appendix to the First Report of the Registrar General, circa 1839, reproduced in Vital Statistics). 17 The London Bills of Mortality illustrate one of the earliest attempts to collect, order (or classify) and analyse causes of death. 18 The initial focus of the Bills of Mortality was to identify deaths due to plague, 19 but subsequently allowed John Graunt in 1662 to publish one of the earliest reports of disease statistical analysis, investigating associations between many causes of death and age, sex and location of residence. Graunt also commented on changes in rates and causes of death over time, birth patterns and population rates. 18 Some impetus for creating classifications of disease (historically sometimes called nomenclatures) came from the successful development of botanical classifications of plants. While better known for their botanical classifications, Sauvages 20 and Linnaeus 21 individually also created disease classifications. In the first place, it is necessary that all diseases be reduced to definite and certain species, and that, with the same care which we see exhibited by botanists in their phytologies; since it happens, at present, that many diseases, although included in the same genus, mentioned with a common nomenclature, and resembling one another in several symptoms, are, not withstanding, different in their natures, and require a different medical treatment. (Thomas Sydenham, Observationes Medicae, 1676) 22 William Farr, as Superintendent of the statistical department of the Registrar General s Office in London, was responsible in the mid 19 th century for the continuing work of creating statistics from the Bills of Mortality. His interest in, and development of, classifications was heavily influenced by his statistical background. 6

29 The advantages of a uniform statistical nomenclature, however imperfect, are so obvious, that it is surprising no attention has been paid to its enforcement in bills of mortality The nomenclature is of as much importance in this department of inquiry as weights and measures in the physical sciences, and should be settled without delay. (William Farr, 16 th Annual Report of the Registrar-General, circa 1856, reproduced in Vital Statistics). 17 For the next century, developments in classifications focussed on the statistical need for data around causes of death. Farr developed the required classification for England s mortality statistics within the Registrar General s office for approximately fifteen years. In 1853 he and Marc d Espine were asked by the Statistical Congress to prepare a uniform nomenclature of causes of death for use internationally. 17,23 Farr and d Espine approached disease classifications from different viewpoints and disagreed on each other s approach towards the classification of disease, resulting in discord at the Statistical Congress in The Congress decided that the classification of diseases was a matter of secondary importance 23 and focussed on a list of diseases of frequent occurrence that could be used to compare causes of death. 23 This is the first reference to classifications being used for international comparisons of data, and illustrates a move towards comparing data on causes of death, not just within countries but also between countries. The next substantial effort to produce a classification of diseases did not occur until the 1890s, with the development of the Bertillon Classification of Causes of Death, again to promote statistical comparisons of causes of death. 23 After revision in 1900 this was known as the International List of the Causes of Death. 24 The need to widen the classification to collect and analyse data about morbidity in living people was identified in 1938 at the Fifth International Conference for the Revision of the International List of Causes of Death, 25 and represented a major shift in the reasons for classifying diseases. The revision was endorsed in 1948, and the name of the classification changed to The Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death. 24,25 This classification became known as the International Classification of Diseases (ICD) and is now in its tenth iteration (ICD-10). 24 ICD-10 is still primarily used for epidemiological purposes in hospital based settings, and for mortality statistics. 7

30 1.2 Development of general practice classifications Coding in general practice began with the use of classifications. ICD was the classification most often used until the mid-20 th century. In 1959 the Research Committee of the Council of the College of General Practitioners in the United Kingdom (UK) developed a new classification of diseases for general practice morbidity coding. The classification was based on ICD and was produced using three principles: frequency of occurrence in general practice, the accuracy with which the term described the appropriate clinical situation and its relationship to ICD. 26 The classification was widely used in the E Book, a research tool used to record morbidity in UK general practice. 27 By the 1960s it was widely acknowledged by general practice researchers that the information needs of general practice were not being met by using the ICD as a statistical classification. A number of researchers commented on the reasons why ICD was not a suitable classification for general practice, for example, that it focussed too heavily on diagnoses rather than symptoms, 26 was too large for use in general practice, and did not contain sufficient content relating to preventive medicine, social problems and administrative issues. 28 However, early general practice classifications were all based on ICD in some way, because that was the gold standard at the time, and a basis through which international comparisons could potentially be made. In their 1967 study, GPs Westbury and Tarrant found that 23% of diagnostic labels used in Canadian general practice (known in Canada as family practice) could not be classified using ICD (Version 8) and 45% could not be classified using the revised Classification of the British College of General Practitioners (1963 version). 29 They subsequently developed a classification for use in Canadian general practice called the Canuck Classification, which was released in At the 5 th World Conference on General Practice in 1972 the World Organisation of Family Doctors (Wonca) was formally established. At this meeting, the following motion was unanimously passed: 8

31 That this combined meeting recommends to Wonca the establishment of an international working party to consider and develop an agreed classification of disease in general practice for presentation at the next meeting of Wonca and that this classification be clearly related to the International Classification of Diseases of the World Health Organisation. 31 Interestingly, the requirement that the classification be related to ICD was an addendum to the original motion. This restricted any opportunity for the new general practice classification to move away from the ICD structure, which is primarily based in disease aetiology (the causes of disease). The resulting classification drew on two existing classifications the Canuck Classification 28 and ICD It was released as the International Classification of Health Problems in Primary Care (ICHPPC) 33 and approved for use by Wonca in 1974 at the Sixth World Conference. 33,34 The revision of the International Classification of Diseases to Version 9 (ICD-9) led to a revision of ICHPPC to retain comparability with ICD. This classification was called ICHPPC-2 and published in Definitions for codes (rubrics) were added to the ICHPPC during the 1983 update and published as ICHPPC-2 Defined. 36 Reasons for encounter (RFEs), the description of the patient s perspective for seeking general practice care, were not adequately captured using ICHPPC. This was not surprising, given ICHPPC s relationship to ICD, and the well known inadequacy of ICD for use in general practice. A Reason for Encounter Classification (RFEC) was developed and pilot tested in the early 1980s. 37,38 This classification was based on localisation (or body systems) rather than disease aetiology. The acknowledgement that a single classification would better satisfy the needs of primary care led to the RFEC becoming the basis of the International Classification of Primary Care (ICPC). ICPC enabled the classification of reasons for encounter, diagnoses, and processes of care. It was released in ICPC was revised in 1998 and subsequently released as ICPC-2. The main areas updated in ICPC-2 were the development of criteria for rubrics for symptoms and complaints and diagnoses and diseases, a conversion table to ICD-10, and corrections to individual rubrics. 40 9

32 1.3 International development of clinical terminologies If we can distinguish diseases, surely we can likewise say in what manner we do distinguish them William Cullen (Lectures introductory to the course on the practice of physic, reproduced in The Works of William Cullen, circa 1800s). 41 In terminological circles, the terms lumper and splitter were first used in the medical context by McKusick in the 1960s 42 and are often used to distinguish between classifications (characterised by less specificity and grouping of similar concepts, or lumping ) and terminologies (characterised by more specificity or splitting ). Even in the very early days of disease classification, there were debates about the best way to include diseases in classifications and terminologies. Many of the early attempts to classify diseases developed hierarchical systems that included varying levels of detail, from the very general to the specific. However, the differentiation between systems needed for lumping and splitting has become more marked in recent years, due mostly to the need for semantic interoperability, where clinical terms stored in an EHR must be unambiguously recorded so they can be electronically transmitted to another EHR and accurately interpreted by the receiving party. The development of detailed clinical terminologies, such as the Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) allows such information to be captured with greater specificity. SNOMED CT is a reference terminology that purports to be the most comprehensive, multilingual clinical terminology in the world. 43 It contains clinical concepts from both human and veterinary medicine and is a multi-hierarchical terminology, containing concepts at various levels of detail. The first version of the SNOMED terminology was the Systematized Nomenclature of Pathology (SNOP), and was created by the College of American Pathologists (CAP) in SNOP was designed as a terminology for anatomical pathology and included four axes for anatomical site, morphology, aetiology and physiological characteristics. This terminology was succeeded in 1974 by a broader terminology, the Systematized Nomenclature of Medicine (SNOMED), which contained additional content for 10

33 diseases and procedures. 45 Various iterations followed until the development of the Systematized Nomenclature of Medicine Reference Terminology (SNOMED RT), in Simultaneously in the UK a system commonly referred to as the Read codes was developed. The Read codes were created in the 1980s by a British GP, Dr James Read. The first version of the Read codes was based on ICD-9, and restricted to diseases commonly found in general practice. Dr Read expanded the coding system to include anything that might be entered into a patient s computerised record. 46 The process used to guide this expansion is not documented. In 1990 responsibility for the Read codes was transferred to the UK Secretary of State for Health. The most commonly used version of the Read codes, known as the 5-byte Version 2 or Read codes Version 2 was developed in The Read codes were seen as a useful tool in the UK for interoperability between different sectors of the health system, and in 1992 the UK National Health Service established the Clinical Terms Project to widen the scope of the Read codes to incorporate all clinical specialties. The resulting product, called Clinical Terms Version 3 or CTV-3, was created by consulting with groups of clinicians representing different clinical specialties and collecting lists of clinical terms from each clinical specialty to create a thesaurus of terms used across medicine. 47 CTV-3 was first released in In the late 1990s, two authors produced key papers that reflected the attitudes towards clinical terminology at the time. James Cimino s work, published in 1998 and now commonly referred to as Cimino s Desiderata, outlined a series of principles for the structure and content of clinical terminologies. Cimino s Desiderata included broad principles, including (but not limited to) the idea that concepts should be made inactive but never deleted (concept permanence), concepts should be clear and unambiguous (concept orientation) and that a clinical terminology should contain information at varying levels of granularity. 49 In the following year, Alan Rector published a paper titled Clinical Terminology: Why Is it so Hard?, which outlined ten difficulties faced by developers of clinical terminologies. 50 There is some 11

34 overlap between these papers, which together provided a level of consensus for both the philosophical path and structural characteristics needed for future clinical terminologies. In 1999 the College of American Pathologists (CAP) (who owned and developed SNOMED) and the UK s National Health Service agreed to combine SNOMED RT with the UK s CTV-3 to create SNOMED CT. According to CAP, SNOMED RT and CTV-3 were the two foremost terminologies available at the time, and combined, they could create a world class terminology. 44 The process undertaken to merge the two terminologies is outlined in detail in a paper by Stearns et al. 51 In summary, the upper level hierarchies of the two terminologies were merged and their content streamlined. A multi-hierarchical structure was developed for the new terminology and populated. Alpha and beta tests were undertaken, 51 and the first version of SNOMED CT was released in January After its release, responsibility for the development of SNOMED CT was retained by CAP. As SNOMED CT was a commercial terminology owned by the US and UK, other countries were reluctant to adopt SNOMED CT because they were concerned about long-term access to the terminology. To dispel these concerns, it was announced in 2005 that a new standards body would be created to oversee governance and maintenance of SNOMED CT on an international level. 52 In 2007 the intellectual property of SNOMED CT was transferred from CAP to the newly formed International Health Terminology Standards Development Organisation (IHTSDO). Nine countries were charter members of the new organisation, including Australia. 53 The organisation was incorporated in Denmark, and established its headquarters there. Editorial responsibility for SNOMED CT was kept with CAP, who had a contract with the IHTSDO to continue providing support. 54 The formation of the IHTSDO removed the proprietary nature of the terminology and allowed countries other than the US and UK to contribute to its development. It also marked the internationalisation of SNOMED CT and increased acceptance of SNOMED CT as a terminology standard. The other Charter members included Canada, Denmark, Lithuania, the Netherlands, New Zealand, Sweden, the US and the UK

35 In 2012, there are 19 Member countries in the IHTSDO. 56 The IHTSDO is overseen by a Management Board and General Assembly, with advisory groups in four main areas content, implementation and innovation, technical, and quality assurance. The wider IHTSDO community of practice contains Special Interest Groups and Project Groups, facilitating the participation of individuals and organisations with specific interests, 57 including a General/Family Practice Special Interest Group Clinical terminologies in Australian general practice As previously stated, the computerisation of general practice led to a need for semantic interoperability, in the form of terminologies containing terms and/or concepts that were more specific than those available in classification systems. The need for appropriate clinical terminologies in general practice was recognised as a priority in Australia as early as 1985, when the RACGP 5 th Computer Conference recommended that the RACGP give the highest priority to the early resolution of difficulties in coding of primary health problems. 59 In the early 1990s a Coding Workshop was held by the Information Management Steering Group (a joint group involving the RACGP, the Australian Medical Association (AMA) and the Australian Commonwealth Government) to review the coding systems available for general practice. No suitable terminologies existed at the time in Australia. During the workshop, reasons for coding were identified, including continuity of care, quality assurance, auditing and decision support. 60 Subsequently Drs Graeme Miller and Helena Britt from the (then) Family Medicine Research Unit at the University of Sydney conducted a review to find a suitable clinical terminology for use in Australian general practice computerised medical records. After consultation internationally with members of the World Organisation of Family Doctors, the Read codes 47 from the UK were identified as the only possibly suitable system available. The Read codes (described in Section 1.3) were subsequently trialled by Miller and Britt in Australian general practice in 1994 through a demonstration grant from the (then) Australian Department of Human Services and Health (the Aus-Read trial)

36 The final recommendation of the Project team was for the purchase of an Australian licence for the Read codes, on the proviso that an Australian edition of the Read codes be developed to account for differences in medical terminology between the UK and Australia (including differences in the meaning of some clinical terms). 61 This did not eventuate. The international distributors of the Read codes, Computer Aided Medical Services, would not permit the development of an Australian edition of Read codes to incorporate Australian medical synonyms and abbreviations. This was a major impediment to use of the Read codes in Australia and the licence was subsequently not purchased by the Australian government (personal communication, Dr. Graeme Miller, March 2011). The Australian Morbidity and Treatment Survey (AMTS) conducted in was a large-scale national study of general practice activity in Australia. During the course of this study 495 GPs recorded in free text 146,940 patient RFEs and 146,799 problems managed on structured paper encounter records. The free text records were subsequently classified to ICPC by trained secondary coders, most of whom had backgrounds in nursing or medical records. The ICPC book contained an alphabetical index, allowing the secondary coders to search for terms and their correct ICPC rubric code. However the AMTS demonstrated that the number of entries in the index was extremely deficient. Additional entries and cross-references were added to the index to help overcome these deficiencies. 62 After the AMTS, a quality assurance program for GPs (the Morbidity and Therapeutic Index) was developed. Together these studies resulted in over 500,000 encounter records secondarily classified to ICPC. 63 By 1995 it was apparent that a licence for the Read codes would not be purchased by the Australian Government, but Miller and Britt recognised an urgent need for a reliable coding system containing sufficient specificity for clinical use within GP computerised medical records. The (then) Family Medicine Research Unit (University of Sydney) therefore developed a general practice clinical terminology based on the expanded ICPC index developed during the AMTS and Morbidity and Therapeutic Index. This terminology contained greater specificity than previously available using ICPC alone, and 14

37 was classified to ICPC. Called ICPC PLUS, the system was released in 1995 and incorporated initially into one general practice clinical software program (personal communication, A. Prof. Helena Britt, March 2011) with subsequent uptake by other software vendors. In 1998 ICPC PLUS was updated to ICPC-2 PLUS to reflect the international update of the ICPC classification to Version 2 (ICPC-2). The link from the clinical terminology to ICPC was an important aspect of this work, to ensure that data entered using the terminology could be extracted for data reporting using an international standard classification. ICPC-2 PLUS is currently incorporated into 12 general practice software systems, with approximately 450 GP practices (incorporating 2,500 GP users), community health centres and Aboriginal medical services around Australia using the system. 64 It is notable that the need for a general practice clinical terminology containing more specificity than a classification was recognised widely, both in Australia and internationally. Some GP software vendors in Australia developed their own termsets or clinical terminologies in parallel to the development of ICPC PLUS. The Doctor Command Language (Docle) classification was designed in Australia for use within computerised medical records in the 1980s. Docle is based on alphabetical algorithms that are built to represent a medical term or concept. Docle was incorporated into a single general practice software system, Medical Director, in the mid 1990s. 65 In Medical Director, GPs use a termset called the Medical Director termset that can be linked to Docle. The other major GP software vendor, Best Practice, also developed its own termset (personal communication, participants at GP RefSet workshop [described in Chapter 4], November 2009). All GP software vendors therefore use proprietary termsets or clinical terminologies. As such, the clinical terminology used in Australian general practice EHRs is not standardised and semantic interoperability between GP EHR software is virtually non-existent. Internationally, a series of clinical terminologies classified to ICPC were also developed for use in general practice between 1994 and These included LOCAS (from French-speaking Belgium), Encode-FM (Canada), the 15

38 ICPC2-ICD10 Thesaurus (the Netherlands), the Belgian Bi-lingual Thesaurus (Belgium) and the Danish ICPC extension (Denmark). 66 Each of these terminologies were classified to ICPC, further demonstrating the importance of linking the terminology used in general practice to an international classification for standardised reporting. In 1999 the Australian General Practice Computing Group (GPCG) established a Coding Jury to evaluate coding systems available, and determine the best coding system for use in general practice in Australia. 67 The coding systems included in the evaluation were ICPC-2 PLUS, Docle, 68 SNOMED RT (all clinical terminologies) and ICD-10-AM (the Australian modification of the ICD-10 classification). In 2000, the Jury released its recommendations, concluding that at that time, there was no coding system available that was suitable for coding in Australian general practice. The Jury deemed ICD-10-AM the closest match to the requirements, and delivered a short term recommendation to implement ICD-10-AM as a terminology for general practice within five years of the report s release, with an additional recommendation that specific and essential general practice terms 67 be added to ICD-10-AM to address shortcomings in its content from a general practice perspective. ICD-10-AM is a classification, not a clinical terminology, and its content is not suitable for use as a clinical terminology. Given the inadequacies of ICD to adequately capture general practice content, recognised for nearly 50 years, this recommendation was unexpected and difficult to justify. 69 The Jury s long term recommendation was for the implementation of SNOMED CT (see Sections 1.3 and 3.3 for an overview of SNOMED CT). 67 However as stated previously, the version of SNOMED reviewed by the Jury was SNOMED RT, the version prior to SNOMED CT. The revision process between SNOMED RT and SNOMED CT was substantial and involved the merging of SNOMED RT with the UK National Health Service s (NHS) CTV SNOMED CT was not available until 2002, two years after the Jury was evaluating coding systems. As such the Jury had no way of evaluating the validity of SNOMED CT for Australian general practice. 16

39 After the Coding Jury recommendation the Australian National Centre for Classification on Health, custodians of ICD-10-AM, investigated improving the structure of the ICD-10-AM index to better meet the needs of general practice. 70 There was also an attempt by the University of Adelaide (funded by the now defunct General Practice Computing Group) to construct a general practice interface vocabulary to support the clinical terminology needs of Australian GPs (the Australian General Practice Vocabulary). 71 Neither ICD-10-AM, nor the GP vocabulary, were ever implemented in Australian general practice. Essentially, the short-term recommendations of the General Practice Coding Jury were abandoned because the recommended short-term solution was not appropriate in general practice. The release of SNOMED CT in 2002 provided a new option for a clinical terminology for general practice. In 2004 a business case was prepared in Australia for a national clinical terminology. The resulting recommendation was for SNOMED CT to be used in Australia as the core national clinical terminology, with the addition of other supporting terminologies for areas such as pathology results and medications. 72 The business case does not reference a substantial evidence base to support this decision. Instead, the decision appears to be primarily based on financial considerations, to leverage the international expertise used to develop SNOMED CT. The report did recommend enhancing SNOMED CT for Australian use, which was an acknowledgement that it may not suit the needs of Australian users. The National e-health Transition Authority (NEHTA) was formed in July 2005 as a partnership between the Australian national (federal) and state governments. 73 Chief Executives from federal and state health departments form the board of directors. NEHTA was given three years funding totalling $18.2 million from to develop timelines for the e-health agenda, develop e-health business cases, develop standards and support e-health implementations. 74 Australia first signed a licence for SNOMED CT in 2006, enabling the free use of SNOMED CT throughout Australia from July 1, NEHTA was given responsibility for the development and maintenance of SNOMED CT in 17

40 Australia. 75 In April 2007 Australia became one of the founding members of the IHTSDO. 53 NEHTA acts as Australia s official delegate to the IHTSDO and has a representative on the IHTSDO s Management Board which has overall oversight of SNOMED CT internationally. 1.5 The use of SNOMED CT-AU in Australian general practice There is a clear need for a reliable and accurate clinical terminology for use within EHRs throughout the Australian health care system, to promote the standardised recording and extraction of clinical terms from EHRs for the purpose of semantic interoperability. In Australia, SNOMED CT was purchased to fill this role. After the purchase of a SNOMED CT national licence, NEHTA released a version of SNOMED CT containing Australian content, known as SNOMED CT-AU. 76 The structure of SNOMED CT-AU and its relationship to SNOMED CT are described in Section However, published evaluations of SNOMED CT (or SNOMED CT-AU) in Australia are rare, as are reports of SNOMED CT implementations. In Australian general practice, there is one published SNOMED CT evaluation, involving 10 GPs. This study briefly trained GPs in the use of SNOMED CT, then asked GPs to use SNOMED CT for at least one day. 77 The results focus on the GPs views of SNOMED CT, and the study did not contain an evaluation of the extent to which SNOMED CT was fit for purpose in general practice. To my knowledge, SNOMED CT-AU has not yet been implemented in any Australian general practice EHRs. It is reasonable to assume that SNOMED CT-AU contains adequate content in the general practice clinical domain, given SNOMED CT s predecessors include a version of the Read codes which originated in general practice. However, the content gaps identified in the Aus-Read trial suggest that SNOMED CT-AU may be missing some content needed in Australian general practice. The extent to which the development of the local version of SNOMED CT (i.e. SNOMED CT-AU) has added content specific to Australian usage is also unknown. NEHTA has not published any information about the methods used to create SNOMED CT-AU. 18

41 In Australia, general practice is a clinical specialty with high levels of computerisation, and thus is a candidate for early implementation of SNOMED CT-AU. The process used to implement a new clinical terminology should include a thorough evaluation of the terminology content prior to implementation, to ensure that the terminology contains content that is accurate, relevant and suitable for clinical use. To date such an evaluation has not been conducted in relation to the implementation of SNOMED CT-AU in Australian general practice, and as a result it is uncertain whether SNOMED CT-AU is suitable for implementation in this setting from a content perspective. In this thesis, suitability will be measured by assessing the extent to which the clinical terms used by Australian GPs are included in SNOMED CT-AU, and assessing the range and scale of deficiencies identified. General practice EHRs are large databases containing many pieces of clinical information. The core data element needed for information transfer (and thus semantic interoperability) is that which is equivalent to problem/diagnosis. This thesis will address issues relating to the current use of clinical terminologies in Australian general practice (focussing on the problem/diagnosis data element) and assess the extent to which SNOMED CT-AU is fit for purpose for Australian general practice. The overall aims of the thesis are described in Chapter 2. Chapter 3 contains an overview of the structure and content of each of the resources used in the thesis. In Chapter 4 I investigate the issues faced by GPs and/or software vendors when using clinical terminologies. Themes identified in Chapter 4 form the basis of Chapter 5, in which specific issues relating to clinical terminology use are investigated using the Bettering the Evaluation and Care of Health (BEACH) study of general practice, in which clinical terms are recorded in free text by GPs and coded using ICPC-2 PLUS. Chapter 6 assesses the extent to which the terms from ICPC-2 PLUS can be mapped to SNOMED CT-AU. Discussion and conclusions are presented in Chapters 7 and 8. 19

42 2 Aims of this thesis and the candidate s contribution 2.1 Aims The principal aim of this thesis is to determine the extent to which the content in SNOMED CT-AU is suitable for use in Australian general practice. More specifically, this thesis aims to: identify issues affecting the present and future use of standardised clinical terminologies in Australian general practice, and the requirements for their inclusion in general practice EHRs. identify patterns in the current use of a clinical terminology currently used in Australian general practice in two commonly used data elements reasons for encounter and problems managed. determine the extent to which SNOMED CT-AU contains the terms used by Australian GPs to describe symptoms and complaints, and diagnoses and diseases. identify deficiencies in SNOMED CT-AU content in its reflection of general practice content. 2.2 Candidate s contribution The candidate was fully involved in all aspects of this thesis, including conceptualising the thesis concept, developing the aims of the thesis, planning, designing and undertaking the research. The candidate undertook a comprehensive literature review involving the selection of databases, libraries, and other sources of information, including standards documents. The candidate managed all aspects of the workshops held with GPs and software vendors, including selection and recruitment of participants, preparation of the presentations for the workshops, presentation of sections of the workshops (in conjunction with Prof. Graeme Miller as supervisor), and collation of the workshop findings. The candidate wrote the requirements and prepared the General Practice Reference Set (GPRS) requirements 20

43 document (included as Appendix 1), consulting with NEHTA staff to ensure the document adhered to the NEHTA document template and NEHTA specifications. After the workshops, the candidate realised that the success of any future implementation of SNOMED CT-AU in general practice was heavily reliant on ensuring that current patterns of terminology use in Australian general practice were considered. The candidate subsequently conceptualised the substudies reported in Chapter 5 to research this issue. The candidate planned and initiated the collection of the free text from the BEACH program described in Chapter 5. Under the candidate s instruction, free text was entered by secondary coders employed by the Family Medicine Research Centre (FMRC). Subsequent to data entry the candidate analysed the free text entries, assisted by a graduate research assistant, Ms Carmen Zhang. Analyses of BEACH data in Sections 5.3 and 5.4 were performed by data analysts employed by the FMRC, Ms Lisa Valenti and Mr Christopher Harrison, under instruction from and direction of the candidate. The candidate independently conducted all analyses described in Sections 5.2 and 5.5. The candidate was responsible for undertaking all mapping activities described in Chapter 6 for the mapping from ICPC-2 PLUS to SNOMED CT-AU. The candidate called on the clinical expertise of A/Prof Graeme Miller to assist in the mapping process where there was uncertainty about the appropriate mapping result. The preparation of this thesis document was entirely the work of the candidate, including the background and literature review, reporting and interpretation of results, the discussion and conclusions made. 2.3 Candidate s involvement in research leading to this thesis The research reported in this thesis was conducted in association with the BEACH study of general practice, a continuous national study of general practice activity in Australia. The candidate completed her Honours thesis 21

44 using BEACH data (Aged patient encounters in the general practice setting 78 ) at the FMRC in 2002, and has been employed at the FMRC since this time. The candidate s primary work role surrounds the application of ICPC-2 and ICPC-2 PLUS in Australia, involving the licensing and education of software developers and users in Australia, updating the ICPC-2 PLUS terminology (quarterly) and preparing ICPC-2 PLUS terms for use in the BEACH study. The candidate also undertook research to classify the ICPC-2 rubrics according to chronicity 79 and develop a set of data elements for the general practice electronic health record and data query minimum data set. 80 In 2005 the candidate performed the manual mapping of a selection of terms from ICPC-2 PLUS to SNOMED CT, as part of the preparation of the business case for the Australian purchase of a SNOMED CT licence. In , the candidate supervised a trial of the development of an automated mapping tool and process to create a map from ICPC-2 PLUS to SNOMED CT, undertaken as a summer school activity between the Centre and the School of Information Technology at the University of Sydney. 81 The purchase of an Australian licence for SNOMED CT in 2006 highlighted the need to evaluate the suitability of SNOMED CT for use in various health settings in Australia. The need to develop change mechanisms for the implementation of SNOMED CT also became evident, including maps from legacy termsets/terminologies such as ICPC-2 PLUS to SNOMED CT. Following discussions with the Director and Medical Director of the FMRC, the candidate decided to pursue the investigation of an Australian general practice SNOMED CT reference set (then known as a subset). This work forms the basis of this thesis. In 2006 the candidate became an associate member of the Wonca International Classification Committee (WICC), and became a full member of WICC in In 2008 the candidate became a member of the IHTSDO s Primary Care Special Interest Group (now known as the International General Practice/Family Physician Special Interest Group). The candidate is currently the Project Manager for an IHTSDO project to create an international general/family practice SNOMED CT reference set, and a map from this 22

45 reference set to ICPC-2. This project began in 2010 and its completion is expected in

46 3 Resources used in this thesis Three resources will be used throughout this thesis. Two of these are terminology products: the Australian ICPC-2 PLUS general practice terminology, and the SNOMED CT-AU reference terminology. The third resource is data from the BEACH study of general practice activity. ICPC-2 PLUS is used to enter data into most of the data elements in the BEACH research database and is used in Australian general practice electronic health records. Backgrounds for each of these resources have been described in detail in Chapter 2. This chapter describes the structure, content and current status of each of the sources. Detailed methods describing how each source was applied in different parts of this thesis are included in each results chapter. 3.1 Bettering the Evaluation and Care of Health (BEACH) Methods used in the BEACH study are detailed extensively in each report of annual results from the study Information provided in Section 3.1 has been summarised from these reports. The candidate has been a co-author of these reports since In summary, BEACH is a cross-sectional national study of general practice activity in Australia. BEACH began in April 1998 and has been running continuously since then. Every year approximately 1,000 GPs from around Australia are randomly selected for participation. Each GP participant records details about 100 consecutive patient encounters. These details are recorded in free text on structured paper forms Glossary of terms used in BEACH Box 3.1 contains a glossary of terms from the BEACH study that are used in this thesis. 24

47 Box 3.1: Glossary of terms from the BEACH study used in this thesis Construct Consultation Diagnosis/problem Encounter Definition See Encounter A statement of the provider s understanding of a health problem presented by a patient, family or community. GPs are instructed to record at the most specific level possible from the information available at the time. It may be limited to the level of symptoms. Any professional interchange between a patient and a GP. GP registrar A medical practitioner who is currently enrolled in a general practice training program. 94 Problem managed Reason(s) for encounter (RFEs) Significant Symptom/ complaint Vocationally registered See Diagnosis/problem The subjective reason(s) given by the patient for seeing or contacting the general practitioner. These can be expressed in terms of symptoms, diagnoses or the need for a service. This term is used to refer to a statistically significant result. Statistical significance is defined at the 95% confidence level. Any subjective evidence of a health problem as perceived by the patient. Registration of a medical practitioner as a general practitioner by Medicare Australia, certifying that the practitioner is suitably qualified to work in general practice, and primarily works in general practice. 95 Note: Definitions have been sourced from the report General Practice Activity in Australia unless otherwise specified Governance and funding The BEACH program is conducted by the FMRC at the University of Sydney. The BEACH study is financially supported through a consortium of government departments, pharmaceutical companies and other organisations. Over the course of this study, the supporting organisations included: Australian Government Department of Health and Ageing Australian Government Department of Veterans Affairs Australian Institute of Health and Welfare The Office of the Australian Safety and Compensation Council, Department of Employment and Workplace Relations Abbott Australasia AstraZeneca Pty Ltd (Australia) Bayer Australia Ltd CSL Biotherapies Pty Ltd GlaxoSmithKline Australia Pty Ltd 25

48 Janssen-Cilag Pty Ltd Merck, Sharp and Dohme (Australia) Pty Ltd National Prescribing Service Ltd Novartis Pharmaceuticals Australia Pty Ltd Pfizer Australia Roche Products Pty Ltd Sanofi-Aventis Australia Pty Ltd Wyeth Australia Pty Ltd. Each financial stakeholder is represented on the BEACH Advisory Board, together with representatives from professional organisations such as the RACGP, the AMA, the Australian College of Rural and Remote Medicine (ACRRM) and the Consumers Health Forum. The Advisory Board meets three times per year to discuss the governance of the study Ethics The BEACH study was conducted in collaboration with the Australian Institute of Health and Welfare (AIHW) under the AIHW Act between April 1998 and March The BEACH study received ethics approval from the Ethics Committee of the AIHW and the Human Research Ethics Committee of the University of Sydney GP sampling and recruitment Each year, a random sample of GPs is drawn from all vocationally registered GPs and GP registrars who have conducted a predetermined number of GP consultations. The sample is provided by the Australian Government Department of Health and Ageing and is updated every three months. Each randomly selected GP is approached to participate in the study first by letter, and subsequently by telephone. A recording date is set several weeks in advance for GPs who choose to participate, and a research pack is sent to the participant with instructions to start on the date agreed during the telephone call. Non-returns are followed up by regular telephone calls. 26

49 GPs who complete the study earn Clinical Audit points from the RACGP or from ACRRM towards their quality improvement requirements Data elements collected in BEACH There are three linked data collection processes in the BEACH study: GP characteristics collected through the GP Profile, encounter data, and patient based health information. Detailed methods for each of these collections, including data elements collected, are provided in each annual report of BEACH results. 96 Figure 3.1 presents the BEACH relational database and shows the relationships between the data collections. The figure shows that: all data elements are directly related to the GP characteristics, patient characteristics and the encounter. reasons for encounter (RFEs) are indirectly related to problems managed. For example, a patient may have a single RFE that relates to multiple problems, or multiple RFEs that relate to a single problem. all management types are directly related to the problem managed. This thesis uses only the encounter data collected in BEACH. GPs are instructed to record information only about the current encounter on the encounter form, creating a snapshot of information. Data about medical conditions not discussed during the current encounter are not captured in BEACH. An example of the BEACH encounter form is included as Appendix 2. This thesis contains information about two types of encounter data collected: reasons for encounter (RFEs) and problems managed (highlighted in blue in Figure 3.1). 27

50 GP characteristics age and sex years in general practice country of graduation number of sessions per week FRACGP status (yes/no) currently a registrar (yes/no) Practice characteristics The encounter date direct (face to face) Medicare Benefits Schedule item number(s) claimable workers compensation other paid no charge indirect (e.g. telephone) Patient substudies (SAND) practice size (FTE GPs) postcode accreditation status after-hours arrangements teaching practice (yes/no) The patient age and sex practice status (new/old) Concession card status DVA status postcode of residence NESB/Indigenous status reasons for encounter risk factors body mass smoking status alcohol consumption other topics Figure 3.1: The BEACH relational database Problems managed diagnosis/problem label problem status (new/old) work-related problem status Management of each problem Medications (up to four per problem) prescribed over-the-counter advised provided by GP drug class drug group generic brand name strength regimen number of repeats drug status (new/continued) Other treatments (up to two per problem) procedural treatments clinical treatments (e.g. advice, counselling) practice nurse involvement Pathology tests ordered (up to five per encounter) individual tests (e.g. glucose test) or batteries of tests (e.g. lipid profile) Other management referrals (up to two per encounter) to specialists to allied health professionals to emergency departments hospital admissions imaging ordered (up to three per encounter) Note: FTE full-time equivalent; FRACGP Fellow of the Royal Australian College of General practitioners; DVA Department of Veterans Affairs; NESB non-english-speaking background; SAND Supplementary Analysis of Nominated Data. RFEs are defined as the subjective reason(s) given by the patient for seeing or contacting the general practitioner. These can be expressed in terms of symptoms, diagnoses or the need for a service. 15 At each BEACH encounter, 28

51 GPs are instructed to record at least one and up to three RFEs in the patient s own words. Problems managed are defined as a statement of the provider s understanding of a health problem presented by a patient, family or community. 15 GPs are instructed to record at the most specific level possible from the information available at the time, which may be limited to the level of symptoms. If more than four problems are managed at an encounter, GPs are asked to record the four problems that best describe the breadth of the consultation Data entry On submission of the data to the FMRC, the paper forms are entered into a Microsoft Access 2003 database by trained secondary clinical coders. The clinical coders are primarily undergraduate students from a health sciences background, including health informatics, psychology or medical science. To facilitate ease and speed of data entry, two coding systems are used ICPC-2 PLUS and a pharmaceutical coding system called the Coding Atlas for Pharmaceutical Substances. Quality assurance of the coded data is undertaken to ensure that the coded data are representative of what was written on the encounter forms. All quality assurance processes are ongoing, and their development continues over time. Computer locks are included in the BEACH database to avoid inappropriate data entry. For example, there is a lock that prevents terms related to pregnancy being entered for a male patient. Secondary coders undergo a rigorous training program, and are regularly audited to check the entry of coded data against information on the encounter form. Coders are also instructed that if they cannot find an appropriate term in the database they should manually tag the encounter form. Tagged forms are then reviewed by the coding supervisor. Further quality assurance checks are undertaken using the statistical software used in the program during analysis Representativeness After the conclusion of each BEACH year the age and sex distribution of the final participating sample of GPs in BEACH is compared with the age sex 29

52 distribution of all GPs in Australia included in the sample frame, to determine the extent to which the sampled GPs are representative of the sample frame population. If participants from particular age or sex groups are over-represented or under-represented, statistical weights are applied to the data to attain comparable estimates and precision. There may also be variation in GP activity levels due to the BEACH sampling method allowing the inclusion of both full-time and part-time GPs. Additional weighting may be applied to the encounter data based on each GPs activity level, proportional to the number of general practice service items claimed by each GP in the previous 12 months. The raw and weighted datasets are compared in each annual report of BEACH data. In most years, the differences between the datasets are minimal. However, weights cannot be applied when multiple years of data are merged. Results reported in this thesis come from data merged across six years of the BEACH study, from to , and As such, there was no weighting applied to data reported in this thesis. However, increasing the sample size from 1,000 GPs (in one year) to 6,000 GPs and 600,000 encounters increases the statistical power enormously and negates the need for such weighting Statistical methods used in BEACH In analyses of BEACH data the primary point of reference is the encounter. Proportions are only used to describe the distribution of an event that can occur once only during an encounter, or to describe the distribution of events within a larger group of the same events (e.g. frequency of a specified problem as a percentage of all problems). Rates per 100 encounters are provided when an event can occur more than once at an encounter Current status of BEACH BEACH is an ongoing study which began in April The 14 th year of data collection was completed in March

53 3.2 ICPC-2 PLUS ICPC-2 PLUS 97 is a clinical terminology classified to the International Classification of Primary Care, Version 2 (ICPC-2). 40 The history of ICPC-2 PLUS has been detailed in Chapter Glossary of terms used for ICPC-2 and ICPC-2 PLUS Box 3.2 contains definitions for the terms used in this thesis when referring to ICPC-2, and/or ICPC-2 PLUS. Box 3.2: Glossary of terms used in conjunction with ICPC-2 and/or ICPC-2 PLUS Construct Chapter Component Code International Classification of Primary Care, Version 2 (ICPC-2) ICPC-2 code ICPC-2 PLUS code ICPC-2 PLUS Keyword ICPC-2 PLUS term Process code Rubric Status Definition One of the two axes forming the main divisions within ICPC-2. The 17 chapters are primarily based on body systems, with additional chapters for psychological and social problems. Chapters are identified through the use of a single alphabetic character in an ICPC-2 code, most of which are mnemonics (e.g. the respiratory chapter is represented by the letter R). The second of the main divisions in ICPC-2. There are seven components that are identical in each ICPC-2 chapter. A fixed sequence of signs or symbols, alphabetical or numerical characters designating an object, concept or term. ICPC-2 code: A three digit alpha-numeric sequence that represents a concept (or group of concepts) in the ICPC-2 classification. The code contains one alphabetic character and two numeric characters. The combination of these characters represents the code s position in the classification. ICPC-2 PLUS code: A six digit alpha-numeric sequence that represents a term in the ICPC-2 PLUS terminology. Each code has two components: the ICPC-2 code and the term code or PLUS code. A clinical and statistical classification of the World Organization of Family Doctors (Wonca) developed and maintained by the Wonca International Classification Committee (WICC). ICPC-2 represents the clinical workload of general practitioners. See Code See Code Words or abbreviations up to 10 characters in length. Keywords create the search mechanism in ICPC-2 PLUS. They are linked to, and used to access ICPC-2 PLUS terms. A word or phrase that encompasses a clinical description used in general practice. Terms have a maximum of 30 characters. An ICPC-2 or ICPC-2 PLUS code included in one of Components 2, 3, 4, 5 or 6 and represented as a code from 30 to 69. The descriptor (or label) of the clinical concept (or group of concepts) that is represented by an ICPC-2 code. An indicator of whether the term is valid for use in data entry. In ICPC-2 PLUS, there are two statuses: Active: indicates the term is current and available for data entry. Inactive: indicates the term has been retired, and is no longer valid for data entry. Terms marked as inactive must be retained in the medical record for historical data reporting. Note: Definitions have been sourced from the ICPC-2 PLUS functionality requirements 98 and the Wonca Dictionary of General/Family Practice

54 3.2.2 ICPC-2 ICPC-2 40 provides the structure for ICPC-2 PLUS. ICPC-2 is owned by Wonca and developed and maintained by the Wonca International Classification Committee (WICC). The first version of ICPC (ICPC-1) was released in 1987, and the second version (ICPC-2) published in ICPC-2 is a biaxial classification, as shown in Figure 3.2. On one axis are 17 chapters. Fourteen of these chapters are based on body systems, with additional chapters for psychological problems (Chapter P), social problems (Chapter Z) and general problems that may be unspecified, or relate to multiple body systems (Chapter A). Each chapter is identified by a single alphabetic character. Where possible, the alphabetic character represents a mnemonic to aid recall. Components A B D F H K L N P R S T U W X Y Z 1. Symptoms, complaints 2. Diagnostic, screening, prevention 3. Treatment, procedures, medication 4. Test results 5. Administrative 6. Other 7. Diagnoses, disease A General & unspecified L Musculoskeletal U Urinary B Blood, blood-forming organs N Neurological W Pregnancy, childbearing, D Digestive P Psychological family planning F Eye R Respiratory X Female genital H Ear S Skin Y Male genital K Circulatory T Metabolic, endocrine, nutritional Z Social Component number Code range Component name 1 01 to 29 Symptoms & complaints 2 30 to 49 Diagnostic and preventive procedures 3 50 to 59 Medication, treatment, therapeutic procedures 4 60 to 61 Results 5 62 Administrative 6 63 to 69 Referrals and other reasons for encounter 7 70 to 99 Diagnoses and diseases Note: To represent a code, the symbol is replaced by an alphabetic chapter character. Figure 3.2: The structure of the International Classification of Primary Care Version 2 (ICPC-2) 32

55 The second axis represents the components of the classification, listed in Figure 3.2 from 1 to 7. These numbers are not overtly shown in the structure of the resulting ICPC-2 code. Each component is represented by a range of two numeric characters that are standard across chapters. Due to lack of space within the biaxial structure of ICPC-2, some clinical conditions could not be placed within their correct component and were mis-classified. For example, the rubric Skin infection, post-traumatic has a code of S11. As an infection, this rubric should be placed in the Diagnoses and diseases component (Component 7) in the skin (S) chapter, and have a code within the range S70 to S99. However, this component contains 30 rubrics, the maximum number of rubrics available for Component 7, so Skin infection, post traumatic was placed in the Symptoms and complaints component of Chapter S and given code S11. In the absence of a solution to overcome the issues around component mis-classification until the next version of ICPC (ICPC-3), WICC developed the ICPC-2 pager, a two page summary of the ICPC-2 rubrics, with each rubric colour-coded according to its correct component in ICPC-2. A colour-coded copy of the ICPC-2 pager is included as Appendix 3. These corrected components have been applied in all analyses undertaken in this thesis ICPC-2 PLUS As outlined in Chapter 2, ICPC-2 PLUS is a clinical terminology designed for use in Australian general practice EHRs and for secondary data entry in general practice research programs such as BEACH. ICPC-2 PLUS can also be used as an index to the ICPC-2 classification, to ensure that terms are correctly classified in ICPC-2. Structure of ICPC-2 PLUS codes Codes in ICPC-2 PLUS retain the biaxial structure of ICPC-2. Each term within ICPC-2 PLUS is classified to an ICPC-2 code, and the associated ICPC-2 code is represented in the ICPC-2 PLUS code as the first three digits, as demonstrated in Figure 3.3 (reproduced from the ICPC-2 PLUS functionality requirements for developers 98 ). The second part of the 33

56 ICPC-2 PLUS code is an automatically generated three digit number with no inherent meaning, allocated as new terms are added. A The first 3 characters are the ICPC-2 code. The alphabetic part (A) indicates the chapter of the ICPC-2 code, while the numeric part (86) identifies the location of the code within the chapter. The last three digits comprise the plus part of the ICPC-2 PLUS code. As each plus code is entered, it is assigned the next available 3 digit number in the rubric. As such, there is no meaning to the order of these codes. Figure 3.3: Structure and meaning of an ICPC-2 PLUS code Structure of ICPC-2 PLUS terms ICPC-2 PLUS terms are a maximum of 30 characters long. They are structured according to one of the following two principles: 1. Common usage expression (e.g. Restless legs syndrome ) 2. Problem/procedure;type;attribute (e.g. Pain;musculoskeletal;leg). In the second structural variation, each part of the ICPC-2 PLUS term is separated by semicolons. Type specifies the nature of the problem. Attribute is an optional inclusion in the term structure, and specifies a characteristic of the problem. Attributes may include chronicity (e.g. acute or chronic) or the site of a problem (e.g. arm or leg). For example, the structured ICPC-2 PLUS term for acute myocardial infarction is Infarction;myocardial;acute. This structure was developed to aid the user when searching for ICPC-2 PLUS terms, as all terms are uniformly structured and terms containing the same root (problem/procedure) word can be seen together in a picklist. Software vendors who incorporate ICPC-2 PLUS in their general practice EHRs are instructed that picklists must be presented to the user in alphabetical order to ensure this grouping occurs. 34

57 Functionality of ICPC-2 PLUS The functionality of ICPC-2 PLUS is outlined in a set of functionality requirements, which are provided to all ICPC-2 PLUS Developers (that is, any GP EHR vendor who installs ICPC-2 PLUS). Users access ICPC-2 PLUS through a keyword system with the keyword acting as the interface for users to access the terminology. ICPC-2 PLUS keywords are words or word acronyms (e.g. AMI for acute myocardial infarction) of up to ten characters in length. Any words containing more than ten characters are abbreviated to ten characters (e.g. international would be abbreviated to internatio ). Every word contained in an ICPC-2 PLUS term is included as a keyword linked to the term. In addition, any expressions that are synonymous with, or logically related to the term are also included as keywords, for example the word neoplasm is available as a keyword linked to every term containing the word tumour. Figure 3.4 shows an example of the relationship between keywords and terms. Keyword link Term DIAB link Advice/education;diabetes link Check up;diabetes link Diabetes;Type 1 link Diabetes;Type 2 Figure 3.4: Example of the representation of links between keywords and terms After entering a keyword, users are presented with a picklist of the associated ICPC-2 PLUS terms from which they select the most appropriate term. Both the term and its code are then saved into the chosen data field in the record ICPC-2 PLUS maintenance ICPC-2 PLUS is updated quarterly and released to ICPC-2 PLUS Developers in January, April, July and October each year. Changes made to ICPC-2 PLUS include: 35

58 additions of new terms and keywords movement of ICPC-2 PLUS terms from one ICPC-2 code to another modifications to ICPC-2 PLUS term descriptions retirement of terms that are no longer valid for inclusion in ICPC-2 PLUS. Identification of suggestions Suggestions for ICPC-2 PLUS come from multiple sources. As described in Section 3.1.6, one of the quality checks undertaken during BEACH is for coders to tag encounter forms containing free text for which they are unable to find an appropriate term in the terminology. The tagged forms are reviewed by the coding supervisor, who makes suggestions for new terms if an appropriate match is not found. Suggestions for new terms are also made by GPs who use the ICPC-2 PLUS terminology in their EHRs. Processing suggestions All suggestions for new content in ICPC-2 PLUS are considered by the FMRC Classification Committee, consisting of staff at the Centre who are involved in coding and classification, including the Director, Medical Director, Classifications Manager (the candidate), BEACH Project Manager and Coding Supervisor. The FMRC Classification Committee considers the suggestions from multiple angles. The following questions are asked: How frequent (or rare) is the problem/procedure in general practice? If the suggested term is rare, is it important from a public health perspective? Is this suggestion a new concept, or reflecting a new way of representing something that already exists within the terminology? Can the suggestion be reflected in the terminology using keyword links to an existing term, rather than adding the term itself? 36

59 How and where does the suggestion fit within the ICPC-2 classification? ICPC-2 PLUS is a general practice based terminology, and terms included in the terminology should have their foundation in general practice. If a suggestion is very specific to a medical specialty that is not general practice, careful consideration is given as to whether or not to include the term. Decisions made during the FMRC Classification Committee meeting are formally recorded, with reasons provided for each decision. Individuals or organisations that make suggestions are sent a summary of their suggestions made, with the Classification Committee s decision for each suggestion. If a suggestion is rejected, the reason(s) is provided. The requestor can then re-submit a suggestion for the next release if they can provide further justification for the suggestion. Approved suggestions are added to the ICPC-2 PLUS maintenance database. Comma separated value (CSV) files are generated and sent to ICPC-2 PLUS Developers for inclusion in their EHR product, and subsequent distribution to end users Relationship between ICPC-2 PLUS and BEACH The inter-relationship between the development of the ICPC-2 PLUS terminology and its use in the BEACH study is unique. In general, terminology developers are not involved in the collection of data using their terminology, and thus are not directly affected by the impact of terminology changes in the data collection process Current status of ICPC-2 PLUS, April 2011 Table 3.1 contains the number of terms available in each chapter, component and chapter/component of ICPC-2 PLUS in the April 2011 release. These will form the denominator for results presented in Chapter 6. 37

60 Table 3.1: Number of ICPC-2 PLUS terms by ICPC-2 chapter and component, April 2011 ICPC-2 chapter ICPC-2 component A B D F H K L N P R S T U W X Y Z Total Component 1 (symptoms and complaints) Components 2 6 (processes of care) Component 7 (diagnoses and diseases) , , N/A 3,288 Total 1, ,928 Note: N/A not applicable. 38

61 3.3 SNOMED CT Information provided in this section has been derived from the SNOMED CT Technical Reference Guide 99 and the SNOMED CT Technical Implementation Guide. 100 In summary, SNOMED CT is a structured clinical reference terminology designed for use in electronic health records. Unique clinical concepts form the basis of SNOMED CT. Each clinical concept is linked to other concepts through multi-hierarchical relationships. Each concept is also linked to one or more descriptions, each of which describes the concept s meaning in an alternate way SNOMED CT glossary Box 3.3 below is a glossary of the terms used in relation to SNOMED CT. These definitions will be used throughout this thesis. Box 3.3: Glossary of terms used in conjunction with SNOMED CT and/or SNOMED CT-AU Construct Concept Concept equivalence Concept identifier (Concept ID) Cross mapping Description Description logic Expression Extension Hierarchy Definition A clinical idea to which a unique Concept ID (i.e. a concept identifier) has been assigned. Equivalence is the state of two SNOMED CT concept codes or postcoordinated expressions having the same meaning. Concept equivalence can occur when a postcoordinated expression has the same meaning as a precoordinated concept code; or when two different postcoordinated expressions have the same meaning. A permanent, unique, numeric identifier used to represent a SNOMED CT concept. A cross map is a reference from a concept code to a cross map target. A description associates a human-readable term with a concept that it describes. A concept is associated with several descriptions. Each of these represents either a preferred term, synonym or fully specified name for the concept in a particular language or dialect. A formalisation of the characteristics of concepts that are always true for each instance of a concept, and differentiate concepts from one another. A structured combination of one or more concept identifiers used to express an instance of a clinical idea. A data table or set of data tables that is created in accordance with the structures and authoring guidelines applicable to SNOMED CT. An extension is ordinarily edited, maintained and distributed by an organisation other than the IHTSDO. An ordered organisation of concept codes linked together through is a relationships. Concept codes are linked to their more general parent concept codes directly above them in a hierarchy. Concept codes with more general meanings are usually presented as being at the top of the hierarchy and then at each level down the hierarchy code meanings become increasingly more specific or specialised. (continued) 39

62 Box 3.3 (continued): Glossary of terms used in conjunction with SNOMED CT Construct IHTSDO International release Is a Member/ member country Morphologic abnormality National Release Centre Postcoordinated expression Precoordinated expression Preferred term Reference set (RefSet) Relationship Release format Release format 1 Release format 2 Semantic tag SNOMED CT identifier Subset Subtype (or child) Supertype (or parent) Synonym Definition The International Health Terminology Standards Development Organisation (IHTSDO) is a not-for-profit association that develops and promotes use of SNOMED CT to support safe and effective health information exchange. The core set of SNOMED CT data files supplied by the International Health Terminology Standards Development Organisation, released twice a year in January and July. The relationship type that defined a supertype subtype relationship between two concepts. Usually expressed as subtype is a supertype. A Member of the International Health Terminology Standards Development Organisation (IHTSDO) in accordance with the IHTSDO Articles of Association. Alterations from normal body structures. The organisation within an IHTSDO Member country that is responsible for maintaining and releasing SNOMED CT content including any national extensions of SNOMED CT. Representation of a clinical meaning using a combination of two or more concept identifiers is referred to as postcoordination. Representation of a clinical meaning using a single concept identifier is referred to as precoordination. The term that is deemed to be the most clinically appropriate way of expressing a concept in a clinical record. The preferred term varies according to language and dialect. Reference sets represent groups of components that share specified characteristics that affect the ways the components are displayed or otherwise accessible within a particular realm, specialty, application or context. A relationship represents an association between two concepts. Each concept in SNOMED CT is logically defined through its relationships to other concepts. A file structure specified by the IHTSDO for files used to distribute SNOMED CT content. The file structure specified for the files used to distribute SNOMED CT content in This file structure was used until The file structure specified by the IHTSDO for files used to distribute SNOMED CT content from The semantic category to which the concept belongs (e.g. clinical finding, disorder, procedure, organism, person etc. Also known as a hierarchy tag. A unique integer identifier applied to each SNOMED CT component (concept, description, relationship, subset, etc). A group of components (e.g. concepts, descriptions or relationships) that share a specified common characteristic or common type of characteristic. Subsets represent information that affects the way the components are displayed or otherwise accessible within a particular realm, specialty, application or context. A specialisation of a concept, sharing all the definitional attributes of the parent concept, with additional defining characteristics Subtype is sometimes used to refer to the concepts in a hierarchy that are directly related to a parent concept via the is a relationship. A concept that is the target of a direct is a subtype relationship from a specified concept. A term that is an acceptable alternative to the Preferred term as a way of expressing a concept. Note: Definitions have been sourced from the SNOMED CT Technical Implementation Guide

63 3.3.2 Scope and structure SNOMED CT includes concepts from both human and veterinary medicine, and contains clinical concepts that represent varying levels of specificity. SNOMED CT content is divided into 18 top level hierarchies, as shown in Box 3.4. Box 3.4: Top level hierarchies in SNOMED CT Body structure Pharmaceutical/biologic product Situation with explicit context Clinical finding Physical force Social context Environment/geographical location Physical object Special concept Event Procedure Specimen Observable entity Qualifier value Staging and scales Organism Record artefact Substance SNOMED CT is based on description logic, formalising the essential characteristics of concepts, that is, those characteristics that are always and necessarily true, and that serve to differentiate concepts from each other. 101 Description logic facilitates the definition of different components of SNOMED CT concepts from a computational perspective. It also provides the extensibility mechanism inherent in SNOMED CT by facilitating permissible linkages, both between different concepts (e.g. the hierarchical parent child relationship between Diabetes mellitus and Type 1 Diabetes Mellitus ), and between individual concepts and their attributes (e.g. Asthma can be linked to severity levels, but cannot be linked to laterality). There are three primary constructs in SNOMED CT: concepts, descriptions and relationships. These three constructs constitute the SNOMED CT core data tables. Concepts and descriptions In SNOMED CT, concepts are defined as unique clinical ideas. Box 3.5 provides an example of the characteristics of a SNOMED CT concept. SNOMED CT concepts are represented by a unique identifier called the SNOMED CT ID or Concept ID, a numeric code of variable character length. The Concept ID has no meaning and cannot be used to establish a concept s position in a SNOMED CT hierarchy. 41

64 Each concept has an unambiguous label called the Fully specified name. At the end of the fully specified name there is a word in parentheses called the hierarchy tag or semantic tag describing the location of the concept in the SNOMED CT hierarchical structure. Concepts are also represented by a number of descriptions, including a single preferred term and one or more synonyms (where applicable). The preferred term represents the form of the concept as it should appear in clinical use. There is only one preferred term for each SNOMED CT concept. Synonyms represent alternate ways of expressing a concept. The example in Box 3.5 demonstrates that acronyms can also be included in synonym labels. There is no limit placed on the number of synonyms that can be linked to a SNOMED CT concept. Identical preferred terms or synonyms may be used in different concepts. The Concept ID and fully specified name are the only two elements that uniquely represent a concept. Box 3.5: Example of SNOMED CT concept and its constituent parts SNOMED CT concept identifier: Fully specified name: Preferred term: Synonyms: Chronic renal impairment (disorder) Chronic renal impairment Chronic renal disease CKD - chronic kidney disease Chronic kidney disease Relationships In SNOMED CT, relationships provide a linkage mechanism. There are two types of relationships in SNOMED CT: hierarchical relationships and defining attribute relationships. The most common relationship type in SNOMED CT is an is a relationship, otherwise called a parent child or a supertype subtype relationship. This type of relationship defines the structure of SNOMED CT by explicitly 42

65 representing the hierarchical relationship that exists between two concepts, and only occurs when all instances of the relationship are always valid. For example, Diabetes mellitus is a Disorder of glucose metabolism. It is not feasible in a practical sense to explicitly include all possible combinations of concepts in a clinical terminology. For example, the explicit inclusion of all variants of severity (e.g. mild, moderate or severe) in SNOMED CT concepts in which it is appropriate to record severity would lead to a fourfold increase in the number of such concepts. For example, a single concept Asthma would be accompanied by additional concepts for Mild asthma, Moderate asthma and Severe asthma. In clinical terminology, the explicit inclusion of attributes such as severity within concept labels is called precoordination. To overcome this problem, the description logics within SNOMED CT allow such attributes to be linked to concepts in an alternate way. There is a set of relationships in SNOMED CT called defining attribute relationships that outline a concept s defining characteristics, such as finding site, morphology and causative agent. These relationships also contain SNOMED CT concept identifiers. These relationships form the basis of the postcoordination functionality in SNOMED CT, allowing concepts containing a single clinical construct (called kernel concepts) to be linked to their defining attributes using a combination of two or more concept identifiers. The combination of two or more concept identifiers to represent a clinical meaning is referred to as a postcoordinated expression. At present, precoordinated concepts are included in SNOMED CT. However, the IHTSDO advocates a move away from the inclusion of precoordinated concepts towards a model that focuses on postcoordination and plans to slowly remove precoordinated content from SNOMED CT, although a timeline has not been finalised for this work Additional functionality Between 2002 and 2011 SNOMED CT was released as a set of files using a format known as Release Format 1 or RF1. Over time, a number of technical deficiencies in this format were identified (as outlined in the 43

66 SNOMED CT Technical Implementation Guide 100 ). A new release format, known as Release Format 2 or RF2 was released in All information contained in RF1 was retained in RF2, ensuring compatibility of the two versions. The updated format reduces areas of ambiguity identified in RF1 and provides SNOMED CT with greater flexibility. Implementation guidance is available in the SNOMED CT Technical Implementation Guide however the information provided is guidance only. The IHTSDO does not impose rules on the implementation of SNOMED CT. SNOMED CT can be implemented using a variety of methods, and at many levels from a SNOMED CT enabled browser to full-scale clinical implementation in an electronic health record. Similarly, the IHTSDO does not publish rules outlining how SNOMED CT concepts should be displayed in a user interface. The IHTSDO provides software developers with a toolkit to aid the implementation of SNOMED CT. The toolkit contains extra data tables to aid text based searches of SNOMED CT content. In SNOMED CT, concepts are retired, or made inactive, but never deleted. If the meaning of a concept changes the original concept is made inactive, a new concept is created to represent the new meaning and appropriate linkages made between the inactive concept and the new concept. In SNOMED CT active concepts are primarily known as current concepts. The functionality of SNOMED CT is further enhanced by a series of optional functions. The first of these are reference sets, or RefSets. SNOMED CT RefSets allow developers or users to restrict the SNOMED CT content available for a particular purpose. RefSets can be created: to represent various languages (e.g. English, Spanish, etc); to create subsets of SNOMED CT content for particular organisational purposes; or for different medical specialties. The term reference set has been used only since the introduction of the new release format for SNOMED CT in 2011, known as RF2. Prior to this time this functionality was called the subset mechanism. The cross mapping mechanism in SNOMED CT facilitates the mapping from SNOMED CT concepts to other coding systems, such as classifications. The 44

67 cross mapping mechanism works only in one direction, linking single SNOMED CT concepts to one or more codes in the target coding system Maintenance The IHTSDO updates the SNOMED CT core international release twice a year, in January and July. During each release, changes may be made to both the structure and content of SNOMED CT. Additions or modifications to concepts or descriptions and re-modelling of concepts in the SNOMED CT hierarchies are examples of changes made to SNOMED CT during the maintenance cycle. Editorial principles for maintenance are outlined in the SNOMED CT Editorial Guide. 102 Management of suggestions Suggestions for SNOMED CT content are made through a formal request submission process. In IHTSDO member countries, the country s national release centre co-ordinates the request submission process. Users in member countries send suggestions to their national release centre. Suggestions are collated by the release centre and those meeting the IHTSDO s criteria for inclusion in SNOMED CT are forwarded to the IHTSDO through an online request submission system. The online system allows requestors to track a submission, identifying its current status. An is sent to the requestor after a successful submission has been included in SNOMED CT. If the submission has been rejected an is sent to the user explaining the reasons for the rejection. 103 Until 2012, the College of American Pathologists SNOMED Terminology Solutions (CAP STS) was contracted to the IHTSDO to perform maintenance on SNOMED CT. 54 This responsibility is gradually being transferred to the IHTSDO SNOMED CT-AU The IHTSDO permits and encourages the creation of national extensions of the international release of SNOMED CT to incorporate local content that is not appropriate for inclusion in the international release. In response NEHTA, Australia s SNOMED CT national release centre, developed an Australian 45

68 version of SNOMED CT, called SNOMED CT-AU. The first release of SNOMED CT-AU occurred in December SNOMED CT-AU differs from SNOMED CT in a number of ways. It contains the SNOMED CT international release but removes some unnecessary content (for example, non-human concepts). It also contains the Australian extension with additional specific content created by NEHTA for Australian usage. Another feature of SNOMED CT-AU is the inclusion of the Australian dialect reference set, a reference set that contains Australian-specific language (e.g. nappy instead of diaper ) and spelling (e.g. diarrhoea instead of diarrhea ) for SNOMED CT international concepts. When necessary, NEHTA also releases additional supporting documentation with SNOMED CT-AU specific to Australian usage. 105 These documents are only available to SNOMED CT-AU licence holders, including the FMRC Status of SNOMED CT-AU The November 2010 version of SNOMED CT-AU was used in this thesis. This version contained 292,262 current (active) concepts and 761,380 descriptions. 3.4 Application of resources in the methods used in this thesis The resources have been applied in this thesis using a variety of methods. Each results chapter contains a Methods section describing the Methods specific to that individual chapter, including an explanation of how each of these resources have been used. 46

69 4 Development of requirements for an Australian SNOMED CT general practice reference set 4.1 Background After NEHTA was formed and the Australian Government purchased a licence for SNOMED CT (as described in Chapter 1), a number of projects were established to aid the implementation of SNOMED CT in Australia. As previously mentioned, one of the first major focuses was the preparation and release of a version of SNOMED CT containing Australian content, known as SNOMED CT-AU. 76 In parallel, the IHTSDO continued to develop the international version of SNOMED CT. The IHTSDO began a process to replace the existing formats used to release SNOMED CT data structures (now referred to as Release Format 1, or RF1) with a new set of specifications for SNOMED CT release formats (known as Release Format 2 or RF2). 106 As part of the upgrade from RF1 to RF2 the subset mechanism in SNOMED CT that enabled developers and/or users to identify small areas of SNOMED CT content for a particular purpose was replaced with a new reference set (RefSet) mechanism. RefSets have previously been discussed in Section The RefSet mechanism was used in Australia to develop RefSets for clinical specialties and facilitate the uptake of SNOMED CT in Australia. The first of these was the emergency department reference set (EDRS) project, commissioned by the Australian Government Department of Health and Ageing. Content in the EDRS was derived from existing termsets used in Australian hospital emergency departments. These terms were mapped to SNOMED CT-AU using a tool developed by the Commonwealth Scientific and Industrial Research Organisation (CSIRO) called Snapper. 107 The EDRS was first released with the July 2010 SNOMED CT-AU release. 108 As the level of computerisation in Australian general practice is high, 109 the FMRC felt that general practice was an ideal candidate for early implementation of SNOMED CT-AU and approached NEHTA to develop an Australian general practice RefSet of SNOMED CT-AU. After negotiations 47

70 with NEHTA over three years, the FMRC was contracted in August 2009 to produce an Australian general practice RefSet of SNOMED CT-AU, which was to be the second of the SNOMED CT-AU clinical content RefSets. During negotiations for this contract, NEHTA requested that stakeholder engagement form the initial phase of the project, and the proposed project was expanded to include an additional component for consultation with stakeholders. This became the first phase of the project. The aim of the consultation process was to identify stakeholder requirements for an Australian general practice RefSet of SNOMED CT-AU, formally document these requirements and prepare a set of measurable criteria for each requirement. These criteria would then be used to evaluate the structure and content of the RefSet against each requirement, providing evidence that the RefSet incorporated the documented stakeholder requirements. Stakeholders were divided into three groups for face-to-face workshops that I organised and conducted. Workshops were conducted for the following groups: GPs; NEHTA Clinical Leads; and GP EHR software vendors. 4.2 Introduction Chapter 1 contained an introduction to clinical terminologies currently used in Australia (see Section 1.4). The aim of the workshops was to evaluate current attitudes towards coding systems and clinical terminologies, and to identify deficiencies in the coding systems or clinical terminologies currently used. Underlying these aims was the need to identify requirements for the SNOMED CT-AU RefSet for general practice, so as to ensure the RefSet would meet the needs of users. Understanding GPs current attitudes to, and experiences with coding systems was vital for planning the future implementation of SNOMED CT-AU in Australian general practice, and to ensure that GP software vendors and GPs themselves would be willing to implement and use the planned SNOMED CT-AU general practice RefSet in GP EHRs. In the workshops some words were used interchangeably by GPs and vendors. For example, some referred to coding systems, others to terminologies. In this chapter I have standardised the language used in the 48

71 body, but direct quotations from GPs or vendors retain the original terms used by the workshop attendees. 4.3 Methods Method of selecting participants GPs GPs were selected to participate in the GP workshop based on their responses to the BEACH GP characteristics questionnaire (Appendix 4) in which they were asked to indicate: the extent to which they use computers in their clinical practice; the functions for which they use computers in clinical practice; and which clinical software program they use. GPs who stated they were either completely paperless, or used hybrid medical records (a combination of computer and paper-based records) were identified, representing GPs who were computer users. This list was further restricted to GPs who were located in New South Wales to improve the likelihood of the GP being able to attend a face-to-face workshop. GPs using the following software programs were approached: Medical Director isoft Medtech 32 Genie Solutions Monet Zedmed Intrahealth Best Practice. The software vendor isoft owns two products used in Australian general practice Practix and Monet and GPs using both products were approached. The Medical Director software has a high proportion of the market share (as shown in Table 4.1 in the Results section) so two GPs using this software were included. I generated approach lists according to the software product used by the GP. In each list GPs were ordered according to their BEACH unique identifier (called a DOCID) which broadly equates to when a GP participated in 49

72 BEACH. GPs who had participated in BEACH more recently (hence having a higher DOCID) were approached first as they were less likely to have moved from the practice at which they participated in BEACH. GPs were approached in a stepwise manner such that GPs were only approached once those higher in the list had refused their invitation. This process removed the likelihood of over-representation from any one vendor. GPs who had completed the BEACH study between April 2009 and September 2009 were initially approached for participation in the workshop (as of 11 September 2009 there were 153 NSW GPs in this sample). Once this sample was exhausted, GPs who had participated in BEACH between April 2008 and March 2009 were approached (339 GPs available). Metropolitan GPs were selected over rural GPs to increase the likelihood of their attendance at a face-to-face workshop. I sent letters of invitation to GPs requesting their participation in the workshop (Appendix 5) and followed up non-respondents by telephone. In total, I sent letters to 62 GPs. NEHTA paid the GPs for their attendance according to the Commonwealth Government remuneration package. There are two other software vendors who provide general practice clinical software Communicare and Northern Territory Health (NT Health) but both supply primarily to the Aboriginal care sector. In BEACH, GPs are included in the BEACH sample frame if they have claimed a minimum number of Medicare GP service items. Many of the Aboriginal Controlled Health Services are financed through block grant funding rather than through Medicare, so it was not surprising that there were no BEACH participants using these software products present in my sample. These two software suppliers were approached independently of the sampling process and asked for the name of a GP user who might attend the workshop. An apparent lack of interest from the GPs saw neither of these software systems represented at the workshop. NEHTA Clinical Leads NEHTA has employed a group of over 40 clinicians to provide expert clinical advice and feedback on NEHTA s projects. 110 NEHTA requested that the 50

73 FMRC conduct a specific workshop with the Clinical Leads to obtain their feedback into the project. I sent an invitation to the NEHTA Clinical Leads through the Director of the NEHTA clinical unit, requesting the participation of Clinical Leads in a workshop designed specifically for them. I received a sufficient number of positive responses from this , and no further recruitment was required. The Clinical Leads who attended were all practising GPs. I subsequently asked Clinical Leads what brand of clinical software they used in clinical practice, but this was not a requirement for participation in the workshop. GP clinical software vendors Using the Medical Software Industry Association (MSIA) list, the Chief Executive Officer of the MSIA sent an invitation on my behalf to GP clinical software vendors to attend the GP clinical software vendor workshop, with a reminder sent two weeks later. After the vendor workshop, GP software vendors who were not able to attend the workshop but registered interest in the project were interviewed individually Method of analysis I conducted the workshops in the following manner: I prepared Microsoft PowerPoint slides that were presented by my supervisor, A/Prof. Graeme Miller and I. The presentations summarised the main themes to be covered in the workshop. A short introduction about each topic was given, followed by discussion around the group on that topic. After each topic was presented participants were encouraged to give feedback and opinions on the topics discussed. Participants were also encouraged to identify related areas of interest or concern for discussion by the entire group. Leading questions asked during the workshops included: How do you enter information into the following data fields: reason for encounter/reason for consultation? problems/diagnoses/reason for prescription? 51

74 prescriptions? referrals? pathology or imaging tests? Are there any deficiencies in the methods you use to enter this information? If so, what are they? Is there anything you would like to improve about how you enter information into your EHR? How do you extract data from your EHR? Are there ways data extraction from your EHR could be improved? The GP session was held on a Wednesday afternoon in October Anecdotally, Wednesday afternoons are often a time during which GPs do not have face-to-face encounters with patients. This date and time was chosen with the expectation this would improve participation. The vendor session ran for one day between 10:00am and 4:00pm in October 2009, and the workshop with NEHTA Clinical Leads ran between 10:00am and 4:00pm in November The workshops were recorded and I transcribed the recordings. I extracted the qualitative data collected during the workshops from the transcripts, identifying themes to capture the current use of clinical terminology in electronic health records, GP and vendor attitudes towards clinical terminology and future terminology needs of stakeholders. After discussion with NEHTA staff, I used the identified themes to create a set of formal requirements for an Australian general practice reference set of SNOMED CT. I prepared the resulting document (Project plan and requirements specification: General practice reference set project) in consultation with NEHTA staff using the NEHTA document template and adhering to NEHTA specifications. The document is attached as Appendix 1. 52

75 4.4 Findings Characteristics of workshops Table 4.1 provides an overview of the breadth of the consultation process. The table compares the market share of GP software vendors with the number of GPs, Clinical Leads and GP software vendors consulted in this process. The market share of each GP software vendor was calculated using responses to a question in the BEACH GP Survey that asked GP participants about the brand of clinical software used. Table 4.1: Comparison of market share against consultation Company Market share (a)(b) (%) Number of GPs consulted Number of Clinical Leads consulted Vendor consulted (Yes/No) Termset/ Terminology used ACS Computing N/A 0 0 Yes N/A Best Practice Yes Best Practice termset Communicare (c) Yes ICPC-2 PLUS Genie Solutions Yes ICPC-2 PLUS/ ICD-10-AM Global Health N/A 0 0 Yes N/A HCN (Medical Director) Yes Docle termset isoft (Practix) Yes ICPC-2 PLUS isoft (Monet) Yes ICD-10 Medical Objects N/A 0 0 Yes N/A Medtech No ICPC-2 PLUS MIMS N/A 0 0 Yes N/A NT Health (c) No ICPC-2 PLUS Stat Health N/A 0 0 Yes ICPC-2 PLUS Zedmed Yes ICPC-2 PLUS (a) Based on responses to the question What clinical software is used? in the BEACH study GP profile between April 2008 and March 2009; n=914. (b) Figures do not total 100%. Other programs = 3.2%. (c) Due to the sampling methods used in the BEACH study the proportion of GPs using software developed by Communicare or NT Health is an underestimate of the true proportion of GP users of these software programs. Both Communicare and NT Health s Primary Care Information System (PCIS) are primarily used in Aboriginal Community Controlled Health Services, many of which are not funded through the Medicare Benefits Schedule. GPs working in these settings may therefore not perform the minimum number of Medicare services to qualify for inclusion in BEACH. Further detail about sampling methods used in the BEACH study is provided in Section Note: GP general practitioner; HCN Health Communication Network; N/A not applicable; NT Northern Territory. Note: N/A (not applicable) in the column outlining each product s market share indicates that the company did not have a marketed general practice clinical software product at the time of the workshops. MIMS develops a medication coding system that is integrated into other GP clinical software. Stat Health s clinical software was in development when the workshops were conducted. ACS Computing, Global Health and Medical Objects did not have a clinical EHR software product used by GPs at the time of the workshops. 53

76 Of the 62 GPs approached to participate in the GP workshops, eight GPs agreed and seven GPs actually attended (11.3%). Six GP Clinical Leads attended the Clinical Leads workshop and eleven vendors participated in the vendor workshop Terminologies or coding systems currently incorporated into GP clinical software The coding system or terminology used in each GP clinical software product represented at the workshops is also shown in Table 4.1. The majority of products incorporate ICPC-2 PLUS, developed and maintained by the FMRC at the University of Sydney. ICPC-2 PLUS is described in detail in Section 3.2. In all systems the use of ICPC-2 PLUS is optional. Best Practice uses a locally developed termset, hereafter called the Best Practice termset. Medical Director uses a termset based on the Docle codes. 68 Genie Solutions incorporates ICD-10-AM as well as ICPC-2 PLUS. As previously described, ICD-10-AM is a classification designed for morbidity and mortality coding, 111 and is primarily used in secondary care (hospital) settings Themes emerging from workshops Results from the workshops will be presented according to the themes discussed Awareness of, and attitudes towards coding Participants in the GP workshops were asked about whether they used a coding system, or entered data from a picklist (indicating use of a coding system or clinical terminology) in their GP clinical software. All the GP participants reported they entered some form of structured data using picklists into their EHR. However, some GPs demonstrated a lack of awareness about the coding system or terminology used in their EHR. GPs could describe how they entered data by searching and picking a term from a picklist, but did not call this coding. Awareness of SNOMED CT varied considerably between the groups. Those in the GP group were mostly unaware of SNOMED CT a few had heard the name but had no knowledge about the product beyond its name. The clinical lead group had varying levels of knowledge some regarded themselves as very familiar with SNOMED CT, and others knew of SNOMED CT but had not 54

77 investigated it in detail. All the vendors had heard of SNOMED CT, but only Medical Objects had implemented it. Medical Objects produces an application for health care providers to send and receive secure messages from their clinical software 113 not a clinical EHR itself. No GPs reported using SNOMED CT in their clinical work. The benefits of coding entering data into EHRs using structured clinical terminologies were not uniformly recognised by workshop participants. GP: I don t see the usefulness of coding at all. GPs from the Clinical Leads group said although awareness about coding had increased in recent years at an organisational level within governments and Divisions of general practice, there was a general lack of awareness about reasons for coding among practising GPs, and this created a barrier to GPs understanding of the benefits of using coding systems in their GP clinical systems. The primary reason GPs saw for coding data was to populate health summaries or problem lists for communication with other health care providers. An example of this would include sending a health summary that outlines the patient s diagnosed health conditions, current medications and past procedures, with a referral letter to a specialist. GP: The main aim of maintaining the datasets (using a coding system) is so that you can put the accurate data into the referral letters. GPs also said that health summaries and problem lists assisted with patient continuity of care, and that this improved their management of individual patients. GP: From my point of view the most important reason to use those codes is really to develop the current problem list. This is the most important thing, so that if you re not there (and) someone else sees your regular patient, blind Freddy can see what their problems are and they don t miss something and give the NSAIDs to the patient who s had a haemorrhagic peptic ulcer. GP: I m trying to create a summary view of the patient because that makes my management of the patient better. Health summaries are documents containing a synopsis of information related to a patient s history. The RACGP Standards for general practices (4 th edition) specifies that GPs must have a health summary for at least 75% of their active patients. These summaries should contain details relating to seven 55

78 core components of patient care, including: adverse drug reactions; lists of current health problems and current medications; relevant past personal history; personal social history and family history; and immunisations and health related risk factors. 95 Vendors saw the linkage between coding systems and decision support tools as an important reason for coding, however this view was not shared by the users. GP: Those prompts are a killer. You put the prompts in (and) they [individual GPs in the practice] turn them off. GP: Most doctors say I know it anyway. It s in my head. Discussion The fact that some GPs could not name the coding system used in their EMR is not necessarily negative, as coding systems are always integrated into an EHR s workflow. This was reflected in a statement made by one vendor that the implementation of SNOMED CT should be seamless, and that users should not be aware that they are using SNOMED CT. One of the underlying themes discussed in the workshops was communication. Comments from both GPs and vendors indicated that communication, particularly with specialists, is one of the primary reasons for coding, to send a list of the patient s current problems and medications with a referral letter. At present, most communication runs from a GP to a specialist or hospital. There is a clear need for communication to occur in both directions to achieve interoperability. Interoperability is defined as "the capability to communicate, execute programs, or transfer data among various functional units in a manner that requires the user to have little or no knowledge of the unique characteristics of those units" Usability issues A common theme from GPs currently using coding systems was their positive attitude towards favourites lists of terms. A favourites list is a file of the codes the GP uses frequently in their clinical software. This is an application that has to be built into the system by vendors, and is not available in all systems. 56

79 GP: I find that (the favourites list) quite useful. It saves you going through what is quite often a frustrating process to find something that will actually fit with what you re trying to say. However some GPs also acknowledged that using a favourites list makes users lazy and less accurate in their coding. The inclusion of a favourites list in clinical software may increase the likelihood that the GP chooses terms that appear on the favourites list because terms on this list are easy to find, rather than searching further for a more appropriate term for the patient s problem. GPs were questioned about how easy it was to use the coding system currently included in their clinical system. GP: It s easy to use, but it s not necessarily efficient to use and it doesn t necessarily find you the answer. In contrast: GP: It s not that easy. I would have to click at least three or four times to get the right thing. These statements include a number of facets relating to usability, including ease of use, efficiency and completeness. Each of these will independently be discussed below. One part of the issue of usability related to how coding systems and terminologies are implemented in clinical systems. When asked how often they could not find a suitable match in their EHR for the term they wished to record, the GPs responses varied from once a month to regularly. There are two primary reasons users would not be able to find a term in a coding system or clinical terminology: 1. The term does not exist in the terminology. 2. The term exists in the terminology, but cannot easily be found. In the GPs descriptions of not being able to find terms, both reasons were identified. GP: I would say 10-15% of the time what I want is not in there and it s not that I can t find it, it s not in there. The same GP gave the following example: 57

80 GP: Everyone knows a sebaceous cyst is actually called an epidermal cyst and has been done for about 15 years. The system does not have epidermal cyst I ve got to hit sebaceous cyst. These statements demonstrate that in some circumstances the termsets or clinical terminologies used by GPs in their EHRs are not sufficiently complete for users. This leads to the question of how many terms are needed for completeness. In previous comments GPs stated that they like the concept of a favourites list because this minimises the effort needed for searching through termsets and clinical terminologies. Put together, these comments indicate that there is a delicate balance between GPs having access to a sufficient range of terms to enter into their EHRs, too many terms, and too few terms. The need for completeness in a terminology is inherently linked to the need for ongoing terminology maintenance, to ensure that terminologies are updated with terms not previously identified, with new clinical conditions as they arise, and in line with changes to the medical terminology used. The second reason listed above, of terms existing in a termset but not easily being found was also a source of frustration. Of particular annoyance to the GPs was the additional time needed to search for terms that could not be located easily. GP: I would have to click at least three or four times to get the right thing. This comment relates to the search and display mechanisms used in the termsets and clinical terminologies. When this issue was discussed further the GPs said that they sometimes gave up entering the term using the termset/ clinical terminology, if it was too hard to find. GP: Occasionally, probably more than occasionally I can t find a match that I m looking for and I might have one more cut at it another way and give up and put it in free text. All codeable fields in an EHR should have an attached free text field to capture information relating to the context of the associated coded term.{miller, /id} The comments made by GPs in the workshops indicate that GPs use these free text fields as a secondary data entry mechanism if they cannot find an appropriate term quickly and easily in the termset or clinical terminology. As a result the information entered in free text 58

81 is not captured in a standardised manner, which may result in spelling errors or acronyms that are ambiguous or not widely understood. This negatively impacts on the usefulness of the data for re-use. The use of clinical terminologies standardises the words used to capture clinical information, particularly when these data are recorded in specified data elements in a relational database. The combination of standardised data elements and clinical terminologies minimises the effort needed to search data as the user does not have to consider every possible synonym, acronym or spelling mistake when conducting a search. It also facilitates grouping for data reporting and analysis and communication with other health care providers (interoperability). GPs also complained about the lack of flexibility in search and display mechanisms in their existing systems. One problem that was shared by GPs who use a commonly used GP EHR was that the software allows them to search only for terms from the start of a string. For example, if the GP was searching for ischaemic heart disease, they had to enter ischaemic as the first word in the search. Searches starting with the word heart did not link to the term ischaemic heart disease. A lack of consistency in how terms are displayed was also identified, as GPs felt this made searching harder. Vendors were acutely aware of users frustration around the issues of ease of use. When discussing ease of use, both GPs and vendors saw time and number of mouse clicks as closely related markers of the performance of a coding system or clinical terminology. GPs and vendors both mentioned that one or two mouse clicks was optimal when searching for terms. In the opinion of one vendor speed was more important than number of mouse clicks: Vendor: Irrespective of that (mouse clicks) it has to be faster than what they re used to. It can t be slower If it s more clicks and faster then they ll still complain but as long as it s faster at least it s dependable. GPs identified some enhancements that would improve the search and display mechanisms in their EHRs. These included: the use of wildcards. A wildcard is a character (e.g. an asterisk) that allows searching of a central portion of a word and all its variants

82 highlighting words included in the search string when results of searches are displayed (e.g. if the user enters DIAB in the search, all terms containing this search string are highlighted as Diabetes). Google type search mechanisms. Utilising synonyms and/or acronyms rather than relying on the exact string being found in the search. Search and display enhancements suggested by the GPs came from their own experiences with using computers. As one GP said, We re all used to Google, and the concept of highlighting words from the search string is one used in Microsoft Outlook, a commonly used application. Another GP suggested the alternative of typing free text into a data field in an EHR that could then be converted to a code automatically. This suggestion is known as natural language processing. Discussion Some of the issues identified, particularly those around search and display mechanisms, relate to the interface used to access the coding system, rather than a deficiency within the system itself. The party responsible for user interfaces varies. In some instances the interface is included as a requirement of the coding system. As described in Section 3.2.3, a software vendor who implements ICPC-2 PLUS is provided with a set of functionality requirements that specifies the search and display mechanisms to be used, including the use of keywords as the interface to access terms, and the display of the picklist in alphabetical order. In other coding systems no (or few) specifications are given to vendors and vendors can choose to implement the system as they wish. In England, the NHS went into partnership with Microsoft in to jointly develop a common user interface (CUI). The aim of the CUI was to minimise variance in the user interface between electronic medical record systems used in England. This in turn aimed to make it easier for clinicians to move between practices or health settings within England, with the clinician being assured of a consistent user interface to enter clinical information in each setting. 116 At present, the Microsoft CUI covers patients 60

83 personal data, allergies and date displays, with future plans to include work on clinical notes for hospital admissions and primary care, and displaying tables and charts. 117 evaluated. The implementation of the Microsoft CUI has not yet been Issues related to the ease of use of EHRs in general practice have not been widely researched. An Australian study conducted in 2001 found that there was no relationship between how easy GPs found using EHRs and their actual computer use for clinical purposes Extracting data from EHRs Attitudes towards extracting data from EHRs varied between GPs and software vendors. Vendors saw data extraction as one of the main purposes of coding: Vendor: They ask for data extraction. They don t ask for coding systems. In some instances, GPs initially said that they did not extract data from their EHRs. When asked specific questions about whether they ran data reports to recall patients, the GPs all responded in the affirmative. GPs indicated that their practices extract data from their medical records to identify: female patients who are due for a Pap smear or mammogram patients who are due for routine immunisations (e.g. the influenza vaccine for patients aged 65 years or more) patients who require follow-up for abnormal blood test results patients who need routine follow-up of chronic conditions. One GP said that he prefers extracting longitudinal data about individual patients to examine a single patient s management over time, rather than analysing data from many patients across his practice. The GP said he finds this approach more clinically useful as it aids continuity of care when he manages his patients. GPs, especially those who were also NEHTA Clinical Leads, stated that the introduction of the Australian Primary Care Collaboratives Program (APCC) had increased the profile of data extraction. The APCC is a program that aims 61

84 to assist Australian general practices improve the quality of care provided to patients. 119 The program began in 2005 with a first phase, 120 and initially focussed its work on three areas: diabetes, coronary heart disease and access and care redesign. 119 Two new focus areas have been added in a second phase, including chronic obstructive pulmonary disease and chronic disease prevention. 119 These GPs reported that general practice as a profession is starting to see the value of data through audit programs such as the APCC and that this has led to more data extraction from general practice records. In turn, this has increased awareness around improving the quality of data entered into EHRs. GP: People are starting to see the value of data. Another issue that arose during discussions about data extraction related to who in a practice is responsible for organising data extraction, and the impact this person has on the data entered by others in the practice. GPs spoke of needing a champion within the practice to encourage others to enter data appropriately to aid extraction. It would appear that these champions are often practice managers or practice nurses. GP: in my practice if I was not driving it rigorously with my practice manager and my practice nurse it just would not happen. The same GP questioned the long-term sustainability of extracting reliable data from EHRs within a practice, particularly if the practice champion leaves the practice. GP: If you have a change of practice nurse or practice manager you go right back to square one. Discussion Reliable data extraction is enhanced by the use of standardised coding systems or clinical terminologies. 60 Data entered into EHRs in a standardised manner increases the reliability of the information that is extracted. For example, the use of coding systems can minimise the amount of typing by the user, which in turn reduces the likelihood of spelling errors. Further, at the data extraction stage searches on a text string (for example, diabetes) will not find mis-spelt words (for example, daibetes, diabets etc). 62

85 In the workshops, GPs said their practices mainly extracted data in relation to reimbursement programs through the Australian Medicare Benefits Schedule (MBS) Practice Incentives Program (PIP). Over the last decade, there has been a trend in MBS funding towards disease-specific payments. For example, GPs are given additional payments if they complete a cycle of care involving at least two asthma consultations within a one year period, for patients with moderate or severe asthma. 121 These incentives provide GPs with motivation to enter and manage data in their EHRs in such a way that it can be easily extracted and entered into recall letters sent to patients or for reporting purposes. No national data are available on the proportion of general practices that extract data from EHRs. One study conducted in a single geographical area of general practice (known as a Division of General Practice) in Melbourne estimated that 44% of practices in that area are extracting data from their EHRs in The culture of general practice A number of the themes discussed in the workshops related to the current organisational culture in general practice. The issue raised above about needing a champion in a practice to extract data from medical records also relates to the organisational culture of general practice. One topic that arose a number of times in both the GP and the Clinical Lead workshop was the need to clean data stored in EHRs. Data cleaning refers to the process of reviewing data, checking for duplicated data and reviewing whether data have been entered into the correct data fields in the EHR. One example of the need for data cleaning given by one GP was reviewing a patient s medical record to find asthma included on the problem list multiple times. A chronic condition such as asthma should be included on a problem list once only, as patients are diagnosed with asthma once. Even though a patient with asthma may experience multiple exacerbations of asthma, the underlying condition does not change. Similarly, another GP expressed frustration that acute, self-limiting conditions (e.g. upper respiratory tract infections) would appear on problem lists long after the condition had resolved. In this GPs software, the problem list was derived of terms entered only into the problems managed field. The GP reported that he entered self- 63

86 limiting conditions into the reason for encounter field only, and not the problems managed field, to avoid this occurring. A number of the GPs in the workshops found the data cleaning process to be necessary but time consuming to ensure that data contained in the record was correct. GPs also stated that people who receive problem lists (e.g. specialists) report they are frustrated when they receive a problem list that contains duplicated and/or out-of-date information. One of the roles of the practice champion mentioned above is to educate each GP in the practice to verify whether a patient s problem is already included in the record before they enter a new problem, to avoid such duplication. A closely related topic was the use of synonyms by different GPs in a single practice. GP: you ll have one doctor calling it cancer of the breast, and another one would call it breast neoplasia and another one would call it breast tumour. So you get all these multiple variations on the same idea. Just for me myself, as a personal thing I choose the one that I feel is most (pauses), that I prefer. This GP was justifying the existence of duplicated data in a medical record. In his view, duplicating data through the use of synonyms was a legitimate form of duplicating data so that he could include his preferred term in the record, rather than another GP s preferred term. The overarching theme in the examples discussed above is that of education. The role of the champion is to educate users about the most effective and appropriate ways to enter data into their EHR. The topic of GP education about using EHRs was specifically discussed, particularly in the workshop with Clinical Leads, a number of whom had roles in training either GPs or medical students. A lack of familiarity about computers in general, and EHRs specifically, was seen as a barrier to the effective use of coding systems in EHRs. This was acknowledged by a number of the GPs: GP: I can write faster than I can type and time is of the essence. GP: A lot of things we leave out because we re not typists. This issue may in part be due to the age distribution of GPs in Australia. In Australia in , 39.9% of GPs were aged 55 years and over, while only 64

87 6.9% were aged less than 35 years. 93 As older GPs retire and are replaced with younger, more computer literate GPs, familiarity with computers will increase and typing speed will become less of a barrier to computer use. However, familiarity with computers and typing speed is only one problem relating to education. One GP Clinical Lead who is involved in training medical students said that students never see the various GP EHRs in a classroom setting, and they are never taught about using computers clinically or how to enter or manage data, and added that this also applies to practising GPs. GP: The problem is the way they (GPs) are using the product. They don t know how to use the product in the first place. Another GP indicated that there was not an effective change management process when general practice, as a profession, became computerised: GP: We ve gone through a decade (long) transition from paper to almost (fully) computer, but we haven t changed the culture of how we actually enter the information [into the medical record]. GPs also said that they were sometimes reluctant to use the computer during patient consultations, and stated this was primarily due to time constraints: GP: We don t have 30 seconds in a consultation to code. GP: I think the danger is we all spend too much time looking at the screen and not the patient and sometimes I just forget the computer and deal with the patient, and then I ll do it all after the patient s gone, or if I m really busy make a note and go back at the end of the session and put in the patient notes then. In relation to the future implementation of SNOMED CT-AU in Australian general practice, GP software vendors requested support for the implementation, specifically requesting vendor education in the form of implementation guidelines and tools (e.g. terminology browsers) to aid vendors during the SNOMED CT-AU implementation process. Vendors also expressed concern that GPs have developed familiarity with the EHR systems they currently use, and the processes used within those EHRs. As stated by one vendor: Vendor: If it s another ten keystrokes for a doctor to code something, I mean, they would just beat us to a pulp so it s got to at least be as useful as what we have and if it s worse by even one keystroke then, you know, we ve got real issues. 65

88 An example of a piece of information important to GPs was being able to include processes of care (such as procedures, immunisations and check-ups) as problems being managed. The specific example discussed during the vendor workshop related to including the procedure appendicectomy as a problem in the problem list. Vendor: It s a procedure, but it s actually historically relevant when someone comes in with tummy pain, so you can t not have them there. There are a variety of views about the relevance of including terms for processes of care within a list of terms for problems managed. The view of the vendors present at the vendor workshop was that Australian GPs regard processes of care as appropriate terms to include in a problem list. This topic will be discussed in greater detail in Chapter 5. Discussion In recent years general practices have become multidisciplinary, with 79.0% of GPs reporting that a practice nurse is employed at their major practice. 93 Practices have also become larger. In % of GPs reported they worked alone (as a solo GP/in a solo practice) and 38.9% stated they worked in a practice with more than 5 other GPs. 82 By % of GPs reported working solo, and 60.8% worked in a practice with 5 or more other GPs. 93 The larger size of general practices means that GPs are now less likely to be familiar with other GPs practising styles, methods of entering information in medical records and preferred clinical terms. A number of the statements made relate to the issue of time. It is widely recognised that GPs are time poor, both during consultations and outside patient consultations. In the mean length of an Australian general practice consultation was 15.3 minutes. 93 A study conducted in 2001 indicated that 95% of Australian GPs undertook work that was not reimbursed by the national health insurance scheme (Medicare). 123 No more recent studies have been undertaken in this area. In the workshops, GPs stated that they write information into free text boxes rather than coding if finding a code is difficult, as an example of how they take shortcuts to save time during a consultation. 66

89 The reported lack of familiarity with computers may in part be due to the age distribution of GPs in Australia. In , 39.9% of GPs were aged 55 years and over, while 6.9% were aged less than 35 years. 93 As older GPs retire and are replaced with younger more computer literate GPs, familiarity with computers will increase and typing speed will become less of a barrier to computer use. One of the strongest messages arising from the workshops was that coding systems and clinical terminologies have to be easy for GPs to use, in terms of containing sufficient content, and in the methods used for implementation. As one GP very succinctly stated: GP: Unless it s simple you won t do it. 4.5 Development of requirements from results of the workshops As stated in Section 4.3.2, after the workshops were completed I used the findings to create a set of requirements for an Australian general practice RefSet of SNOMED CT-AU. The themes and issues discussed in Section 4.4 provided the basis for the requirements, as written in the document Project plan and requirements specification: General practice reference set project (included as Appendix 1). Every requirement included in the document is outlined below, including justifications for including the requirement. Requirements were divided into three categories: Content: these requirements relate to the content of the RefSet Implementation: requirements describing how the RefSet should function. Governance: requirements outlining responsibilities for development and maintenance of the RefSet. Each requirement contained fit criteria, a set of criteria that outlines the scope of the requirement and provides additional specifications for each requirement. Where relevant, the fit criteria are also discussed in describing the reasons for including the requirement. 67

90 4.5.1 Content requirements Although the requirements prepared for the GPRS refer to the SNOMED CT-AU terminology, all requirements relating to content can be applied to any terminology used in general practice. GPs had a number of concerns about the structured terminologies they currently use not containing a sufficient number of terms. Such concerns were addressed by including the following requirement: The GPRS must be sufficiently comprehensive to cover clinical terms used in general practice (BR.GPRS.0001). To do this, development of the GPRS was going to utilise four clinical terminologies/termsets currently used in Australian general practice as sources for general practice terms: the Classification and Terminology of Community Health (CATCH), the University of Adelaide s GP termset (created using terms from the Medical Director software), ICPC-2 PLUS and the termset used in the Best Practice software. Utilising terminologies and codesets already used in general practice during GPRS development would ensure a bottom-up approach when creating the GPRS and would increase the likelihood of the terminology being suitable for GPs. Of course, GP complaints about not being able to find appropriate terms in specific termsets or clinical terminologies was also acknowledged, and thus four termsets were chosen, with the intention that a broad approach would ensure comprehensiveness in the resulting GPRS. Another aspect of comprehensiveness specifically mentioned within this requirement was that of contextual information, to ensure that users are able to include attributes such as finding site, laterality and chronicity, severity and certainty of diagnosis with the concepts/terms included in the GPRS. Another requirement included in the project plan and requirements specification relating to content in the GPRS was that: The values in the GPRS must be capable of being restricted based on the context of their usage (BR.GPRS.0002). The criterion for this requirement was that the GPRS should include terms relating to the following data elements: reason for visit/reason for encounter 68

91 problem/diagnosis/reason for prescription past and family history health interventions allergies reason for medication change. The list of data elements above was selected because these data elements represent the most commonly used data elements in general practice EHRs, and are important in the transfer of patient information between health care providers. The original list did not contain the data element reason for medication change. This element was requested by the MSIA, and subsequently added to the list by NEHTA. Creating links between concepts/terms in a clinical terminology and data elements in EHRs is known as terminology binding, 124 and can be used to streamline the terms available for a particular data element. For example, it is not clinically relevant to include acute and transient illnesses such as the common cold in a data element for family history of disease. Including concepts such as common cold in a terminology list for family history of disease results in a larger than needed list, making it harder and more time consuming for GPs to search the list to find the most appropriate term. A consideration listed in this requirement was that the GPRS content must populate health summaries and problem lists. The close relationship between the data elements listed above and the RACGP requirements for health summaries (previously discussed in Section ) should be noted. Based on this requirement the GPRS could populate information in five of the seven core components outlined in the RACGP Standards for general practice. 95 Two components of the RACGP health summary standards relate to medication lists and immunisations. The disease for which a patient is immunised can be captured in the GPRS (for example, hepatitis immunisation). However, medication names and dosages (including medications provided in the course of an immunisation) are not available in 69

92 SNOMED CT-AU, and therefore these two components cannot be included in the GPRS. However, the requirement does not state the method to be used to bind the terminology to the data elements. This is problematic how do we know what terms are used as reasons for encounter, and what terms are used as problems managed? Most of the source termsets provided are single termsets with no inherent bindings included in them. It was therefore recognised that consideration would to be given to the methods used to undertake the terminology binding. A third requirement relating to the previous two requirements listed was the need for exception handling. The requirement was listed as: Users must be able to record clinical terms that do not exist within the GPRS (BR.GPRS.0003). It was clear in the workshops that GPs were frustrated by searching for terms that did not exist within their current terminology/termset. In the Project plan and requirements specification: General practice reference set project this is listed as a content requirement, although it has features that are closely related to the implementation of the GPRS, and SNOMED CT-AU overall. The next requirement was associated with the need for managing legacy information. The implementation of the GPRS must enable the retention of legacy information (BR.GPRS.0004). All GPs consulted in the workshops used some form of structured data entry in the form of a termset or clinical terminology. As such, most GPs using EHRs in Australia will have information stored electronically in a coded format. The implementation of SNOMED CT-AU through the GPRS should not overwrite or delete information that was previously stored using a different termset or clinical terminology. Legacy information can be retained by creating a linkage from a legacy data set (referred to hereafter as a source termset or terminology) to SNOMED CT-AU. These linkages are called maps, and the process used to create the linkage is called mapping (see Chapter 6 for a full explanation of mapping). 70

93 I included a second form of linkage in the requirements, with a requirement stating that: The GPRS must be linked to relevant international classifications (BR.GPRS.0005). As stated in Chapter 1, classifications are used to group and analyse clinical data for epidemiological purposes. During the workshops extracting data from EHRs was identified as an important reason to enter data using structured terminologies. Creating linkages from the SNOMED CT-AU GPRS to classifications for grouping and extracting data from EHRs is therefore vital. Two classifications were explicitly mentioned in this requirement, stating that the GPRS must be linked to ICPC-2, and should be linked to ICD-10-AM. The history and background of ICPC-2 is discussed in detail in Chapter 1 and the structure of ICPC-2 is outlined in Chapter 3. ICD-10-AM is primarily used in the hospital sector, and is briefly introduced in Chapter 1. The need for continuing maintenance of terminologies was discussed in detail in the workshops. As previously reported in Section , GPs expressed frustration that terms they require are sometimes not available in their current terminology or termset. There were three separate requirements included relating to maintenance of the GPRS: 1. Values contained within the GPRS should be kept up to date (BR.GPRS.0006). 2. The GPRS must be released at least every six months (BR.GPRS.0017). 3. There must be a process for maintaining the ongoing development and maintenance of the GPRS that has been endorsed by the GPRS governing body (BR.GPRS.0015). These requirements were included to ensure that maintenance of the GPRS was prioritised after its creation. The first of these requirements is general, stating only that maintenance of the GPRS needs to occur. The second requirement provides more detail, specifying that maintenance must occur in line with the standard SNOMED CT-AU release cycle, which at present is every six months. The requirement was deliberately worded as at least every six months to allow for additional releases of the GPRS outside the standard SNOMED CT-AU release cycle if required due to user need. The third 71

94 requirement addresses the governance requirement for maintenance by ensuring that there would be a formal maintenance process built into the governance of the GPRS Implementation requirements The implementation requirements included in the requirements document were derived from the issues raised during the workshops and NEHTA s protocols relating to SNOMED CT-AU development. Implementation requirements resulting from the workshops included: Implementation of the GPRS should have a minimal impact on the users operation of a vendor s system (BR.GPRS.0009). This requirement was included to acknowledge vendors concerns about the change management process required during implementation of the GPRS. The majority of user complaints during GPRS implementation would be directed in the first instance to vendors. Minimising the number and scale of changes required to an EHR during the transition to the GPRS should lessen the impact of the change from the users perspective. A related requirement providing greater specificity about the usability of the GPRS was: Terms within the GPRS should be fast and easy to find, and minimise the amount of typing at the point of entry (BR.GPRS.0010). This requirement was directly related to GP and vendor concerns about the ease of using termsets and clinical terminologies, discussed in Section Two requirements were included in the document to address the topic of education, raised during both the GP and vendor workshops, one relating to end user education, and the other to vendor education. End users of the GPRS must receive education in its use (BR.GPRS.0011). Implementers of the GPRS are required to undergo appropriate training as to the usage and implementation of the GPRS (BR.GPRS.0012). In the second of these, the word implementers was used in place of the word vendors. This was to acknowledge that the GPRS may be used in research 72

95 projects and for other non-commercial purposes. These requirements are quite broad, reflecting the need for education without specifying the method(s) of education. Further detail about potential educational methods are provided, with suggestions that end user education may take the form of user guides, face-to-face training, online tutorials or context-sensitive help systems as appropriate. For vendors, suggestions for education include an implementation guide, release notes, data structures specifications or reference set specifications. Other implementation requirements were written to incorporate NEHTA s SNOMED CT-AU obligations to ensure that the GPRS adhered to national and international specifications. The first of these was: The GPRS must release all files in SNOMED CT-AU Release Format 2 (RF2) and fully comply with the SNOMED CT Release Format 2 specifications (BR.GPRS.0007). Another NEHTA requirement was that: The terminology viewers included in the SNOMED CT-AU product must allow viewing of the GPRS in addition to its current specification (BR.GPRS.0008). NEHTA includes a terminology viewer (alternately called a browser) with each update of the SNOMED CT-AU release files. This viewer also includes any SNOMED CT-AU RefSets developed by NEHTA. This requirement ensures that the GPRS would be included in the terminology viewer with each release Governance requirements There were also requirements around issues of governance. In Australia, the licence holder for SNOMED CT is the Australian Government. This licence provides Australian users with free access to SNOMED CT through NEHTA. NEHTA sub-licences SNOMED CT to Australian users through the SNOMED CT Affiliate licence agreement and the Australian National Terminology Release licence agreement. The associated requirement for this issue stated: (The) GPRS will be released under two licences: 1) the SNOMED CT Affiliate Licence Agreement and 2) the Australian National Terminology Release Licence Agreement (BR.GPRS.0013). 73

96 There was another requirement covering the governance of the GPRS: There must be a governance structure in place that is responsible for the ongoing management of the GPRS (BR.GPRS.0014). Although each requirement was written to be complete within itself, many of the stated requirements were closely related, with a number of overlaps between the requirements. For example, there were three requirements encompassing the issue of maintenance of the GPRS. Two of these related to content and one to governance. This demonstrates the complexity of ensuring that the development of a clinical terminology such as the SNOMED CT-AU GPRS satisfies the needs of all relevant stakeholders. 4.6 Limitations The sample of GPs participating in the workshops was small, and therefore may not be a representative sample of computer-using GPs. Similarly, the logistic difficulties of organising face-to-face workshops during working hours meant that all participants in the GP workshop were metropolitan GPs from New South Wales. Representation in the Clinical Leads workshop was broader, with two metropolitan GPs from Victoria, one metropolitan GP from South Australia and two GPs from Queensland (one metropolitan and one rural). While additional workshops with GPs from a broader range of geographic locations would have been preferred to discuss the themes further, this was not possible due to time constraints in the project, set by NEHTA. The group of GPs who were clinical leads are a group of GPs who have a particular interest in e-health, so their views and opinions were more developed than the views expressed by GPs in the GP workshop. 4.7 Conclusion The results reported in this chapter formed the basis of the document Project plan and requirements specification for the General Practice Reference Set project. The candidate was the primary author of this document under contract to NEHTA. This document is included as Appendix 1. 74

97 The underlying focus for the workshops was the future introduction of SNOMED CT-AU in Australian general practice. GPs in the Clinical Leads workshop were aware of SNOMED CT and SNOMED CT-AU, however GPs in the GP workshop were not. Some of the comments made, particularly by the vendors, were directed towards SNOMED CT specifically, and may have reflected the vendors preconceived ideas about SNOMED CT. There is no doubt that the introduction of SNOMED CT-AU in Australia will have an impact on vendors, as they will be responsible for converting the termset or clinical terminology they currently use to SNOMED CT-AU. This may take considerable time and resources, and vendors expect to be reimbursed for this additional resource use. In addition, vendors have a responsibility to their users, and vendors will be the main people receiving complaints if the implementation of SNOMED CT is not optimal, or requires GPs to make changes to their work practices. GPs have their own preferred terms, and do not like having terms dictated to them, as indicated by the GP who said s/he choose(s) the one that I prefer. This has implications for a terminology such as SNOMED CT that has a list of preferred terms. The preferred term in SNOMED CT may not be the same preferred term that is used in general practice, or that is even preferred by a single GP in a practice with other GPs. The workshops with GPs and vendors proved to be a valuable first step in the project to develop a SNOMED CT RefSet for Australian general practice. After the workshops were completed, the requirements for a SNOMED CT RefSet for Australian general practice were developed. Each requirement was linked to a set of criteria. After the RefSet was released and developed, these criteria would be used to determine whether each requirement had been met. It soon became apparent that creating measurable criteria was not easy for some of the requirements. For example: (a) What is an appropriate size for the RefSet? This has to take into account GP views that medical terminology had to be easy to use or they would not use the terminology (the one or two clicks principle), 75

98 and that they did not like having to search extensively for terms they could not find. (b) Should processes of care be available for selection in problem lists? I realised that data collected from GPs through the BEACH study of general practice formed a unique dataset I could use to further investigate the topics discussed during the workshops, and to determine answers to the questions outlined above. These form the basis for Chapter 5: Current use of medical terminology by GPs in Australia. 76

99 5 Current use of medical terminology by GPs in Australia 5.1 Background Workshops with GPs and software vendors about the development of a SNOMED CT-AU RefSet for Australian general practice, described in Chapter 4, highlighted a number of issues about the optimal characteristics of general practice terminologies, particularly in terms of size, scope and content. As described in Section 3.3, SNOMED CT-AU (the Australian version of SNOMED CT) is a large clinical terminology containing over 300,000 concepts. The IHTSDO acknowledges that the size of SNOMED CT is an impediment to its implementation, and have therefore created a technical solution called reference sets or RefSets that allow users to identify and group subsets of SNOMED CT concepts for a particular purpose. 100 After the workshops I was convinced that the development of an Australian general practice SNOMED CT-AU RefSet needs to have a sound evidence base to ensure it is representative of the terminology currently used by Australian GPs. Although four termsets were identified as sources for the RefSet during the development of the GPRS requirements (Section 4.5.1), the usage patterns of each terminology had not been analysed in terms of their size, scope and content. I realised I was in a unique position to study this using ICPC-2 PLUS, one of the terminologies listed as a source for the RefSet and the terminology used in the BEACH study of general practice. Most of the data elements in BEACH are coded using ICPC-2 PLUS (as shown in Figure 3.1), both of which have been described in detail in Chapter 3. The use of ICPC-2 PLUS and BEACH data to assess the current use of terminology in Australian general practice is somewhat dependent on the extent to which ICPC-2 PLUS contains the medical terms used by Australian GPs. One of the concerns raised by GPs in the workshops related to clinical terminologies not containing a sufficient number of clinical terms to record the 77

100 breadth of their practice (Section 4.5.1). As a result of this discussion, the following statement was included in the GPRS requirements: The GPRS must be sufficiently comprehensive to cover clinical terms used in general practice (BR.GPRS.0001). For many years there has been debate about how much information should be explicitly included in a clinical terminology. Some terminologists have stated that the terms contained in clinical terminologies should be expressed as closely as possible to the term s usage in everyday language. 125,126 The concepts of precoordination and postcoordination were introduced in Section If clinical terminologies explicitly contain all the terms (or concepts) that exactly represent the term s usage in common parlance, they may contain considerable precoordination, and the opportunity for combinatorial explosion arises. Combinatorial explosion is the function by which the size of a terminology increases exponentially due to the explicit inclusion of every possible precoordinated term. 50 It is commonly recognised that combinatorial explosion should be avoided in clinical terminologies. Often, the phrases atomic concept or atomic term are used to describe a term that contains a single meaning. Atomic is defined as the least portion of a thing or quality 16 and atomic terms or concepts cannot be deconstructed into smaller parts. 127 If interpreted strictly, this definition implies that a term such as malignant neoplasm is precoordinated, because there are multiple grades associated with neoplasms (e.g. benign, pre-malignant). In this thesis, I will use the term kernel term or kernel concept instead of referring to atomic terms or concepts. Kernel terms/concepts may be atomic or precoordinated terms/concepts that act as the primary expression in an overall term or concept 127 and describe terms that have a single meaning or definition. For example, in the term Left otitis media the kernel term is otitis media. Detail can be added to kernel terms or concepts by appending descriptions to the kernel term, such as attributes, contextual information or another kernel term. An attribute represents a characteristic of the meaning of a concept or the nature of a refinement. 128 In this thesis, attributes refer to the second aspect of this definition, and are used to describe items that refine or provide 78

101 supplementary meaning to a term or concept. There are multiple types of attributes, including modifiers, which change the meaning of a term clinically by adding a disease stage or grade; qualifiers, which change the meaning of a term in a temporal or administrative sense; and statuses, such as episodicity. 127 Precoordinated terms or concepts can also be created by combining two or more clinical concepts (or meanings) into a single term or concept, known as a composite concept/term. 16,127 Contextual information can be used in conjunction with kernel concepts to provide additional meaning about the status of the kernel concept, for example, whether the patient has a personal or family history of the kernel concept, or whether the kernel concept is not present in a patient (negation). 129 The aim of this chapter is to investigate selected issues (identified in Chapter 4) relating to the current use of clinical terminology in Australian general practice, using the ICPC-2 PLUS terminology as a research tool. Each section in the chapter addresses a single topic raised during the GP and/or vendor workshops relating to terminology size, scope or content, and contains its own aims, methods and discussion. Four areas will be addressed: the extent to which the terms included in ICPC-2 PLUS reflect the free text terms used by GPs. the number of ICPC-2 PLUS terms used to describe RFEs and/or problems managed. the scope of terms used by GPs to describe problems managed. the size and scope of changes made to ICPC-2 PLUS during maintenance of the terminology. Further discussion about common issues identified in these studies is included in Section

102 5.2 Comparison of GP free text descriptions and ICPC-2 PLUS coded terms Background The extent to which ICPC-2 PLUS can be used to assess the patterns of GP terminology use is dependent on how well ICPC-2 PLUS represents the clinical terms used by Australian GPs. Therefore I conceptualised the following study to determine the extent to which ICPC-2 PLUS contains the terms used by Australian GPs, and to identify characteristics of terms used by Australian GPs that are not included in the ICPC-2 PLUS terminology, including the use of attributes and composite terms. Results from this study will inform other aspects of current terminology use discussed later in this chapter Aims To determine the extent to which coded ICPC-2 PLUS terms are representative of the free text recorded by GPs for the problems managed data element. To determine the characteristics of those free text terms not captured in ICPC-2 PLUS, when the free text and ICPC-2 PLUS terms are not synonymous Methods This section describes a study undertaken in conjunction with the BEACH national study of general practice. As previously discussed in Section 3.1, GPs in the BEACH study record at least one and up to four problems managed at each encounter. These data are recorded on structured paper encounter forms in free text by the participating GPs, and entered into the BEACH Microsoft Access database by secondary coders. Each free text entry on the structured encounter form is coded into the equivalent data field in the BEACH database. Free text data written in the problems managed data fields must be coded using ICPC-2 PLUS. For this study the BEACH database was modified to include an additional field. This additional field was a free text field of 100 characters, allowing the 80

103 coders to type the full free text into the database as written by the GP. The free text field was linked to a field containing the ICPC-2 PLUS coded term. This study was undertaken over two months, starting in September 2010 and ending eight weeks later in November After two months the sample size was deemed sufficient to provide meaningful results and it was decided the sample had reached saturation. Coders were given the following instructions: Code the problem managed as you always would using ICPC-2 PLUS (Section 3.2.3). Then, type the problem managed into the free text field exactly as it was written. Do not expand abbreviations. Include question marks if they are written. Correct the doctor s spelling if you know that it is wrong. Do not type in punctuation (e.g. if the doctor has written O.A. or O/A type it in as OA. If the doctor has written: # type fracture L type L R type R + type and. Coders were deliberately instructed to code the problem prior to typing the free text, to ensure that the problem managed was coded in the usual way and to minimise the likelihood of the coded entry differing from that normally selected. Data were exported from the BEACH database into an Excel spreadsheet, with each row in the spreadsheet containing the ICPC-2 PLUS code, the 81

104 associated ICPC-2 PLUS term and the related free text description for a single entry. I drafted a set of criteria against which each free text term and its associated ICPC-2 PLUS term would be assessed. Three characteristics were chosen for this assessment. 1. The match type: an assessment of the quality of the match between the full free text and the selected ICPC-2 PLUS term. If the match type was assessed as more specific (that is, the free text contained information not present in the coded term, and therefore had greater specificity than the coded term) the match was further assessed as described in Step 2 below. 2. Type of specificity lost: an assessment of the type of information not captured in the coded term. If the information lost was an attribute the match was assessed further, as described in Step Attribute detail: specifies the type of attribute included in the free text term, when the presence of an attribute was recorded in Step 2. I selected a random sample of 1,000 entries from the spreadsheet, and pilot tested each of these entries according to the draft criteria. I surmised that 1,000 entries would be a sufficient cross-section of entries to test the draft criteria. The criteria were modified during this process to include additional criteria needed for full assessment of the test entries. After 1,000 entries had been evaluated and assigned criteria, I reviewed and updated the set of criteria. The final list of criteria against which each entry was assessed is presented in Box One entry only was permitted in Columns 1 (match type) and 2 (type of specificity lost). Many free text terms contained more than one attribute, so multiple entries were permitted in Column 3 (attribute detail). After testing, all assessments made during the pilot phase were deleted. I then repeated the assessment process, ensuring that all entries were evaluated against the same set of criteria. 82

105 Box 5.2.1: Criteria used to assess relationship between free text and ICPC-2 PLUS term Match type STEP 1 MATCH TYPE Abbreviation used for match type Definition Exact match E A lexical and semantic match between the free text and the coded term (i.e. the free text and the coded term are identical) Synonym S A semantic match between the free text and the coded term (i.e. the two terms mean the same thing, expressed in different ways) More specific* ML The free text description contains more detail, and is more specific than the coded term Less specific SL The free text description contains less detail, and is less specific than the coded term Best fit BF The free text description does not have an equivalent lexical or semantic match to the ICPC-2 PLUS coded term. The free text has therefore been coded to an ICPC-2 PLUS term that is broadly comparable to the free text term Not representative NR The coded term is not representative of the free text Composite term C The presence of two or more clinical descriptions in a single free text description (e.g. hypertension and diabetes) STEP 2 TYPE OF SPECIFICITY LOST (* INCLUDED ONLY WHEN A MATCH TYPE OF MORE SPECIFIC HAS BEEN RECORDED) Attribute # A An attribute modifier was included in the free text description, but not represented in the coded term Detail lost DL Detail about the kernel concept included in the free text description has been lost in the coded term STEP 3 ATTRIBUTE DETAIL INCLUDED IN FREE TEXT ( # INCLUDED ONLY WHEN THE PRESENCE OF AN ATTRIBUTE HAS BEEN RECORDED IN COLUMN 2) Certainty of diagnosis C The problem under management has not been confirmed (e.g. possible hepatitis;? pregnancy) Causation CA The problem s causal factor(s) (e.g. viral pharyngitis) Duration D The amount of time a problem has been experienced, expressed either as a frequency (e.g. diarrhoea for 2 days) or a grade of duration (e.g. acute rhinitis; chronic shoulder pain) Laterality L The side of the body in which the problem was experienced (e.g. conjunctivitis of left eye; bilateral otitis media) Recurrence/ Episodicity R Indicates that the problem is recurrent or episodic in nature (e.g. recurrent thrush; flare-up of rheumatoid arthritis) Review RE Indicates that a pre-existing problem is being reviewed by the GP (e.g. review hypertension). Severity SE A grading of how badly the patient is affected by the problem (e.g. mild asthma; severe pharyngitis) Site SI The anatomical site where the problem is experienced (e.g. lumbar scoliosis; wart on hand). Age Y The age (or age level) of the patient experiencing the problem (e.g year check; newborn examination). Other O Other attribute not included in the list of attributes 83

106 5.2.4 Results Over the period of this study, 101 recorded encounter pads were entered into the BEACH database, equating to 10,100 encounters. Within these encounters, 15,447 problems managed were coded and the associated free text recorded (results not tabled). Of all problems, 41.1% were assigned as an exact match, indicating that the free text and associated code were identical. A further 18.2% of matches were assessed as synonyms, where the coded ICPC-2 PLUS term had the same meaning as the free text (Table 5.2.1). For one-quarter of problems managed (25.5%) the free text description was assessed as more specific than the coded ICPC-2 PLUS term (that is, the free text contained more detail than the coded term) and as less specific than the ICPC-2 PLUS term for 4.5% of problems managed. The match type of best fit was assigned to 3.9% of ICPC-2 PLUS terms. This category indicates that there was no exact or synonymous match found in the ICPC-2 PLUS termset for the free text written, but the term coded was the most appropriate term available without being more specific or less specific than the source term. There were 1.5% of the selected ICPC-2 PLUS terms assessed as being not representative of the free text. For 5.4% of the problems managed terms, the free text was assessed as a composite term (where more than one kernel term was included in the free text), and the match between the free text and ICPC-2 PLUS term was not considered further (Table 5.2.1). 84

107 Table 5.2.1: Assessment of the match type for problems managed between free text and ICPC-2 PLUS coded terms Match type Number Per cent Exact match between free text and coded term 6, Free text and coded term are synonymous 2, Free text more specific than coded term 3, Free text less specific than coded term Code is the best fit for the free text term Code is not representative of the free text term Free text is a composite term Total 15, These results indicate that approximately 60% (59.3%) of the free text problems managed terms were captured at the same level of specificity after being coding with ICPC-2 PLUS, either as exact matches or synonyms. Further consideration was subsequently given to the remaining 40% of problems managed for which free text and the ICPC-2 PLUS term was not synonymous, particularly for those matches assessed as more specific, where the free text contained more detail than was captured in the coded ICPC-2 PLUS term. The uncaptured data in the coded term was classified as one of two types: as an attribute of the kernel term or as other detail lost. Nearly three-fifths (59.2%) of the free text terms with more specificity than the coded ICPC-2 PLUS terms contained attributes such as laterality, severity or duration. The remaining 40.8% had other detail lost in the term that was not classifiable (Table 5.2.2). Table 5.2.2: Assessment of terms marked as more specific Attribute type Number Per cent Attribute 2, Other detail lost 1, Total 3, There were 2,741 attributes recorded within the 2,335 free text terms that contained attributes. The most common unencoded attribute was body site (26.0% of all attributes recorded), followed by laterality (20.5%) and level of certainty of the problem label (14.3%) (Table 5.2.3). 85

108 Table 5.2.3: Distribution of attributes included in free text Attribute type Number Per cent of all attributes recorded Site Laterality Certainty Causation Duration Review Recurrence/episodicity Severity Age Other Total 2, Discussion Findings from this study indicate that a considerable amount of information is lost when GPs free text is coded using ICPC-2 PLUS. Most of the information lost relates to attributes included in the free text that are not explicitly represented in ICPC-2 PLUS. This result is not surprising. Some attributes are included in ICPC-2 PLUS terms when the presence of the attribute is clinically important. For example, abdominal pain is divided into quadrants based on laterality and position left upper quadrant pain, right upper quadrant pain, left lower quadrant pain and right lower quadrant pain. The location of the pain is clinically important during the diagnostic process, e.g. right lower quadrant pain is indicative of appendicitis. ICPC-2 PLUS contains some such terms when the presence of the attribute is of clinical or diagnostic importance. However, the inclusion of attributes has been actively limited during the terminology s development to avoid combinatorial explosion. Avoiding combinatorial explosion is seen as a good practice in terminology development. 50 As described in Section 3.2.3, keywords are used in the ICPC-2 PLUS terminology to access terms. Synonyms form one type of keyword, and the high proportion of problems managed assessed as synonyms may indicate that the inclusion of synonyms as keywords does facilitate the correct identification of an ICPC-2 PLUS term. 86

109 GPs are known to use medical records as an aide memoir, a record to assist their memory of a patient s medical history, and it is possible that some of the inadequate or ambiguous information recorded on the BEACH encounter form could be an indication of this habit. A GP who writes veins as the problem managed may know that the patient has varicose veins, and utilise their memory of the patient s history, rather than explicitly recording the problem in their medical record. Another issue I noted while assessing the free text descriptions was the ambiguous inclusion of abbreviations. For example, the letter L was commonly used in the free text, and in some terms could be interpreted as either Left or Lumbar. This ambiguity can lead to misinterpretation and must be avoided to minimise any risk to patient care when information is transferred between clinicians (e.g. when a patient is referred from one clinician to another). I also identified ambiguity in the free text terms written by the GPs. Terms including temp, temperature and got a temperature were all included in the free text descriptions evaluated in this study. These terms are ambiguous from the machine-readable perspective; however human reasoning interprets this text to mean that a patient s core body temperature is above normal. I also noted the inclusion of Australian colloquialisms in the free text data, for example, crook which means that the patient feels unwell. Finally, some GPs included non-standard abbreviations in the free text. For example, the term XS sweating was used to describe excessive sweating. Similar issues have been discussed by others in the Australian setting. 130 The 4.5% of matches that were less specific is somewhat concerning, as it indicates that there was specificity contained in the selected ICPC-2 PLUS term that was not provided in the GP s free text description. In general, this means that the free text written by the GP was insufficient in some way, and the coder has used additional information recorded elsewhere on the BEACH encounter form to identify an appropriate ICPC-2 PLUS term. For example, a number of the ICPC-2 PLUS terms coded about contraception (e.g. Contraception;OC pill or Contraception;implant) were more specific than the free text (i.e. the GP wrote contraception ). In these situations the secondary coder was able to ascertain the type of contraception from other information 87

110 included in the encounter form, such as the inclusion of a prescription for the oral contraceptive pill, or a procedure for an intra-uterine device or contraceptive implant. The coders used this information to find the most appropriate ICPC-2 PLUS term. The inclusion of composite terms, containing more than one clinical meaning, can effectively be captured only using postcoordination. Similarly, the ability to create postcoordinated expressions that link kernel terms and attributes would facilitate the standardised capture of attributes and avoid combinatorial explosion. The extent to which the results presented in this study are reproducible is not known. Ideally, two independent assessors would have performed the assessment, with discrepancies between the assessors resolved by a third party. However, there was insufficient time for this, and I was the only assessor. It should be noted that this study did not include an assessment of coding reliability. Rather, it focussed on the accuracy of the coding system in representing the content of free text terms. As previously mentioned, I did find some instances where the free text had been incorrectly coded, and these were passed on to the Coding Supervisor. However, in the vast majority of cases, the ICPC-2 PLUS term coded was correct, even if some information was lost during the coding process. 88

111 5.3 Number of terms used in a general practice clinical terminology Background During the workshops described in Chapter 4, terminology users reported that there is a delicate balance between a terminology being too small (containing insufficient content) or too big (containing irrelevant or unnecessary content). The resulting requirement was: The GPRS must be sufficiently comprehensive to cover clinical terms used in general practice (BR.GPRS.0001). This requirement included a set of criteria outlining the source general practice termsets currently used in Australia, and indicated that content from these termsets should be included to ensure that the requirement is met. However, there was no overt definition in the requirement for the concept of sufficiently comprehensive. Such a concept is difficult to define, and largely subjective. As previously stated, SNOMED CT contains more than 300,000 concepts and has been designed to include many aspects of medicine. It is clear that in its entirety SNOMED CT is too large to be effectively used in general practice. However, the optimal size for a SNOMED CT general practice RefSet is currently unclear. In this section I will assess the utilisation of terms coded in two data elements of the BEACH study using ICPC-2 PLUS reasons for encounter and problems managed, and describe the frequency with which terms are used in each. This will help to determine an appropriate size for the general practice RefSet. There have been very few studies formally reporting utilisation data from standardised medical terminologies where the terminology itself is the variable of interest. Fung et al reported the use-patterns of source termsets used in the development of the Clinical Observations Recording and Encoding (CORE) problem list, and found that 10% of unique terms accounted for 85% of usage across termsets from six hospital and ambulatory care institutions in the USA and Hong Kong. Approximately 21% of terms accounted for 95% of usage

112 General practice data from the UK entered using the Read codes provided similar results 20 individual Read codes accounted for one-quarter of all Read codes entered in the THIN research database in 2009 and 3,871 Read codes accounted for 95% of all codes used. 131 The total size of the Read termset was not reported Aims The aim of this section is to determine the number of ICPC-2 PLUS terms used to describe reasons for encounter and/or problems managed in the BEACH study, and the frequency with which these terms were used Methods This is a secondary analysis of ICPC-2 PLUS utilisation data from the BEACH study. A detailed description of the BEACH study was provided in Section 3.1. In summary, GPs record up to three patient reasons for encounter and up to four problems managed. GPs write these in free text on a paper encounter form, as shown in Appendix 2. These free text entries are secondarily coded and entered into a Microsoft Access database using ICPC-2 PLUS. For this analysis, five years of BEACH data, from 2004 to 2009 were combined. During this time there were no changes in the data element labels or definitions for RFEs or problems managed in BEACH. Two lists of ICPC-2 PLUS terms were generated, one for reasons for encounter and one for problems managed. Each list included every active ICPC-2 PLUS term used at least once in BEACH between April 2004 and March 2009 to populate the RFE and the problems managed data elements respectively. The lists included the number of times each ICPC-2 PLUS term was used, the rate per 1,000 encounters and the rate per 10,000 encounters. The lists were sorted in decreasing order of term usage. Usually in BEACH data, similar or synonymous terms are classified to ICPC-2 for reporting or, where required, grouped to describe clinical concepts that cross ICPC-2 classes. For example, when reporting the clinical concept of Diabetes, type 2, synonyms such as non-insulin dependent diabetes mellitus, type 2 diabetes mellitus and adult onset diabetes mellitus are grouped together to represent the single clinical concept. However, this study 90

113 examines the use of individual ICPC-2 PLUS terms, and the results presented will therefore not be comparable to published BEACH reports Results Between 2004 and 2009, 4,864 GPs participated in BEACH, recording a total of 486,400 encounters. There were 7,580 terms in ICPC-2 PLUS that could be used for data entry between 2004 and 2009 (Table 5.3.1). Over the five year period: 6,098 terms (80.4% of all active terms) were used at least once to describe either RFEs or problems managed. three-quarters (74.4%) of active terms were used at least once to describe problems managed, equating to 92.6% of terms used at least once in either of the two data elements. more than two-thirds of all active terms (69.1%) were used at least once to describe RFEs, and these accounted for 85.9% of terms used at least once. there were 4,786 terms (equating to 63.1% of all active terms) used commonly to describe both RFEs and problems managed. Table 5.3.1: Use of terms from ICPC-2 PLUS in BEACH, Data element types Number of active ICPC-2 PLUS terms Per cent of all active terms (n = 7,580) Per cent of terms used at least once (n=6,098) Active (a) ICPC-2 PLUS terms used at least once to describe: RFEs 5, Problems managed 5, RFEs and problems managed 4, RFEs and/or problems managed 6, Note: RFEs reasons for encounter. (a) The definition of an active ICPC-2 PLUS term is given in Box 3.2. Distribution of terms used to describe RFEs Figure describes the utilisation of ICPC-2 PLUS terms used to code the RFE data element. 91

114 The most frequent ten terms describing RFEs accounted for 21.0% of all the RFEs recorded in BEACH between 2004 and Half (50.0%) of all RFEs recorded in BEACH were coded using only 66 ICPC-2 PLUS terms. One thousand terms accounted for 92.8% of usage. 1,290 ICPC-2 PLUS terms accounted for 95% of all RFEs recorded. These represented 24.6% of all ICPC-2 PLUS terms used to describe RFEs, and 17.0% of all ICPC-2 PLUS terms. The remaining 5% of RFEs were described using 3,946 ICPC-2 PLUS terms. 5,236 terms (100.0%) 1,290 terms (95.0%) 1,000 terms (92.8%) 200 terms (69.5%) 100 terms (57.4%) 66 terms (50.0%) 10 terms (21.0%) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Figure 5.3.1: Number of terms describing reasons for encounter expressed as a proportion of total usage 92

115 Table lists the ten ICPC-2 PLUS terms most frequently used by GPs to describe patient RFEs between 2004 and Blood pressure check-up was the most frequently used term, recorded at a rate of 49.2 per 1,000 encounters, and accounting for 3.2% of all terms used to describe RFEs over the five year period. The term cough accounted for a further 3.2% of all RFEs. Table 5.3.2: ICPC-2 PLUS terms most frequently used to describe reasons for encounter, Term used to describe RFE Rate per 1,000 encounters Per cent of total RFE terms (n = 5,236) Cumulative per cent Check up; blood pressure Cough Prescription(s) Test result(s) Sore throat Medication; renew Rash; localised Pap smear Immunisation Pain; back Note: RFE reason for encounter. Distribution of terms used to describe problems managed Figure shows the number of terms used to describe problems managed, expressed as a proportion of all terms used for problems managed. The most frequently used ten ICPC-2 PLUS terms accounted for 19.0% of all the problems recorded in BEACH over the five year period. Half (50.0%) of all problems recorded were coded using 85 ICPC-2 PLUS terms. One thousand ICPC-2 PLUS terms accounted for 89.5% of all usage in problems managed. 1,703 terms were needed to describe 95.0% of all problems managed. These represented 30.2% of all ICPC-2 PLUS terms used to describe problems managed, and 22.5% of all ICPC-2 PLUS terms. 93

116 The remaining 5% of problems managed were described using 3,945 ICPC-2 PLUS terms. 5,648 terms (100.0%) 1,703 terms (95.0%) 1,000 terms (89.5%) 200 terms (64.5%) 100 terms (52.8%) 85 terms (50.0%) 10 terms (19.0%) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Figure 5.3.2: Number of terms describing problems managed expressed as a proportion of total usage Table lists the ten ICPC-2 PLUS terms most frequently used to describe problems managed by GPs participating in BEACH between 2004 and Hypertension was the term recorded most often, at a rate of 94.9 per 1,000 encounters, accounting for 6.2% of all terms used to describe problems managed over this period. This was followed by upper respiratory tract infection, which accounted for 2.8% of all terms used to describe problems managed. Cumulatively, the ten terms most frequently used accounted for 19.0% of all terms used for problems managed. 94

117 Table 5.3.3: ICPC-2 PLUS terms most frequently used to describe problems managed, Individual ICPC-2 PLUS term used to describe problem managed Rate per 1,000 encounters Per cent of total (n = 5,648) Cumulative per cent Hypertension Infection; upper respiratory tract Depression Asthma Oesophageal reflux Pap smear Immunisation; flu Hypercholesterolaemia Immunisation Diabetes mellitus Note: flu influenza Discussion There are two aspects of medical terminology use identified in this study. Firstly, a small number of medical terms are used to describe a large proportion of the content of both the RFE and problem managed data elements. In other words, the use of medical terms is concentrated, with GPs using a few medical terms a lot of the time when describing RFEs and problems managed. Secondly, there is considerable breadth of usage, with nearly 4,000 terms accounting for the lowest 5% of usage in both the RFE and problem managed data elements. These results reinforce the idiom that common things occur commonly. However they also indicate that GPs use a wide variety of terms if they are made available, even if many of these terms are not used often. About 6,000 medical terms were needed to capture all RFEs and problems managed in this study, equating to 80% of all the terms available in the ICPC-2 PLUS terminology. Approximately 5,500 terms were recorded within each of the data elements individually, and nearly 4,800 terms were used in both (Table 5.3.1). Even a small constraint in the number of terms available, for example restricting the termsets to 95% of the terms used, would mean the loss of approximately 4,000 individual terms, and would inhibit the ability of GPs to accurately record patient RFEs and problems managed using a terminology. 95

118 It is notable that the results presented for RFEs (Figure 5.3.1) and problems managed (Figure 5.3.2) are alike, indicating similar patterns of concentration and breadth of terminology use. Four in five (78.5%) terms used to describe either of these data elements were used to describe both, demonstrating considerable overlap of terms used (Table 5.3.1). The results presented in this section closely reflect those reported by Fung et al. In that study, usage data from problem lists in six institutions covering both hospital and ambulatory care were presented, and each problem list had a similar skewed distribution to that reported in this study. The proportion of terms needed to represent 95% of usage was also similar in both studies. When these six problem lists were rationalised, Fung et al found that an average of 21% of all unique terms represented 95% of the terms used in problem lists. 125 The current study showed that 22.5% of all ICPC-2 PLUS terms were used to represent 95% of the problems managed in BEACH. The similarity of the results is particularly notable given that these data were collected in different settings (hospital based compared with general practice). However, this comparison is based only on the distributions of terms used. I have not attempted to compare the individual use of terms in any way, although I hypothesise that there may be considerable differences given the different settings in which the data were collected. Rogers 131 also demonstrated the same skewed distribution using data coded in Read codes in the United Kingdom. However, it is difficult to directly compare Rogers data with this study, as the results are presented with the raw number of Read terms used rather than proportions, and the denominator (the total number of Read terms available) is not given to allow me to calculate the proportion to compare the results. Results from both Fung et al 125 and Rogers 131 indicate that the pattern of these results are reproducible, although this could be examined further by future authors. As stated in Section 5.3.1, in most research individual terms are grouped together into larger groups when reporting data to aid the statistical significance and clinical relevance of their results. For example, Vikstrom et al described the distribution of coded diagnoses, including a 96

119 comparison of data coded using ICD-10 with that coded in SNOMED CT. Unfortunately, as with many reports comparing code distributions, the codes in this paper have been grouped at a higher level, in this case at ICD-10 chapter level and equivalent high-level concepts in SNOMED CT. 132 The originality of the current study therefore lies in the technique used to report the results. Data from research studies coded using a terminology may have access to more granular data than is published. It would be interesting to use such data to identify the extent to which the pattern found in the current study is reproducible, particularly in general practice but also in other health settings. Similarly, it would be useful to repeat this study with the other source termsets identified in the GPRS requirements: CATCH; the University of Adelaide GP termset derived from Medical Director terms and the Best Practice termset. It is important to remember that clinicians want to be able to use the terms they like to use, rather than having a set of the preferred terms dictated to them. As stated by GPs in the workshops (Chapter 4), if users are not able to quickly locate a term in a standardised clinical terminology that they wish to use, they will not use the terminology. This study demonstrates that although there is a small number of terms used very often in general practice, there is also considerable breadth in the range of terms used. Limiting general practice termsets to those terms used very often may therefore inhibit the usability of medical terminologies by GPs. 97

120 5.4 The content of the problems managed data element Background Creating a structured list of a patient s problems in his/her medical record was first proposed by Dr Lawrence Weed in the 1960s. He created the problem oriented medical record (POMR), the core element of which is the problem list. Weed defined the problem list as a combination of a table of contents and an index and stated that it should include all patient problems at the level of information known to the clinician, including symptoms, diagnoses, social and preventive problems. 11 During the workshops described in Chapter 4, GPs stated that the population and organisation of problem lists and health summaries was one of their primary motivations for using clinical terminologies. GPs identified problem lists as a tool that supports continuity of care, and associated them with good quality clinical care (Section ). This was supported by vendors, who also requested that processes of care, including procedures such as appendicectomy, be available for inclusion in the problem list (Section ). It became clear that a general practice SNOMED CT-AU RefSet must support the management of problem lists to be successfully implemented. This was reflected in the considerations section of the GPRS requirement BR.GPRS.002: The GPRS must be sufficiently complete to be used to populate health summaries, problem lists, referrals and linked to decision support systems. In EHRs, problems are not always directly entered onto a problem list. They may be entered into a data element equivalent to problem/diagnosis at an encounter, and the problem list derived from data included in such a data element. At present in Australia, there is no standard information model used to structure general practice EHRs, and thus no standardised labels or definitions used to define the data element(s) equivalent to problem/diagnosis. When applying the GPRS requirement to the RefSet design with NEHTA (Chapter 4), the definition used for problems became a contentious issue. 98

121 The primary area of contention was the inclusion of processes of care, such as surgical and preventive procedures. Processes of care are also called interventions, and defined in the Wonca Dictionary of General/Family Practice as: the actions undertaken by a health care provider for the management of a reason for encounter or a health problem. Includes preventive and administrative activities, investigation, diagnosis, treatment, rehabilitation, and co-operation. 15 The NEHTA data specifications do not explicitly exclude processes of care from problem lists, as shown in the definition of problem/diagnosis below: An account of relevant identified health related problems as reported by a healthcare provider. This can include a disease, condition, injury, poisoning, sign, symptom, abnormal finding, complaint or other factor influencing health status as assessed by a healthcare provider. 133 The implicit nature of this exclusion led me to consider other definitions for problem/diagnosis to determine whether there is a consistent approach to the inclusion or exclusion of processes of care in equivalent data elements for problem/diagnosis, in various national and international standards. I compiled a summary of labels and definitions for concepts equivalent to problem/diagnosis (Box 5.4.1). These definitions have been compiled from a variety of Australian and international sources, and include international standards for EHRs (e.g. the European standard CEN 13940). The definitions indicate there is considerable variance in the labels and definitions used to describe the problem/diagnosis or equivalent concept. In most definitions, processes of care are neither explicitly nor implicitly included or excluded, indicating that there is no clear guidance on the use of processes of care within a problem/diagnosis field. 99

122 Box 5.4.1: Labels and definitions used to describe the concept of problem/diagnosis Author/Title Label Definition Processes included Processes excluded Processes not mentioned Wonca Dictionary of General/Family Practice 15 Health problem A concern in relation to the health of a patient as determined by the patient and/or the health care provider. Problems should be recorded at the highest level of specificity determined at the time of the encounter. GPs/FPs see patients with health problems that are not diseases and may not develop into a disease. CEN Health issue Issue related to the health of a subject of care, as identified or stated by a specific health care party. In the notes: According to this definition, a health issue can correspond to a health problem, a disease, an illness. But it may not, such as when it is simply a request for a procedure (therapeutic or preventive) by the subject of care or another health care party etc.? Continuity of care Problem Problems contains (sic) data defining the patient s relevant current and historical clinical problems, record (CCR) 135 conditions, diagnoses, symptoms, findings and complaints at the time the CCR is generated. HL7 EHR System Problem list A problem list may include, but is not limited to: chronic conditions, diagnoses or symptoms, Functional Model 136 functional limitations, visit or stay-specific conditions, diagnoses or symptoms. NEHTA data specifications 133 Problem/ diagnosis An account of relevant identified health related problems as reported by a healthcare provider. This can include a disease, condition, injury, poisoning, sign, symptom, abnormal finding, complaint, or other factor influencing health status as assessed by a healthcare provider. RACGP Standards for General Practice (4th edition) 95 Health summary Medical history The information in this element could include current conditions (e.g. hypertension), and inactive conditions (e.g. childhood asthma now resolved) as well as procedures and clinical interventions in the past. A current pregnancy status should also be recorded where applicable. BEACH recording instructions Diagnoses/ problems Diagnoses/problems managed at this encounter. Diagnose at the highest level with the information available. This may be at symptom level or specific diagnosis level e.g. cough, Pap smear, immunisation, contraception, check-up, Type 1 diabetes, asthma, etc. Medical records, medical education and patient care (Dr L Weed) 11 Problem Problems may be medical or social. If medical, it should fall into one of four types: a diagnosis a physiological finding a symptom or physical finding an abnormal laboratory finding. Note: although not explicitly included in this definition, procedures such as appendicectomy and gallbladder surgery are included as examples of items on a problem list. 100

123 The importance of the variance of the definitions included in Box is somewhat dependent on whether GPs view the exclusion of processes of care in a problem list as a restriction of their ability to populate a problem list or health summary. I therefore speculated whether GPs actually use processes of care to describe problems when they are able to do so. In the BEACH study of general practice, there is no restriction placed on the use of processes of care in the problem/diagnosis data element. In BEACH, GP participants are instructed to record at least one and up to four problems/diagnoses managed at an encounter in free text, to the highest level possible with information available to date. Processes of care, including Pap smears and immunisations, are explicitly included as examples of data that can be entered in this field (as shown in Box 5.4.1). As such, the extent to which GPs use processes of care to describe the problem/diagnosis at BEACH encounters will provide a measure of the importance of this issue in Australian general practice Aims The aims of this section are to: determine the range and distribution of terms used to describe problems managed at BEACH encounters, as coded in ICPC-2 PLUS, classified in ICPC-2 and grouped according to ICPC-2 components. identify changes in the use of process of care terms as descriptions of problems managed at BEACH encounters over the decade to identify the process of care terms most frequently used to describe problems managed at BEACH encounters Method This is an analysis of data from the BEACH study (see Section 3.1 for full BEACH methods) using the corrected ICPC-2 components (explained in Section 3.2). The primary unit of analysis is the problems managed data element. Results are presented as a proportion of all problems managed, with 95% confidence intervals (where relevant) to identify statistically significant differences. To identify changes over time, each of the ten years of BEACH 101

124 data from to were analysed. The total number of encounters recorded in each BEACH year is reported in Table Results presented in Table are similar to those reported in General practice activity in Australia to : 10 year data tables. 137 However, some data have been grouped differently for the purposes of the current study, with all process codes grouped into a single result. The distribution of process terms used in the problems managed data element (according to their placement in ICPC-2) was analysed for the data year. These results are expressed as a proportion of all problems managed, and as a rate per 100 encounters Results Distribution of problems managed, There were 155,373 problems recorded in BEACH in , during 101,349 encounters recorded by 988 GPs. Two-thirds (66.6%) of problems managed at BEACH encounters were expressed in terms of diagnoses or diseases. Problems expressed as symptoms and complaints accounted for 17.4% of all problems managed. Processes of care were used to describe 15.9% of problems (Table 5.4.1). Within the process component, diagnostic and preventive procedures were most frequently recorded (11.0% of all problems managed; 69.2% of process terms recorded as problems), with medications, results, referrals or administrative procedures together accounting for the remaining 5%. 102

125 Table 5.4.1: Distribution of problems managed, by ICPC-2 component, ICPC-2 component label Number Per cent of total problems (a) (n = 155,373) 95% LCL 95% UCL Per cent of processes (n = 24,727) Diagnosis, diseases 103, Symptoms & complaints 27, Processes (all) 24, Diagnostic & preventive procedures 17, Medications, treatments & therapeutics 3, Results 1, Referrals & other RFEs 1, Administrative 1, Total problems 155, (a) Figures do not total 100 due to rounding. Note: LCL lower confidence limit; UCL upper confidence limit; RFE reason for encounter. Distribution of problems managed, changes over time Table shows changes in the distribution of problems managed according to ICPC-2 component over the decade from to As a proportion of all the problems recorded at BEACH encounters, problems were less often described in terms of diagnoses/diseases in (66.6% of all problems managed, 95% CI: ) than in (69.1%, 95% CI: ). They also were less frequently described as symptoms and complaints in compared with , decreasing from 18.7% (95% CI: ) to 17.4% (95% CI: ). In contrast, the proportion of problems that were described in terms of processes of care was greater in (15.9%, 95% CI: ) than in (12.2%, 95% CI: ). In particular, as a proportion of all problems/diagnoses recorded, the use of process codes as problems managed increased in the following areas: diagnostic and preventive procedures, increasing from 8.7% (95% CI: ) in to 11.0% (95% CI: ) in test results, increasing from 0.5% (95% CI: ) in to 1.2% (95% CI: ) in administrative processes, from 0.2% (95% CI: ) in to 0.7% (95% CI: ) in

126 Table 5.4.2: Distribution of problems managed, by ICPC-2 component, BEACH, to Per cent of all problems managed (95% CI) ICPC-2 component (n = 143,528) (n = 139,092) (n = 146,336) (n = 144,674) (n = 137,330) (n = 149,088) (n = 136,333) (n = 146,078) (n = 149,462) (n = 155,373) (a) Diagnosis, diseases 69.1 ( ) 68.5 ( ) 67.5 ( ) 67.8 ( ) 68.0 ( ) 68.6 ( ) 68.2 ( ) 67.8 ( ) 68.1 ( ) 66.6 ( ) Symptoms & complaints 18.7 ( ) 18.8 ( ) 18.4 ( ) 18.0 ( ) 18.2 ( ) 17.6 ( ) 18.0 ( ) 18.4 ( ) 17.9 ( ) 17.4 ( ) Processes (all) 12.2 ( ) 12.7 ( ) 14.0 ( ) 14.2 ( ) 13.9 ( ) 13.9 ( ) 13.8 ( ) 13.8 ( ) 14.0 ( ) 15.9 ( ) Diagnostic & preventive procedures 8.7 ( ) 8.6 ( ) 9.3 ( ) 9.3 ( ) 9.1 ( ) 9.4 ( ) 9.3 ( ) 9.4 ( ) 9.7 ( ) 11.0 ( ) Medications, treatments & therapeutics 2.0 ( ) 2.3 ( ) 2.5 ( ) 2.7 ( ) 2.5 ( ) 2.2 ( ) 2.1 ( ) 1.9 ( ) 2.1 ( ) 2.2 ( ) Referral & other RFE 0.8 ( ) 0.8 ( ) 1.1 ( ) 0.9 ( ) 0.9 ( ) 0.8 ( ) 0.9 ( ) 0.8 ( ) 0.6 ( ) 0.8 ( ) Results 0.5 ( ) 0.7 ( ) 0.7 ( ) 0.8 ( ) 1.0 ( ) 1.0 ( ) 1.0 ( ) 1.2 ( ) 1.0 ( ) 1.2 ( ) Administrative 0.2 ( ) 0.3 ( ) 0.4 ( ) 0.4 ( ) 0.4 ( ) 0.5 ( ) 0.5 ( ) 0.6 ( ) 0.6 ( ) 0.7 ( ) (a) The direction and type of change from to is indicated for each result: / indicates a statistically significant linear change, indicates a non-linear significant change, and indicates there was no change. Note: CI confidence interval; RFE reason for encounter. 104

127 To examine this further, Table shows the frequency distribution of processes used to describe problems managed in , as classified to ICPC-2. Data from the most recent BEACH year ( ) were used in this analysis to reflect the current usage of process terms as problems managed. Preventive immunisation/medication (ICPC-2 code group 44) was the process most frequently recorded as a problem managed, accounting for 4.7% of all problems recorded, at a rate of 7.3 per 100 BEACH encounters. Partial medical examinations (2.3% of all problems managed), medication prescriptions/requests/renewals/injections (1.5%), results of tests and procedures (1.2%) and complete medical examinations (1.1%) were also commonly recorded (Table 5.4.3). The top ten rubrics shown in Table were used relatively frequently to code the recorded problems managed. The remaining rubrics were used rarely 29 rubrics each described 0.2% or less of all problems managed. The single process rubric not used as a problem managed was 61 Results examination/test/record/letter from other provider. Table 5.4.3: Distribution of process codes used to describe problems managed, ICPC-2 label ICPC-2 code (a) Number Per cent (n = 24,727) Rate per 100 encounters Preventive immunisation/medication 44 7, Medical examination/health evaluation partial 31 3, Medication-prescription/request/renewal/injection 50 2, Results tests/procedures 60 1, Medical examination/health evaluation complete 30 1, Histological/exfoliative cytology 37 1, Observation/health education/advice/diet 45 1, Administrative procedure 62 1, Blood test Other reasons for encounter NEC Repair/fixation suture/cast/prosthetic device (apply/remove) Excision/removal tissue/biopsy/destruction/ debridement/cauterisation Microbiological/immunological test (continued) 105

128 Table (continued): Distribution of process codes used to describe problems managed, ICPC-2 label ICPC-2 code (a) Number Per cent (n = 24,727) Rate per 100 encounters Encounter/problem initiated by provider Clarification/discussion of patient s RFE/demand Other preventive procedures Referral to specialist/hospital Therapeutic counselling/listening Other referrals NEC Other laboratory test NEC Consultation with primary care provider Other therapeutic procedures/minor surgery NEC Dressing/pressure/compression/tamponade Follow-up encounter unspecified Referral to other primary health care provider Diagnostic radiology/imaging Incision/drainage/flushing/aspiration/removal body fluid (excl. catheterisation) Diagnostic endoscopy Instrumentation/catheterisation/intubation/dilation Physical function test Other diagnostic procedures Local injection/infiltration Urine test Electrical tracings Physical medicine/rehabilitation Sensitivity test Faeces test Encounter/problem initiated by other than patient/provider Consultation with specialist Results examination/test/record/letter from other provider Total 24, (a) The symbol indicates that all ICPC-2 alphabetic chapter characters were grouped together for this analysis. Note: RFEs reasons for encounter; LCL lower confidence limit; UCL upper confidence limit; NEC not elsewhere classified. 106

129 5.4.5 Discussion Results from this study indicate that GPs have a broad view of the definition of problem/diagnosis, with a wide range of medical terms used when GPs completed this data element in free text. Findings from this study must be viewed in the context of the BEACH data used. The definition of problem/diagnosis used in the BEACH recording instructions explicitly includes examples of processes of care terms, including immunisations and check-ups (see Box 5.4.1). As such, it could be argued that GPs entered processes of care as problems managed in BEACH when they would not normally do so in normal practice, particularly if the inclusion of processes as problems managed is not supported by their EHR. BEACH collects information about the content of single GP patient encounters. In an EHR, data entered into a field labelled problem/diagnosis or problem managed can represent a symptom, diagnosis or process of care managed at a single encounter. These data may subsequently be used for many purposes, such as: a label for an episode of care (equivalent to Weed s definition of a longitudinal problem in a POMR) populating a problem list or health summary populating an event or encounter summary. The differing uses of data for each of these purposes are discussed below. A problem/diagnosis labelled at an encounter may become the start of an episode of care for the stated problem, and subsequently be managed at future encounters. An episode of care is defined as the period from the first presentation by the patient of a health problem or illness until the completion of the last encounter for the same health problem. 15 The notion of the episode of care is equivalent to the concept of problem as defined by Weed. Episodes of care relate to a single problem; that is, one patient with many problems may have many simultaneous episodes of care. An important feature of the episode of care is that the label or name of the problem in an episode can change over time. 138 For example, a problem labelled as a 107

130 symptom (e.g. sore throat) may be revised to a diagnosis label after investigation (e.g. tonsillitis). An evaluation of problems from an EHR that focuses on episodes of care rather than encounters may have a considerably different profile than the results presented here, because the same problem can be managed at many encounters during a single episode of care. As such, chronic conditions that require ongoing management may occur more frequently in an encounter based analysis of data from an EHR than in an episode based analysis. Although a process of care (such as a check-up) may be used as a problem managed at an encounter, it is unlikely that such a process would be used as a label for an episode of care, which represents a longitudinal view of a problem s evolution. 38 In particular, diagnostic and preventive procedures, which make up the largest proportion of processes used as problem labels, could be grouped into a single category in an episode of care using a label such as preventive care. In practise, procedures (including surgical procedures) are often included on problem lists. 125,139,140 In a Dutch study where patients were given the opportunity to make additions to the problem list, they commonly added procedures such as appendicectomy and hysterectomy, 139 which indicates that patients may associate their medical history with procedures performed. The inclusion of procedures is also important for patient safety, as evidenced by the recent death of a patient after surgery to remove the gallbladder. The patient s medical history stated that the gallbladder had previously been removed, but this information was ignored. 141 In its standards for general practice, the RACGP stipulate that 75% of patients in a practice must have a valid health summary that includes both current and past health problems for the practice to be accredited. 95 One of the core elements of the health summary is the patient s medical history, which includes procedures and clinical interventions in the past. 142 This inclusion reinforces the importance of the request from the workshop participants that users must be able to enter procedures into a problem list and reflects the results reported earlier in this chapter, with the group medications, treatments 108

131 and therapeutic procedures (incorporating surgical procedures), comprising 2.3% of all terms used to describe problems managed. The inclusion of processes of care as procedures is currently an important issue in Australia with the planned introduction of a Personally Controlled Electronic Health Record (PCEHR). The PCEHR will be a national, online repository for patients health information. Patients can register for a PCEHR from 1 July 2012, 143 with implementation due to commence shortly thereafter once software vendors have updated their EHRs to enable participation. 144 The PCEHR will not be a full clinical electronic health record, but a summary of the patient s health for transfers of patient care (e.g. referrals or emergency department admissions). 145 Two core aspects of the PCEHR are an event summary 146 and a shared health summary. 147 An event summary is equivalent to an encounter summary in general practice, or a discharge report from a hospital. A shared health summary is equivalent to the concept of a health summary as previously discussed in this section. Both summaries contain information about patients health problems, grouped into a high-level concept called medical history. There are three aspects to the medical history component of the summaries; problem/diagnosis (for which the definition has been sourced from the original NEHTA data specifications definition, as shown in Box 5.4.1); procedure; and medical history item. A medical history item is an entry belonging in medical history that is neither a problem/diagnosis nor a procedure. 148,149 According to these specifications, it appears that NEHTA has chosen to include a broad concept of medical history to encapsulate all health issues within the summary, regardless of the type of health issue. It is not stated in the NEHTA documentation whether the inclusion of a separate group for procedures in the medical history component precludes the recording of procedures in the problems/diagnoses section. Of the processes used as labels for problems managed, more than one in ten were classified as diagnostic and preventive procedures in ICPC-2. Medical examinations (i.e. check-ups), pathology tests, diagnostic imaging, preventive immunisations, counselling and education are all included in this group. In 109

132 Australia, GPs are seen as important providers of preventive care, with approximately 83% of the Australian population attending a GP in (personal communication, Department of Health and Ageing (DoHA), June 2010). The RACGP has published the Guidelines for preventive activities in general practice, also known as the red book (currently in its 7 th edition), 150 since The considerable proportion of problems managed described in this study as diagnostic or preventive procedures, in particular immunisations and check-ups, and the increased use of such terms over time indicates a trend towards GPs recording preventive procedures as problems in their medical records. Over time, the Australian government has increasingly recognised the contribution of GPs in providing preventive care, creating items in Australia s national health insurance scheme (Medicare) to reimburse GPs for defined preventive activities, including health assessments and screening for children, older people, Aboriginal and Torres Strait Islanders and other high risk population groups. 152 ICPC provides mixed guidance about the inclusion of processes of care on problem lists. All versions (ICPC-1, ICPC-2 and ICPC-2-R) acknowledge that a proportion of the health problems managed in primary care cannot be classified as a symptom, disease or diagnosis. However, the guidance given in ICPC as to how to classify these issues differs according to the version used. ICPC-1 states that the process components will occasionally be needed (e.g. for a vaccination, a Pap smear, or some advice) when ICPC is used in the diagnostic mode. 38 In subsequent versions (ICPC-2 and ICPC-2-R) the guidance changed but is not definitive. One statement indicates that health problems that cannot be expressed as symptoms, complaints or diagnoses should be labelled using rubrics such as A97 No disease or A98 Health maintenance/preventive medicine, both of which sit in Component 7. However, another statement some paragraphs later indicates that process codes are acceptable as problem labels: Some systems require that problems be coded only from components 1 and 7; others also accept codes from other components, so that if, for example, the patient attends for a tetanus immunization without a current injury, the problem could be coded as N

133 Such guidance is contradictory, perhaps in part due to the differing use cases for ICPC internationally. Further, these statements primarily relate to the use of preventive activities as problems managed. Although A98 Health maintenance/preventive medicine is an appropriate label for an episode of care in a problem list, it would not provide sufficient detail for the recording of the preventive activity managed at the encounter the check-up, immunisation, or advice given. The inclusion of processes of care as problems managed is highly subjective and scant evidence is available to support their inclusion or exclusion. It appears that most definitions used in the standards reported in Box have ignored the issue. Weed, the father of the POMR does not explicitly state whether processes of care should be included on problem lists, but does include surgical procedures including appendicectomy, breast removal and gall bladder surgery as examples of problem list entries. 11 This indicates that Weed regarded surgical procedures as valid entries on the problem list. GPs may not want all problems that meet the definition of problem/diagnosis to be included on the problem list. As reported in Section , the automatic inclusion of acute, self-limiting conditions on problem lists was seen by GPs during the workshops as a cause of frustration. One GP stated that he entered self-limiting problems (e.g. upper respiratory tract infections) in the reason for encounter data element (instead of the problem/diagnosis data element) so that these problems would not be included on the problem list. This provides a good example of the issues involved with recording a single piece of data for different purposes. In the example above, an upper respiratory tract infection (URTI) is a self-limiting condition that usually resolves without complication, and involves a short episode of care. However, URTIs can occur many times over a patient s lifetime, with each instance resulting in a separate episode of care. Even if each episode is marked with a resolution date (as is recommended in EHR standards 12 ) a self-limiting condition such as URTI may occur many times on a problem list. GPs in the workshops stated that problem lists and health summaries needed regular cleaning to remove these unnecessary entries, and regarded this as a timeconsuming but essential task. Another solution is to separate current 111

134 problems and past history. However, the RACGP standards for general practice for health summaries have combined these into a single medical history item. 142 The solution of including self-limiting conditions in a separate data element that is not linked to the problem list, as proposed by a GP in the workshops, could be regarded as a pragmatic solution to this problem, although a long term, more sustainable solution is needed. At present, there is no international standard definition for problem/diagnosis, and EHR vendors are able to create their own labels and definitions for concepts that are equivalent to problem/diagnosis. This lack of standardisation limits the interoperability of the data collected in different EHRs (that is, for information contained in one EHR to be transferred to another EHR and accurately interpreted by the receiving party). The European standard CEN 13940: Health informatics - System of concepts to support continuity of care is currently being revised for inclusion as an International Organization for Standardization (ISO) standard. 154 If accepted as an ISO standard, this initiative may encourage the standardised use of the term and definition for health issue (included in Box 5.4.1) on an international scale. The label health issue is broad, and would probably be sufficiently inclusive to satisfy most requirements. However, I am concerned that the current definition of health issue in this standard is ambiguous in terms of whether processes of care are valid for inclusion. I find the notes section of the definition particularly concerning. As previously stated in Box 5.4.1, the notes state: According to this definition, a health issue can correspond to a health problem, a disease, an illness. But it may not, such as when it is simply a request for a procedure (therapeutic or preventive) by the subject of care or another health care party etc. 134 The second sentence of this quotation is ambiguous. The phrase but it may not means that it (the health issue) may correspond to something other than a health problem, a disease, an illness. An example of an other item that may correspond to a health issue is given (i.e. a request for a procedure ), but no comprehensive definition of other inclusions is provided. In addition, the example states that requests for procedures may be health issues, but does 112

135 not state whether procedures themselves are valid for inclusion as health issues. Overall, the structure of this definition is poor. There is no explicit definition given for what can be included as a health issue that is not a health problem, disease or illness. This must be rectified in two ways. Firstly, the wording used in the notes must be clarified to remove the ambiguity in the second sentence caused by the phrase but it may not. Secondly, the definition of health issues that are not health problems, diseases or illnesses must be made explicit. The results from this study indicate that GPs do use processes of care to describe problems managed and that the use of processes as problems managed has increased over the last decade. This study was possible because GPs wrote the problems on BEACH structured encounter forms in free text which has no restrictions. In an EHR that uses a standardised clinical terminology or coding system, it is easier to restrict term usage, by limiting the terms available to a subset of valid terms. However, one of the principles in EHR data standards is that there should be a free text field linked to each coded field. 12 Although systems can prohibit users from entering processes of care as problems in a coded data element, GPs may well choose to leave the coded data element empty, and record their preferred terms in the associated free text field. In this event, the data are lost for interoperability, audits, statistics, and potentially for disease-related payments. It is also clear that effort is needed to further define the breadth of problem/diagnosis (or equivalent concepts) in national and international standards, especially in relation to the inclusion of processes and procedures in such data elements. The inclusion criterion in the NEHTA requirements that the implementation of a GP RefSet should not require GPs to change the current approaches they use (BR.GPRS.0009) indicates that users must be allowed to continue entering processes of care as problems if they currently do so. 113

136 5.5 Maintenance of a general practice terminology Background GPs who participated in the workshops described in Chapter 4 indicated that there were gaps in the content of terminologies or termsets they currently use in their EHRs (Section ). They described these gaps as two-fold terms not existing in a terminology, and not being able to easily find terms that do exist. These comments led to the creation of three separate maintenance requirements for the GPRS: the GPRS must be maintained updates should occur at least every six months there must be a clearly defined process for maintaining the GPRS (Section 4.5.1). There are three aspects of terminology maintenance: principles for maintenance maintenance processes implications resulting from maintenance. The GPRS requirements address the principles for, and process of maintenance, but do not cover issues regarding the implications of maintenance. In relation to the principles of maintenance, Cimino included terminology maintenance in his desiderata for controlled medical vocabularies, asserting that terminologies should gracefully evolve over time, defining graceful evolution as the need to ensure that users are not adversely affected by terminology updates, and that changes made are documented including clear reasons for the change. 49 Rector included change management as one of the reasons that creating and maintaining terminologies is hard. 50 The processes involved with updating various terminologies, including the Medical Entities Dictionary (MED), 155,156 the Read codes, 157 and SNOMED 114

137 CT 158 have been described in the literature. In addition, a number of studies have outlined proposals to organise the process of terminology maintenance, suggesting ways to standardise how changes in terminologies are represented, 159 or standardising the criteria for change and workflow mechanisms. 160,161 It is clear that the complexity of the terminology increases the difficulty of terminology maintenance, particularly when simultaneously maintaining multiple versions 157 or when dealing with complex hierarchical structures. 157,158 The implications of maintaining a medical terminology have been discussed, particularly in relation to SNOMED CT. Spackman described the changes in SNOMED CT in the three years after its inception, noting that the rate of changes made in SNOMED CT appeared to be slowing. 158 However there is a growing body of literature that describes the extensive implications of the bi-annual SNOMED CT updates within local implementations While the GPRS requirements address the principles of maintenance and state the need for a maintenance process, little evidence is available in the Australian setting about the content, process and implications of terminology maintenance. ICPC-2 PLUS, which was released in 1995, contains such information, although this has never previously been analysed. This section will therefore describe some terminology maintenance issues in the Australian general practice setting using ICPC-2 PLUS Aims The aim of this section is to describe the number and type of new term additions in ICPC-2 PLUS since , to determine the growth of the terminology since this time. To examine the growth of the interface used to access ICPC-2 PLUS terms, the number of keyword additions will also be investigated. This section will also identify the parties who suggest additions for ICPC-2 PLUS, and the frequency with which suggestions are accepted for inclusion in the terminology. 115

138 5.5.3 Methods The process of updating ICPC-2 PLUS is described in detail in Section 3.2. In this section, all dates used for the time periods shown are 1 May to 30 April in the following year. For example, means 1 May 2009 through to 30 April 2010 inclusive. This represents the yearly cycle of ICPC-2 PLUS updates, with the April update finalised for inclusion in the following year s BEACH study database. At present, the ICPC-2 PLUS April update is released to developers on 12 April, however the date of release has changed over time. I therefore decided to run all analyses for the year 01 May to 30 April to ensure that all April releases are included in the data period. I reviewed the release date of all April updates, ensuring that the April release had never been deferred until May. This approach also aligns with the reporting of data years in other sections of this thesis. Data reported in this section have been analysed from three sources: the ICPC-2 PLUS maintenance database; individual ICPC-2 PLUS CSV files generated from the maintenance database; and the ICPC-2 PLUS suggestions database. The methods used to analyse data using each of these sources are described below. In the maintenance database, all ICPC-2 PLUS terms have an attached term creation date. In Figure the number of new terms is derived from dates entered into the field TermCreatedDate in the table TblTerm in the ICPC-2 PLUS maintenance database. When terms in ICPC-2 PLUS should not be used for data entry, they are made inactive and can no longer be used for data entry by GP end users or by secondary data coders in BEACH. Terms are never deleted in ICPC-2 PLUS, to ensure that users have continued historical access to terms they have previously used in their EHRs. Terms made inactive are linked to a currently active term to ensure continuity of data across time. Prior to April 2004, a date was not attached to terms when they were made inactive in the maintenance database. It was therefore impossible to calculate the number of active ICPC-2 PLUS terms at a specific point in time prior to 2004 using this source. 116

139 Reporting inactive terms in the total size of the database artificially inflates the size of the terminology available for use. During each ICPC-2 PLUS release CSV files are exported from the maintenance database and saved. The ICPC2TRM CSV file contains a list of all terms present in ICPC-2 PLUS at the time of the release with their associated status, creating a fixed view of ICPC-2 PLUS terms at the time the CSV files were generated. I used the ICPC2TRM CSV file to calculate the number of active terms in each April release since April 1997 (reported in Figure 5.5.1). I cross-checked this result against the term Inactive date in TblTerm in the maintenance database for data from 2004 to 2010 to ensure this method was reliable. The percentage increase in the number of terms each year was calculated according to the following formula: (number of new terms/number of total active terms) x 100. To identify patterns in new term additions according to their ICPC-2 component (Figure 5.5.2), I used the maintenance database to extract all new terms added with their associated ICPC-2 component, and calculated the proportion of terms assigned to each component during each year. The date on which keywords were added to ICPC-2 PLUS was not included in the maintenance database until I therefore calculated this using the ICPC2KEY CSV files from the April release in each year. The ICPC-2 PLUS suggestions database was used to analyse the results reported in Tables and These results are restricted to the period The ICPC-2 PLUS suggestions database was created in April 2004, with suggestions prior to this being collated manually. Unfortunately, some accepted suggestions were added to the ICPC-2 PLUS maintenance database without being entered into the suggestions database, and therefore the number of suggestions in the database does not match the number of new terms added over this time. As the records of these suggestions were not kept, it is impossible to know exactly how many additional suggestions were made as a result of them. 117

140 5.5.4 Results New term additions Figure shows the number of new terms added to ICPC-2 PLUS and the total number of active terms in the terminology over fourteen years between and The figure demonstrates that the total number of active terms in ICPC-2 PLUS grew from 6,048 active terms in to 7,817 terms in , representing a 22.6% increase over the fourteen years. The number of new terms added to ICPC-2 PLUS decreased over the time period, indicating that growth slowed over time. The highest number of additions was in , with over one thousand new terms added, increasing the size of the terminology by 15.7%. There were 68 terms added in , representing the lowest number of terms added in a year and a 0.9% increase in the terminology from the previous year. In , there were 117 terms added, equating to an increase of 1.5% from (Figure 5.5.1). 118

141 Number of ICPC-2 PLUS terms No. of new terms(a) 0 1, No. of active terms 6,048 6,395 6,503 6,728 6,868 6,994 7,161 7,257 7,379 7,434 7,497 7,593 7,708 7,817 Figure 5.5.1: Size and growth of ICPC-2 PLUS, to (a) (a) Each year is calculated from 01 May to 30 April. Note: The number of active terms cannot be calculated by subtracting one year s active terms from another year s, due to variance caused by terms made inactive over the same period. A total of 2,688 terms were added to ICPC-2 PLUS between 1997 and 2010 (Table 5.5.1). When grouped according to their ICPC-2 component, diagnoses and diseases (Component 7) accounted for the highest proportion of new term additions (41.7%). Diagnostic and preventive procedures accounted for 23.9% of additions, and 22.2% of terms added were symptoms and complaints. New terms were rarely added to the administrative and results components of ICPC-2 PLUS. 119

142 Table 5.5.1: ICPC-2 PLUS terms added according to ICPC-2 component, ICPC-2 component name ICPC-2 component number Number of terms added Per cent of all new terms Symptoms and complaints Diagnostic and preventive procedures Medication, treatment, therapeutic procedures Results Administrative Referrals and other reasons for encounter Diagnoses and diseases 7 1, Total N/A 2, Notes: ICPC-2 International Classification of Primary Care, Version 2; N/A not applicable. The results presented in Table were analysed further to identify the presence of patterns in the addition of new terms, as a proportion of all terms added each year (Figure 5.5.2). There were few recognisable patterns, indicating that the addition of new terms does not occur in a systematic way. In almost all years, excluding and , diseases and diagnoses (Component 7) accounted for the highest proportion of new term additions. The addition of terms in Component 2 (diagnostic and preventive procedures) spiked in Two consultancies were completed by the FMRC at this time to improve the coding of pathology and imaging orders. New terms to describe pathology and imaging orders were created during the consultancy, accounting for the spike in Component 2 additions at this time. 166 New term additions in Components 6 (referrals and other reasons for encounter), 5 (administrative) and 4 (results) were consistently low throughout the study period. 120

143 Per cent Component 1 Component 2 Component 3 Component 4 Component 5 Component 6 Component Data year Figure 5.5.2: New term additions as a proportion of all terms added, by ICPC-2 component, to Keyword additions The number of keywords in ICPC-2 PLUS increased between and , from 3,167 to 4,520 keywords (Table 5.5.2). The largest increase was 9.2%, between and Growth then slowed to an increase of 1.9% in

144 Table 5.5.2: Growth in keywords in ICPC-2 PLUS, to Year (a) Number of new keywords Total number of keywords Percentage increase per year NAv 3,167 NAv , , , , , , , , , , , , , (a) Each year is calculated from 01 May to 30 April. Note: NAv not available. Term suggestions Between April 2004 and December 2010, there were 808 new terms entered into the ICPC-2 PLUS suggestions database (Table 5.5.3). Suggestions for new terms were split almost evenly between those made internally within the FMRC (49.3%) and external users (including other parties) (50.0%). Other parties included ICPC-2 PLUS developers who may request terms on behalf of their users, and external researchers using ICPC-2 PLUS in small projects. Also included in this category are changes needed to ICPC-2 PLUS due to updates in ICPC-2. The person or organisation making the request was not known for seven suggestions (0.7%). Table 5.5.3: Requests for new terms by requestor, Requestor Number of suggestions Per cent of suggestions made Family Medicine Research Centre Users Other (a) Unknown Total (a) Includes EHR software developers, researchers, and changes required through updates to the International Classification of Primary Care, Version 2 (ICPC-2). 122

145 Of the term suggestions entered into the database, two-thirds were accepted for inclusion in ICPC-2 PLUS (64.5%) (Table 5.5.4). The remainder of the suggestions were rejected (32.3%), or a decision about their inclusion in ICPC-2 PLUS was deferred until a later time (3.2%). Table 5.5.4: Results of new term suggestions, Result Number of suggestions Per cent of suggestions made Accepted Rejected Deferred Total Discussion The results presented support the notion that maintenance of a terminology is an important task, even for terminologies with long term implementations. It is clear from the results that in the 17 years since its first release, ICPC-2 PLUS has grown substantially, both in the number of terms available (Figure 5.5.1) and in the keywords used to access terms (Table 5.5.2). Figure indicates that there was significant growth in ICPC-2 PLUS in This was mostly due to the transition from ICPC-1 to ICPC-2. One of the major changes in ICPC-2 was the inclusion of a set of criteria in each rubric in Components 1 and 7. In ICPC-2 PLUS all inclusions in each ICPC-2 rubric were included as terms in the PLUS terminology, and this action was responsible for the majority of the additions in Spackman reported that the number of new active concepts added to SNOMED CT slowed in the three years following its first release in The results of the current study are consistent with Spackman s findings. In ICPC-2 PLUS growth slowed over the years for both new term and keyword additions. However the growth of the terminology is still noteworthy, with additions of 1.5% of the size of the terminology in new terms, and 1.9% in keywords in Given that terms in Components 1 and 7 were the most often used in ICPC-2 PLUS (Section 5.3), it is not surprising that diagnoses and diseases 123

146 (Component 7) and symptoms and complaints (Component 1) were two of the components with the highest number of additions. There lacks an overall pattern or trend in the addition of new terms in ICPC-2 PLUS, when considered as a proportion of all terms added each year (Figure 5.5.2), indicating that each update is unique and driven by user needs at the time. There were two notable results supporting this conclusion. In and there was a spike in the proportion of term additions in Component 2 (diagnostic and preventive procedures). This component includes pathology and imaging ordering terms. At this time the FMRC was contracted by the Australian Government Department of Health and Ageing to prepare a feasibility study about pathology coding and classification systems in general practice, and to enhance the pathology ordering codes in ICPC-2 PLUS. 167 This was followed in 2000 by a consultancy to develop codes for imaging orders in general practice. 168 It is clear from the figure that these consultancies led to a much higher proportion of new terms added in Component 2 than in previous or future years. Similarly, in there was a spike in the proportion of codes in Component 3 (medication, treatment and therapeutic procedures). This spike was due to a higher proportion of terms added from users in Aboriginal and Torres Strait Islander (ATSI) health centres. The consistently low proportion of additions made to Components 4, 5, and 6 may indicate that ICPC-2 PLUS is sufficiently complete in these areas, or more likely, that terms classified to these components are not used often. The suggestions database provided interesting insight into the parties influencing the development of ICPC-2 PLUS. It is an unusual situation that an organisation is both a developer and a user of a terminology, and undoubtedly this has a significant influence on ICPC-2 PLUS development, with half of the suggestions being internal (Table 5.5.4). Internal suggestions came primarily from the BEACH study, where GPs write free text reasons for encounter and problems managed that are secondarily coded by staff at the FMRC. This acts as a valuable source of information about trends in medical terminology use, including the identification of new terms being used by GPs, 124

147 and changing patterns of terminology use. Another advantage is that the FMRC can respond quickly to such changes without users needing to make suggestions, which may pre-empt some suggestions from users. The result that one-third (32.4%) of suggestions were rejected (Table 5.5.4) is similar to the results of a 1997 evaluation of suggestions for the Read coding system, which stated that 35% of requests for new codes were not added, primarily because the code already existed, but also due to ambiguous or precoordinated suggestions, or due to the incompatibility of the suggestion with the structure or purpose of the Read coding system. 157 Although these results are similar in the proportion of rejected suggestions, the reasons for these rejections differ. ICPC-2 PLUS is an interface terminology designed for use in general practice. As such, the Classification Committee assesses each suggestion against a set of criteria outlined in Section 3.2, one of which is the suggestion s relevance in general practice. When maintaining any terminology, developers must be aware of both intended and actual use of the terminology. In ICPC-2 PLUS, the intended use is for general practice EHRs. However, the implementation of ICPC-2 PLUS has been broader than this, and ICPC-2 PLUS is regularly used within specialist, allied and community health, in the Aboriginal health care sector and in prisons. The balance between its original intended use and actual use must be considered carefully, respecting both the original and actual uses. ICPC-2 PLUS is used in some areas due to the absence of an appropriate and useable terminology for these specialty areas. It is intended that the Australian version of SNOMED CT (SNOMED CT-AU), be used across the entire Australian health care setting. At present the breadth of SNOMED CT-AU is too wide to be useful in most settings. I hypothesise that the development of a SNOMED CT-AU general practice RefSet may assist with its implementation in such specialties, using the SNOMED CT-AU hierarchies to access more specific concepts when needed. However, there were other reasons for rejection of new term suggestions. Sometimes users suggest new terms that already exist in the terminology, or 125

148 are synonymous with an existing term. Although ICPC-2 PLUS does support the inclusion of synonyms, the Classification Committee sometimes decides that the suggestion should be included as a keyword to the original term, rather than as a new term itself, particularly when the suggestion has been made by one individual or arose once only in BEACH. Similarly, if the term already exists in ICPC-2 PLUS, the Committee concludes that there is an issue accessing the term and considers whether additional keyword links are required. As such, the high proportion of rejected suggestions for new terms does not indicate that these suggestions are not dealt with, but may be managed in an alternate way. The results of suggestions are always communicated to the user(s) who made the suggestion, and users are given the opportunity to re-submit suggestions with additional justification for the next release. There are some limitations to this study. ICPC-2 PLUS is a reasonably flat terminology with a bi-axial structure based on the ICPC-2 classification. As such, there are few structural implications in its maintenance, so I was unable to assess this aspect of maintenance. In addition, it would be interesting to have recorded the reasons for new term suggestions from the users perspective however this information is not collected when users make suggestions. The limitations of the ICPC-2 PLUS suggestions and maintenance databases have limited my ability to analyse some information. The maintenance database was designed and created in 1997, and while major updates have occurred over the years, there are some limitations in the database design that I have not been able to overcome. For example, the date on which terms were made inactive was not included as a data element in the database until 2004, limiting my ability to study terms made inactive prior to this time. There is no doubt that terminology maintenance is resource intensive, 155,162,163 and the results of this study indicate that terminology maintenance must be considered as an ongoing task. As stated by Rector, terminologies are not static entities. 50 Even if the existing content and structure of a terminology 126

149 stays the same there are always changes due to improving medical knowledge, or new diseases occurring. GP participants in the workshops discussed the frustration of not being able to find appropriate terms, and described this issue as twofold terms not existing in the terminology, and not being able to find terms. This emphasises the importance of maintaining the interface to the terminology, as well as the terminology itself. 127

150 5.6 Discussion This research describes the current use of clinical terminologies in Australian general practice, using ICPC-2 PLUS as a research tool. Results indicated that ICPC-2 PLUS provides a good reflection of the terms used by Australian GPs, and is therefore a suitable tool with which to assess clinical terminology use in Australian general practice. There is a skewed distribution of term use, with GPs using a few terms a lot of the time. Terminology developers face a difficult task in balancing terminology size, maintenance schedules and user satisfaction. Clearly the most frequently used terms must be included in a clinical terminology. However, determining an appropriate cut-off for the remainder of the terminology (which was shown to contain many terms used infrequently) is a difficult undertaking, and is highly subjective. There is no algorithm or formula that terminology developers can use to ensure that their terminology is of an appropriate size the only measure available in a practical sense is user satisfaction. Terminology developers must therefore conduct internal reviews of terminology content (where possible), encourage feedback from users and demonstrate willingness to respond to feedback received Labelling problems managed It has been acknowledged for approximately 50 years that the clinical spectrum seen in general practice differs to that seen in secondary or tertiary care. 26,28-29 The acceptance that a large proportion of GPs by necessity use symptom labels to record problems was largely responsible for the development of classifications and terminologies specifically for use in general practice (such as ICPC) rather than continued reliance on the ICD. ICD has a basis in hospital and mortality statistics and places a far greater emphasis on diseases, than on symptoms. The reasons ICD is not suitable for use in general practice were described in Chapter 1. The concepts disease and diagnosis differ markedly, although they are often used interchangeably (for example, there is a component in ICPC-2 labelled Diseases and diagnoses ). The Wonca Dictionary for General/Family Practice defines disease as a physiological or psychological dysfunction on 128

151 the basis of well-known symptoms and signs or with a well-known aetiology. 15 In contrast, the term diagnosis refers to the process of determining the name given to a health problem, which may be a disease or a symptom. 15 That these terms can incorrectly be used interchangeably is one of the reasons I primarily use the term problem managed in this thesis, rather than disease or diagnosis. The definitions above apply when the terms diagnosis or disease are used in this thesis. This chapter has focussed on the terms used by GPs to describe clinical concepts, particularly those relating to problems managed. There are a number of complex issues in the labelling of medical conditions, particularly because diagnostic labels in general practice are often working diagnoses. 169 In the early 1970s Howie theorised that the care pathway taught to medical students (primarily through hospital based training) did not prepare them for general practice. Howie suggested that the general practice care pathway often follows the pattern symptom treatment, rather than symptom diagnosis treatment, suggesting that the symptoms and signs presented by the patient may sometimes provide a better indicator of the treatment course than the disease label assigned by the clinician (called the symptom-sign complex). 170 This may partially explain why GPs often label some problems as symptoms rather than diagnoses. Since the 1970s a number of researchers have attempted to describe the diagnostic process in general practice. 169,171,172 A hospital contact may incorporate a reason for presentation, investigation, diagnosis of a disease and treatment, 170 in a single entity usually called an admission. In contrast, in general practice a disease label is often not reached after a single encounter. From the episode of care perspective, the first presentation of a problem may often be presented at the level of a symptom, with investigations (e.g. pathology or imaging tests) occurring outside the consultation, followed by patient re-attendance to determine the outcome of investigations. 173 A label that is equivalent to a disease may then be assigned to the problem, or, where investigations have been inconclusive, the problem may continue to be expressed as a symptom. 129

152 Howie s symptom-sign complex hypothesis has been supported over the last 40 years, with evidence that 40 50% of problems managed at general practice encounters are not labelled as diseases. 171,174,175 Results from the current study differ, with two-thirds of problems labelled as diseases and one-third labelled as symptoms or processes of care, according to the differentiation of symptoms and diagnoses from ICPC-2. Differences between the current study and prior studies are marked. All three previous studies were relatively small in comparison to the sample size used in BEACH, on which the current study was based. The denominator used is not clear in two of the earlier works one refers to consultations as the denominator rather than problems/diagnoses, 175 and another states that they reviewed consultations, and then refers to cases, but does not define whether their analysis is patient based or consultation (or encounter) based. 171 A clear statement specifying the denominator used is vital in drawing conclusions from each of these studies and the lack of a clear denominator leads to questions about the validity of the conclusions. None of the three previous studies define the concept of diagnosis, and this is another factor that makes it difficult to make valid comparisons of my results with theirs. The current study used the label of diagnoses and diseases as described in ICPC-2 and it is possible that this label may contribute to the higher proportion of disease labels reported in this study. In ICPC the concept diagnoses and diseases is not explicitly defined, except insofar as each rubric contains criteria for inclusion or exclusion. Some rubrics included in the diagnoses and diseases component are not diseases, such as Elevated blood pressure (K85). This is a sign, defined as elevated blood pressure that does not meet the criteria for hypertension (two or more readings of elevated blood pressure per encounter, taken at two or more encounters). This suggests that the use of the ICPC-2 component labels to differentiate between symptoms and diagnoses is somewhat flawed. This flaw has been noted by WICC, who have corrected the placement of some rubrics in the ICPC-2 pager (as described in Section 3.2.2). The higher proportion of disease labels in my study might also be partially explained by the ongoing management of chronic conditions that have been 130

153 labelled as a disease at an earlier encounter. Recorded encounters are therefore for ongoing disease management rather than establishment of a disease label. In Australia, the management of chronic problems at general practice encounters has increased over the last decade from 49.3 per 100 encounters to 53.1 per 100 encounters Certainty of diagnosis In our discussions with GPs about the scope of a general practice RefSet of SNOMED CT, many emphasised the importance of being able to indicate that a diagnosis is provisional or differential, with participants stating that they are hesitant to record a diagnosis unless they are certain that the patient has the stated condition, particularly if the recorded diagnosis then automatically appears on the problem list. These concerns were acknowledged in the GPRS requirements, which stated that the GPRS must be sufficiently comprehensive to allow the recording of attributes and contextual information, including certainty of diagnosis (BR.GPRS.0001). In describing the scope of the POMR, Weed states that provisional or differential diagnoses should not appear on the problem list, advocating that problems recorded should be limited to the level of information known to the clinician. 11 Likewise in BEACH, GPs are not asked to indicate whether a diagnosis is provisional or confirmed, but to record the problem at the most specific level possible based on information known to them at the time. As such, it could be expected that GPs should not record provisional or differential diagnoses in BEACH. However, the free text analysis indicated that 14.3% of free text terms recorded in BEACH had an additional piece of information attached to the coded problem managed that related to the certainty of the diagnosis (Table 5.2.3). Clearly GPs have mixed methods of recording certainty of diagnosis. On one hand the results presented in Section 5.4 suggest that a considerable proportion of GPs choose to report problems managed as symptoms or processes rather than diseases, supporting Weed s proposition that problems should be recorded at the level of information known to the clinician. On the other hand, this is not done consistently, with another group of GPs adding 131

154 question marks or other indications of uncertainty to the problem label they have recorded Conclusion Few have examined the relationship between the content and use of clinical terminologies in a formal manner. Most terminology developers do not have easy access to data about the use of their terminology and thus their ability to assess the use of the clinical terminology they develop is very limited. Over time, many statements about the use of clinical terminologies in clinical settings have been made, and some of these have become self-evident truths, whereby the statement is made so often that an assertion is made that it is fact, when there is no evidence base to support the statement. Some of these self-evident truths were stated during the workshops with GPs and software vendors about the size of terminology, the types of terms used and the need for a terminology maintenance cycle. This chapter has created a much-needed evidence base for some of the commonly held assertions about the use of clinical terminology in Australian general practice. There were other research questions identified during the workshops (Chapter 4) that I have not investigated in this chapter. For example, the differing use of terminology in rural and metropolitan areas was discussed during the workshops, but has not been investigated in this study. Additional topics not covered in this chapter will be considered for future study outside this thesis. The development and planned uptake of SNOMED CT marks a turning point in the use of standardised clinical terminologies. SNOMED CT is marketed as an international clinical terminology for use in all clinical settings. Having identified patterns of terminology use in Australian general practice using ICPC-2 PLUS as a research tool, in the next chapter I will evaluate the suitability of SNOMED CT-AU for use in Australian general practice, in light of the terms contained in ICPC-2 PLUS. 132

155 6 Mapping terms from ICPC-2 PLUS to SNOMED CT-AU 6.1 Introduction In this chapter I assess the suitability of the content of SNOMED CT-AU in Australian general practice to determine whether it is suitable or fit for purpose in this setting. In this thesis, fit for purpose is defined as well equipped or well suited for its designated role or purpose. 176 SNOMED CT is promoted as the most comprehensive, multilingual clinical terminology in the world. 43 There is one important word included in this statement relating to the purpose of this thesis: comprehensive, which implies that SNOMED CT contains all content needed in clinical use. This chapter will assess the extent to which it meets that expectation in one clinical specialty in one country general practice in Australia. SNOMED CT-AU is a version of SNOMED CT that has been built for Australian use cases, and includes content specific to Australian use. Each country has its own cultural differences in medicine. For example, outside Australia it is virtually impossible to need medical treatment for a brown snake bite, unless you are in a zoo. SNOMED CT-AU therefore has the potential to increase the suitability of SNOMED CT in Australia by adding content needed for use in Australia. There is very little information publicly available about the current use of SNOMED CT or SNOMED CT-AU in Australia in clinical settings. My review of the literature showed that SNOMED CT is available in hospitals that use EHRs produced by Cerner, 177 and in 2010 was actively used in 41 emergency departments using the Cerner FirstNet product. 178 However, the implementation of SNOMED CT in FirstNet has been severely criticised. 179 SNOMED CT is being used in at least one hospital intensive care unit. 180 I have found no evidence of current use in Australian general practice. Internationally, clinical use of SNOMED CT is also limited. A 2008 study of electronic health and medical record vendors in the US reported that one third of respondents indicated they had implemented SNOMED CT in their clinical systems. 181 However, implementation was not defined in this survey, and no 133

156 information was given about the extent to which SNOMED CT was used in these clinical systems. In 2010 Elhanan et al conducted a survey of SNOMED CT users. The 215 respondents were primarily researchers and educators (48% of respondents), with physicians comprising only 25% of respondents. Two-thirds of respondents (65%) were using SNOMED CT for research purposes, and half were using it for EHR system development. Only 39% of respondents indicated that they used SNOMED CT in live systems. 182 The authors acknowledge this study had limitations, particularly in relation to the characteristics of respondents, who were primarily from research institutions, and many were from the US. These results therefore cannot be generalised to the population using SNOMED CT internationally. Clinical implementations of SNOMED CT are primarily found in the US, with users at the health maintenance organisation Kaiser Permanente, 183 the US Department of Veterans Affairs, 184 the Mayo Clinic 185 and the University of Nebraska Medical Center. 186 SNOMED CT is also used at the Hospital Italiano de Buenos Aires in Argentina. 187 Prior to implementation, a clinical terminology should be evaluated to ensure that its structure and content is suitable for the relevant clinical setting. Many evaluations of SNOMED CT have been published relating to different clinical use cases. In general practice however, only one relevant evaluation is available. A small Canadian study involving one general practice evaluated the extent to which terms included on problem lists were represented in SNOMED CT, concluding that approximately 92% of the terms could be located in SNOMED CT. 188 In , the FMRC in collaboration with the Sydney Language Technology Research Group (University of Sydney) conducted a trial study to investigate the extent to which Australian general practice terms from ICPC-2 PLUS were represented in SNOMED CT using a variety of automated mapping techniques. I was involved in this study, and the results were published by Wang et al

157 Of the 7,410 active ICPC-2 PLUS terms at that time, 60.3% were mapped to SNOMED CT using the automated mapping program. After manual evaluation of results to determine the accuracy of the automated matches, the proportion of matches decreased to 45.9%. 81 The project became a valuable first step in learning about the methods and issues involved with mapping clinical terminologies. However, this study was not intended to result in the clinical implementation of a map from ICPC-2 PLUS to SNOMED CT. The vast majority of GPs in Australia currently use computers for clinical purposes. 9 Findings previously presented in Chapter 4 describe the current use of clinical terminologies in general practice EHRs. At present, all GP EHRs used in Australia contain structured termsets/codesets or clinical terminologies. The introduction of SNOMED CT-AU will therefore require a change management process during which the termsets and terminologies currently used in Australian general practice EHRs are used in conjunction with, or converted to, SNOMED CT-AU. The process through which one clinical terminology is linked to another clinical terminology is called mapping. Mapping is defined as a relationship between the code and term used to represent a health concept in one system, and the code or term that would be used to represent the same concept in another coding or terminology system. 16 Terminologies are usually created for a particular use case or purpose. However, the data coded using a particular terminology may be re-used for other purposes. Rather than re-coding every record using the new terminology, implementing a map from one clinical terminology to another facilitates the re-use of clinical information in a standardised manner, and enables linkage with historical data collected in one terminology after it is mapped to a second terminology. The GPRS requirements (described in Chapter 4) stated that the content in the GPRS would be based on terminologies and termsets currently used in Australian general practice, including: the Classification and Terminology of Community Health (CATCH) 135

158 the University of Adelaide s GP termset (created using terms from the Medical Director software) ICPC-2 PLUS the termset used in the Best Practice software. Developing a RefSet using this approach necessitates mapping the content of each of these termsets to SNOMED CT-AU, to identify concepts in SNOMED CT-AU that are equivalent to the terms/concepts in the source termsets. This work was supposed to be undertaken as part of the GPRS project, for which I was the project manager. In September 2010 NEHTA cancelled the GPRS project before the mapping work had commenced. However, the need for a map from ICPC-2 PLUS terms to SNOMED CT-AU was still a priority, particularly for GP software vendors who wanted to implement SNOMED CT-AU in some form. I therefore decided to continue with the map development, to determine the extent to which terms from ICPC-2 PLUS could be mapped to SNOMED CT-AU, and therefore assess the suitability of SNOMED CT-AU in Australian general practice. 6.2 Aims 1. To assess the coverage of SNOMED CT-AU from the perspective of the general practice clinical domain, identifying the extent to which SNOMED CT-AU is fit for purpose and suitable for implementation in the area of symptoms and complaints, and diagnoses/diseases in Australian general practice. 2. To assess the extent to which ICPC-2 PLUS terms used in Australian general practice are explicitly represented in SNOMED CT-AU as preferred terms or descriptions. More specifically, in this chapter I aim to: examine the extent to which there is a relationship between the validity of matches in the map from ICPC-2 PLUS to SNOMED CT-AU and the 136

159 frequency with which ICPC-2 PLUS terms are used in the BEACH data elements of reason for encounter and/or problems managed. identify the extent to which the mapping results from ICPC-2 PLUS to SNOMED CT-AU contain matches that are acceptable for clinical implementation. identify content areas in SNOMED CT-AU that are deficient from the perspective of Australian general practice, including missing concepts and descriptions. 6.3 Methods Tooling Software programs to facilitate mapping, known as mapping editors or mapping tools, have increased in both number and quality in recent years. The Commonwealth Scientific Industrial and Research Organisation (CSIRO) have developed a software tool called Snapper that supports mapping from local terminologies (such as ICPC-2 PLUS) to SNOMED CT. 189 Snapper was used by NEHTA to create the Australian Emergency Department Reference Set (EDRS) 107 and is gaining recognition from NEHTA and the IHTSDO as a valid tool for the mapping of local terminologies to SNOMED CT. I therefore chose Snapper as the software tool for the map from ICPC-2 PLUS terms to SNOMED CT-AU. A software licence for Snapper was obtained by the FMRC from the CSIRO for this work. Subsequently in 2011 NEHTA purchased a national licence for the use of Snapper, and it is now available free of charge to Australian users. 190 The map was created using the most recent releases of the terminologies available at the start of the mapping process: the April 2011 release of ICPC-2 PLUS and the November 2010 SNOMED CT-AU release Map result categories I manually evaluated each mapping result to determine the quality of the automated match between ICPC-2 PLUS and SNOMED CT-AU, and assigned a map result category from the options listed in Box 6.1, with definitions. 137

160 Box 6.1: Categories used to evaluate mapping results Map result category Abbreviated map result category Definition Exact E A lexical and semantic match from ICPC-2 PLUS to a SNOMED CT-AU fully specified name. Stop words (including and, or etc) and word order are excluded and do not affect the evaluation of a result as an exact match. Description D The ICPC-2 PLUS term has a lexical and semantic match to a SNOMED CT-AU preferred term or description (including designated SNOMED CT-AU synonyms). Stop words (including and, or etc) and word order are excluded and do not affect the evaluation of a result as a description match. Synonym S The ICPC-2 PLUS term is synonymous with a SNOMED CT-AU description (which may be a fully specified name or preferred term) but is not specifically included in SNOMED CT-AU. The category of synonym in this context is not equivalent to the definition of a synonym explicitly included in SNOMED CT-AU. Best fit BF The ICPC-2 PLUS term does not have an equivalent lexical or semantic match and is not synonymous with an existing SNOMED CT-AU concept in the form of a fully specified name, preferred term or description. The source ICPC-2 PLUS term has been mapped to a SNOMED CT-AU concept that is broadly comparable to the source term. These maps should be regarded with caution. Best fit precoordinated BF/PC The ICPC-2 PLUS term is precoordinated with a qualifier (precoordination by qualification) and an equivalent precoordinated concept is not present in SNOMED CT-AU. The term has been mapped to the equivalent SNOMED CT-AU kernel concept, and requires postcoordination to be completed. No match NM There was no equivalent match found in SNOMED CT-AU for the ICPC-2 PLUS term. This term is not conceptually captured in SNOMED CT-AU Mapping principles I decided to focus on clinical conditions rather than processes of care in this map, as these are coded most often in general practice, and I therefore regard them as the highest priority for harmonisation with SNOMED CT-AU. so I limited the mapping to those ICPC-2 PLUS terms in ICPC-2 Components 1 (symptoms and complaints) and 7 (diagnoses and diseases). During the first round of automated mapping, target concepts in SNOMED CT-AU were restricted to the following foundation RefSets (equivalent to top level hierachies in the SNOMED CT international releases): clinical findings events situations with explicit context. 138

161 This restriction served to limit the scope of the target SNOMED CT-AU concepts to those likely to contain matches for the source terms. If an automated match could not be found in one of the foundation RefSets listed above, I manually searched for the source term among all SNOMED CT-AU concepts, regardless of the foundation RefSet. In ICPC-2 PLUS, mapping was restricted to terms with a status of active, and in SNOMED CT-AU, to concepts with a status of current. All map results have a relationship of one ICPC-2 PLUS term to one SNOMED CT-AU concept, creating a one-to-one map of best fit. However, due to its relationship to ICPC-2, ICPC-2 PLUS has some precoordinated content (for example, in the male and female genital chapters). As discussed in Section 3.3.2, SNOMED CT-AU does permit the user to create postcoordinated expressions. However, implementation of postcoordinated expressions has not occurred widely, and none of the GP EHR vendors in Australia have indicated they are willing to implement postcoordination. Rather than creating postcoordinated expressions in this map, I decided to create a map of best fit by mapping to one SNOMED CT-AU concept only, which may be a precoordinated concept. There are two forms of precoordination: precoordination by combination, in which two separate kernel terms are combined in a single precoordinated concept (e.g. ischaemic heart disease with angina). precoordination by qualification, in which a qualifier (representing severity, laterality or chronicity for example) is combined with a kernel concept to create a precoordinated concept (e.g. severe asthma). 81 Source terms in ICPC-2 PLUS that were precoordinated by combination, for which an equivalent precoordinated concept did not exist in SNOMED CT-AU were assigned a map result category of no match. ICPC-2 PLUS source terms that were precoordinated by qualification were mapped to the kernel concept found in SNOMED CT-AU, and marked with a map result category of best fit precoordinated to indicate that the map is not complete due to precoordination in the source term. 139

162 6.3.4 Mapping process All active ICPC-2 PLUS terms belonging to ICPC-2 Components 1 and 7 were loaded into Snapper. The mapping process was as follows: 1. I ran the automap with scope function in Snapper to generate the initial automated mapping results. This method used a lexical search algorithm to identify potential target mappings in SNOMED CT-AU for each source term, allowing me to limit the scope of target concepts to a single SNOMED CT-AU foundation RefSet. For the initial mapping I restricted the scope to the Clinical Findings foundation RefSet (equivalent to the Clinical Findings hierarchy in the SNOMED CT international release). 2. Once the automap with scope function had retrieved all possible mappings, I manually evaluated all results produced to ensure that the automatically generated mapping results were accurate, and assigned a map result category (as described in Box 6.1) to each completed mapping result. 3. For those ICPC-2 PLUS terms for which an automap was not found, I manually searched for an appropriate match in SNOMED CT-AU using the manual search functionality in Snapper. This involved entering synonyms for the source term and/or navigating the SNOMED CT-AU foundation RefSets. A map result category was also assigned to each mapping result found manually. 4. The remaining ICPC-2 PLUS source terms without a mapping result had not been mapped for one of two reasons: I had found a possible map, but needed clinical input from my GP supervisor (who also has experience in classifications and terminologies) to confirm that the possible map was correct; or I could not find an equivalent match in SNOMED CT for the source term. I consulted my GP supervisor to assist in finding matches for these remaining source terms. All matches were assigned a map result category. 140

163 5. At the completion of this process, any remaining source terms for which a match could not be found in SNOMED CT were assigned the map result category of No match. 6. Mapping results were exported from Snapper for analysis and reporting as CSV files and Release Format 2 (RF2) files Analysis For the purposes of this study, the terms usage and utilisation refer to the frequency usage with which ICPC-2 PLUS terms were used in the coding of the RFE and/or problems managed data elements in the BEACH study. The mapping files were analysed in combinations of the following groupings for different purposes: map result category SNOMED CT-AU semantic tag ICPC-2 chapter ICPC-2 component. To determine the extent to which there was semantic equivalence in the mapping results from ICPC-2 PLUS to SNOMED CT-AU, I grouped the map result categories into acceptable and unacceptable matches. The results of this analysis are presented in Section These grouping were defined according to the following criteria: Acceptable matches were those regarded as a semantic match between an ICPC-2 PLUS term and a SNOMED CT-AU concept description (fully specified name, preferred term or synonym), and equated to those matches with a map result category of exact, description or synonym. Unacceptable matches were those for which no precise or complete semantic match was identified in SNOMED CT-AU to represent a source ICPC-2 PLUS term. Unacceptable matches were assigned a map result category of best fit, best fit precoordinated or no match. 141

164 The results presented in Section include modified versions of the utilisation data from the BEACH study presented in Chapter 5, which was limited to the BEACH years 2004 to However, the mapping was undertaken using terms from the April 2011 ICPC-2 PLUS release, which contained more terms than were available for use in BEACH between 2004 and Data from two additional years ( and ) were added to the five years of data previously presented in Chapter 5 to include ICPC-2 PLUS terms that were added and used in the latter two years. The results presented in Section therefore include data collected about RFEs and/or problems managed in the BEACH study between April 2004 and March Since mapping from ICPC-2 PLUS to SNOMED CT-AU was limited to terms from ICPC-2 Components 1 (symptoms and complaints) and 7 (diseases and diagnoses) all ICPC-2 PLUS process terms were removed from the utilisation data. New ICPC-2 PLUS terms are added into the BEACH database once a year at the start of the new data year in April, and contain all ICPC-2 PLUS terms added during the previous year. For this analysis, BEACH data were available for ICPC-2 PLUS terms added until the April 2010 ICPC-2 PLUS release. ICPC-2 PLUS terms added after April 2010 and before April 2011 (n=16 in Components 1 and 7) were therefore excluded for this analysis. Having viewed demonstrations of some of the general practice clinical software incorporating ICPC-2 PLUS, I know anecdotally that ICPC-2 PLUS is implemented most often in the problems managed data element, and is rarely used to record RFEs. In Chapter 5, the utilisation data in BEACH were reported separately for the RFEs and problems managed data elements. In this study I report the utilisation data in two ways: combining the utilisation data for RFEs and problems managed to provide a single reference point for linkage with the mapping results. for problems managed alone, as most general practice software vendors limit the use of ICPC-2 PLUS for this purpose. 142

165 The usage data were sorted in descending order of frequency in a Microsoft Excel spreadsheet, and the cumulative frequencies were calculated. The cumulative frequency results were divided into five groups. As shown in Figures and 5.3.2, a small number of terms accounted for a large proportion of usage in both the RFE and problems managed data elements in BEACH. To account for the skewed distribution, the groups were divided according to a series of criteria based on the frequency distribution of terms used. Group 1 contained the ICPC-2 PLUS terms that cumulatively accounted for the top 50% of term frequencies in the RFE and/or problems managed data elements in BEACH. Each subsequent group was divided into half of the previous group, as follows: Group 2 contained the next 25% of terms according to the frequency distribution for RFEs and/or problems managed, equating to terms that cumulatively accounted for between 50.1% and 75.0% of terms in the frequency distribution. Group 3 contained the next 12.5% of term frequencies that cumulatively accounted for % of terms in the frequency distribution for RFEs and/or problems managed. Group 4 contained the group of ICPC-2 PLUS terms that accounted for the next 6.25% ( %) of term frequency according to the frequency distribution. Group 5 contained the remaining 6.25% of ICPC-2 PLUS terms that cumulatively accounted for % of term frequency according to the frequency distribution. A sixth group was created for those ICPC-2 PLUS terms in ICPC-2 Components 1 and 7 that were never used in the RFE or problem managed data elements in BEACH. Mapping results were subsequently matched with utilisation data to facilitate a combined analysis between the validity of matches and term utilisation. To calculate the association between map validity and utilisation, the map result categories were collapsed into three groups: acceptable matches (as 143

166 described in Section 6.4.2), unacceptable matches (those categorised as best fit or best fit precoordinated) and no matches. 6.4 Results Overall mapping results There were 5,453 ICPC-2 PLUS terms from ICPC-2 Components 1 (symptoms and complaints) and 7 (diagnoses and diseases) available for mapping to SNOMED CT-AU. A match was found in SNOMED CT-AU for 4,987 terms (91.5%). There was no valid match found in SNOMED CT-AU for 8.5% of ICPC-2 PLUS terms (Table 6.1). When separated into map result categories, an exact match to a SNOMED CT-AU fully specified name was found for 32.0% of ICPC-2 PLUS terms, and maps to SNOMED CT-AU descriptions were identified for a further 15.5% of ICPC-2 PLUS terms. Together, these results indicate that 47.5% of ICPC-2 PLUS terms were represented in SNOMED CT-AU as both semantic and lexical matches. Equivalent synonyms were found in SNOMED CT-AU for 22.1% of ICPC-2 PLUS terms, indicating that a concept was semantically included in SNOMED CT-AU without being explicitly included as a preferred term or description. The remaining ICPC-2 PLUS terms (21.9%) were best fit matches in SNOMED CT-AU. This category was broken down into precoordinated best fit matches (4.5% of the total), and other best fit matches (17.4%) (Table 6.1). Matches categorised as best fit should be viewed with caution, as a best fit match is not synonymous with an existing concept in SNOMED CT-AU. Table 6.1: Overall mapping results, ICPC-2 PLUS to SNOMED CT-AU Mapping result category Number Per cent Exact match 1, Description match Synonym match 1, Best fit all 1, Best fit match Best fit precoordinated match No match Total 5,

167 Figure 6.1 represents the mapping results grouped according to ICPC-2 chapters (largely based on body systems) and map result categories. The figure demonstrates that there was considerable variation between chapters in the validity of matches from ICPC-2 PLUS terms to SNOMED CT-AU. The highest proportion of exact matches was found for terms in the Neurological (45.6%), Circulatory (42.5%) and Respiratory (40.6%) chapters of ICPC-2. The proportion of ICPC-2 PLUS terms for which no match could be found in SNOMED CT-AU was highest in the ICPC-2 Social chapter (48.5%), followed by the Pregnancy/family planning (16.3%) chapter. The proportion of ICPC-2 PLUS terms for which a best fit match was found in SNOMED CT-AU was highest in the General and unspecified (27.2%), Ear (24.4%) and Endocrine/metabolic/nutritional (22.5%) chapters. The proportion of matches categorised as best fit due to precoordination in the ICPC-2 PLUS term were highest in the Male genital (30.5%), Female genital (22.1%) and Pregnancy/family planning (17.7%) chapters. Most of the ICPC-2 chapters are based on body systems, including separate chapters for the male and female genital systems. However, many symptoms or diseases in the genital chapters can be experienced by both males and females, creating concepts that include the symptom or disease precoordinated with patient sex. As such, it is not surprising that the best fit precoordinated category was used most often in the mapping for the Male genital, Female genital and Pregnancy/family planning ICPC-2 chapters. 145

168 ICPC-2 chapter A B D F H Exact Description Synonym Best fit Best fit-precoordinated No match K L N P R S T U W X Y Z Per cent Note: ICPC-2 chapters: A General and unspecified; B Blood and blood forming organs; D Digestive; F Eye; H Ear; K Circulatory; L Musculoskeletal; N Neurological; P Psychological; R Respiratory; S Skin; T Endocrine/metabolic/ nutritional; U Urinary; W Pregnancy and family planning; X Female genital; Y Male genital; Z Social. Figure 6.1: Distribution of mapping results within each ICPC-2 chapter, by map category 146

169 Figure 6.2 displays the mapping results according to map result categories for each ICPC-2 component. In this figure the best fit precoordinated category has been combined with the best fit category, and is reported as best fit all. The figure demonstrates that the pattern of mapping results varies considerably between ICPC-2 component 1 (symptoms and complaints) and 7 (diseases and diagnoses). In Component 1, matches were most often rated as best fit all, accounting for 29.8% of all results in Component 1. Exact matches followed (23.7%), then synonyms (21.6%). In contrast, exact matches (37.5%) accounted for the largest proportion of results in Component 7, followed by synonyms (22.4%) and descriptions (17.7%). Differences in the map result categories between the two components were notable results were categorised as best fit all or no match in twice the proportion of Component 1 terms compared with Component 7. It is clear from these results that maps in Component 7 were more often valid than those in Component 1, indicating that SNOMED CT-AU represents diseases and diagnoses more accurately than symptoms and complaints, from the perspective of Australian general practice. Per cent Component 1 Component Exact match Description Synonym Best fit-all No match Map result category Note: Component 1 Symptoms and complaints; Component 7 Diagnoses and diseases. Figure 6.2: Distribution of mapping results according to ICPC-2 components and map result category 147

170 The process of mapping terms from ICPC-2 PLUS to SNOMED CT-AU resulted in a single link between each ICPC-2 PLUS term and a single SNOMED CT-AU concept, where such a link was possible to create. In this process, it was acceptable for multiple ICPC-2 PLUS terms to be mapped to a single SNOMED CT-AU concept, creating a many-to-one relationship. Overall, 4,981 ICPC-2 PLUS terms were mapped to 3,789 SNOMED CT-AU concepts, an average of 1.3 ICPC-2 PLUS terms mapped to each SNOMED CT-AU concept (results not tabled). Table 6.2 shows the pattern of SNOMED CT-AU concepts included in the mapping results, as defined by the SNOMED CT-AU semantic tag. A semantic tag is attached to the fully specified name of each concept, and represents the concept s position within the hierarchical structure of SNOMED CT-AU. 100 The majority of concepts included in the map were regarded as Disorders in SNOMED CT-AU (71.2%), with Findings incorporating an additional 27.1% of concepts. The other seven semantic tags arising in the map, together represented only 1.8% of concepts included as mapping targets. Table 6.2: Distribution of SNOMED CT-AU concepts included in the map according to the SNOMED CT-AU semantic tag SNOMED CT-AU semantic tag Number Per cent Disorder 2, Finding 1, Event Situation Procedure Life style Observable entity Physical object Regime/therapy Total 3, Comparison of acceptable and unacceptable matches in the mapping results An important consideration when developing a map is to ensure that the target identified is semantically equivalent to the source, that is, the source and the target have the same meaning. If a map target is not semantically equivalent to a map source, the result does not have semantic equivalence 148

171 and is unacceptable. Maps that are not semantically equivalent should not be implemented for clinical use. In the map from ICPC-2 PLUS terms to SNOMED CT-AU, semantic equivalence was defined as maps with a match result category of exact match, description, or synonym. In total, 69.6% of ICPC-2 PLUS terms were acceptably mapped to SNOMED CT-AU (Table 6.1). Figure 6.3 presents the semantically equivalent, or acceptable results for each component within each ICPC-2 chapter. Overall, the figure shows that in every ICPC-2 chapter the proportion of semantic matches was higher for ICPC-2 PLUS terms from Component 7 than from Component 1. In Component 7, the chapters with the highest proportion of acceptable matches were the Respiratory (86.7%), Ear (85.7%), Eye (85.5%), Circulatory (85.0%) and Urinary (84.7%) chapters. Levels of acceptable matches in Component 1 (symptoms and complaints) were more variable across chapters. The highest proportions of acceptable Component 1 matches were in the Neurological (77.5%) and Musculoskeletal (74.9%) chapters. However, in the Blood and blood forming organs, and Social chapters, less than a third of matches in Component 1 were regarded as acceptable (31.8% and 31.9% respectively). These results indicate that SNOMED CT-AU is consistently more accurate when representing diseases and diagnoses than symptoms and complaints, although the coverage of symptoms and complaints varies markedly according to ICPC-2 chapter. 149

172 ICPC-2 chapter A B Component 7 Component 1 D F H K L N P R S T U W X Y Z Per cent Note: Acceptable matches were those with a map result category of exact, description or synonym, as described in Section ICPC-2 chapters: A General and unspecified; B Blood and blood forming organs; D Digestive; F Eye; H Ear; K Circulatory; L Musculoskeletal; N Neurological; P Psychological; R Respiratory; S Skin; T Endocrine/metabolic/ nutritional; U Urinary; W Pregnancy and family planning; X Female genital; Y Male genital; Z Social. Component 1 Symptoms and complaints; Component 7 Diagnoses and diseases. Figure 6.3: Chapter component specific proportion of acceptable mapping results 150

173 6.4.3 Relationship between map result category and utilisation of terms To determine the relationship between the usage of ICPC-2 PLUS terms and the suitability of the map from those ICPC-2 PLUS terms to SNOMED CT-AU, term usage was divided into five groups, as shown in Figure 6.4. There were 82 terms that accounted for the top 50% of usage. The subsequent 25% of usage (from % of usage) required the use of 256 terms. The five groups shown in Figure 6.4 will be used subsequently in Table % 3,511 terms (usage between and 100.0%) 90% 526 terms (usage between 87.5 and 93.75%) 80% 418 terms (usage between 75.1 and 87.5%) 70% 60% 256 terms (usage between 50.1 and 75.0%) Per cent 50% 40% 30% 82 terms (top 50% of usage) 20% 10% 0% Figure 6.4: Relationship between number of terms used to describe RFEs and/or problems managed and term utilisation 151

174 Table 6.3 describes the relationship between the validity of the mapping results from ICPC-2 PLUS to SNOMED CT-AU and the usage of the source ICPC-2 PLUS terms within either the RFEs or problems managed data elements in BEACH. Of the 5,441 ICPC-2 PLUS terms mapped to SNOMED CT-AU and included in this analysis, 4,793 (88.1%) were used at least once in either or both data elements. Mapped ICPC-2 PLUS terms with a map result category of description were those most likely to have been used at least once in BEACH (91.4% of these terms being used at least once), followed by exact matches (90.4%) and synonyms (88.1%). ICPC-2 PLUS terms categorised as best fit precoordinated or no match were the least likely to have been used at least once (80.6% and 80.4% used at least once respectively). Table 6.3: Distribution of mapping results according to ICPC-2 PLUS terms used at least once as reasons for encounter and/or problems managed in BEACH (a) Mapping result category Number of ICPC-2 PLUS terms assigned to category (b) Number of ICPC-2 PLUS terms used at least once Per cent used at least once Number of PLUS terms not used Per cent not used Exact match 1,742 1, Description match Synonym match 1,201 1, Best fit match Best fit precoordinated match No match Total ICPC-2 PLUS terms 5,441 4, (a) (b) Term utilisation is derived from ICPC-2 PLUS terms from ICPC-2 components 1 and 7, used in either the reasons for encounter or problems managed data elements in BEACH between April 2004 and March Twelve mapping results were removed. These results related to ICPC-2 PLUS terms that were added between May 2010 and April 2011, and not included in the BEACH data collection period. The results that follow present the combined mapping and utilisation data in two forms: ICPC-2 PLUS terms used at least once as RFEs and/or problems managed; and problems managed alone. The results in Table 6.4 are divided into the three collapsed match result categories of acceptable match, unacceptable match and no match. The table shows that the proportions of match types both used and not used in the two groups were very similar. Acceptable matches were found for 71.8% of terms used in the problems 152

175 managed data element alone, compared with 71.0% used in either the RFE and/or problems managed data elements. Table 6.4: Comparison of mapping results and utilisation for RFEs and/or problems managed combined, and problems managed alone Problems managed alone RFEs and/or problems managed combined Match result type Used (n, %) Not used (n, %) Used (n, %) Not used (n, %) Acceptable match 3,339 (71.8) 445 (56.5) 3,404 (71.0) 380 (59.0) Unacceptable match 962 (20.7) 226 (28.7) 1,015 (21.2) 173 (26.9) No match 349 (7.5) 116 (14.7) 374 (7.8) 91 (14.1) Total 4,650 (100.0) 787 (100.0) 4,793 (100.0) 644 (100.0) In Table 6.5, the utilisation results have been further divided into usage groups, as described in Section The combined results for RFEs and/or problems managed show that ICPC-2 PLUS terms in the lowest frequency group (Group 5) were those with the highest proportion of no match results (22.6%), followed by the terms that were never used in BEACH (14.1%) (Table 6.5). Reasons for encounter and problems managed The proportion of acceptable matches was highest in Group 1 (91.5%), which accounted for the top 50% of usage, and remained high for terms included in Groups 2 (89.8%) and 3 (84.7%). This represents a high degree of concordance between terms used frequently and acceptable matches. The proportion of acceptable matches decreased in the lower usage groups, including 77.0% of Group 4 and 53.6% of Group 5. Unacceptable match results accounted for less than 10% of ICPC-2 PLUS terms in Groups 1 and 2 (8.5% and 9.0% respectively), which represented the terms used most frequently. Almost one-quarter (23.8%) of the unacceptable matches were in Group 5, and 26.9% of unacceptable matches were never used in BEACH as RFEs and/or problems managed. These results indicate that there is a high association between the utilisation of ICPC-2 PLUS terms and the extent to which such terms can be acceptably mapped to SNOMED CT-AU. 153

176 Table 6.5: Distribution of map result types according to ICPC-2 PLUS term utilisation in either the RFE and/or problems managed data elements Group 1 Group 2 Group 3 Group 4 Group 5 Not used Total Match result type Number Per cent Number Per cent Number Per cent Number Per cent Number Per cent Number Per cent Number Per cent REASONS FOR ENCOUNTER AND/OR PROBLEMS MANAGED COMBINED Acceptable match , , Unacceptable match , No match Total , , PROBLEMS MANAGED ONLY Acceptable match , , Unacceptable match , No match Total , , Group 1 the ICPC-2 PLUS terms that together accounted for the top 50% of terms utilised in the reasons for encounter or problems managed data elements in the BEACH study, Group 2 the ICPC-2 PLUS terms that together accounted for % of terms utilised in the reasons for encounter or problems managed data elements in the BEACH study, Group 3 the ICPC-2 PLUS terms that together accounted for % of terms utilised in the reasons for encounter or problems managed data elements in the BEACH study, Group 4 the ICPC-2 PLUS terms that together accounted for % of terms utilised in the reasons for encounter or problems managed data elements in the BEACH study, Group 5 the ICPC-2 PLUS terms that together accounted for % of terms utilised in the reasons for encounter or problems managed data elements in the BEACH study, Not used the ICPC-2 PLUS terms were not used in the reasons for encounter or problems managed data elements in the BEACH study,

177 Problems managed alone When the utilisation data were restricted to those terms used to describe problems managed only, the highest proportion of terms that were not mappable to SNOMED CT-AU (i.e. no matches ) were also not used in BEACH as problems managed (14.7%) (Table 6.5). In Group 1, 95.8% of matches were acceptable, and there were zero no matches. Similarly in Group 2, 90.2% of matches were acceptable, with most of the remainder being unacceptable matches (9.4%). The proportion of matches that were acceptable was lower in Group 3 (85.4%). Unacceptable matches were highest for ICPC-2 PLUS terms that were not used in BEACH (28.7%), and represented nearly one-quarter of matches in Group 5 (23.6%). The two sets of results were compared. In Group 1, there was a higher proportion of acceptable matches for problems managed alone (95.8%) than when RFEs and problems managed were combined (91.5%). In contrast, in Group 5 (which represented the lowest usage) there was a considerably higher proportion of no matches in the results in which RFEs and problems managed were combined (22.6%) than for problems managed only (8.9%). These results indicate that mapping results were more often acceptable matches for problems managed. Summary of quantitative findings Of the ICPC-2 PLUS terms included in the mapping to SNOMED CT-AU, 69.6% were acceptably mapped, although less than half (47.5%) of the ICPC-2 PLUS terms were explicitly included as concepts in SNOMED CT-AU (that is, they were an exact or description match). The highest proportions of exact matches were found in the Neurological, Circulatory and Respiratory chapters of ICPC-2. Exact and description matches were more often found for terms in ICPC-2 Component 7 than in Component 1. There was a high association between the extent to which terms were acceptably mapped to SNOMED CT-AU and their utilisation in the BEACH study. 155

178 6.5 Qualitative findings and recommendations This section describes issues identified in SNOMED CT-AU during the mapping process. The issues have been grouped into themes. Each theme is described according to the following structure. The overall issue is described, examples of the issue are provided, implications of the issue are identified and discussed, and suggested solutions for the issue are presented. I have included examples to illustrate each of the issues identified but these examples are in no way exhaustive. It would be impossible to provide a list of all instances of each issue. In some examples I have included the approximate number of concepts affected by each issue, where this could be easily calculated. In this section, labels used to describe SNOMED CT-AU descriptions (including fully specified names, preferred terms and synonyms) are represented using italics (e.g. Miscarriage). The number included in parentheses immediately following a SNOMED CT-AU label is the concept identifier Ambiguity in SNOMED CT-AU concepts Issue One of the principles included in Cimino s seminal work outlining desiderata for controlled medical vocabularies was that of concept orientation, whereby concepts included in a clinical terminology should not be ambiguous or redundant. 49 In other words, each concept must be mutually exclusive and have one, and only one, meaning. The work reported below suggests that the structure of SNOMED CT-AU puts it at risk of violating this principle. The issue of mutual exclusivity can be divided into two types, as described below. i. Identical clinical concepts with differing term structures located in the same hierarchy Examples and discussion I found some examples in SNOMED CT-AU where the same expression was included twice with a slightly different term structure. In the examples below, both concepts are found in the Clinical Findings foundation RefSet. 156

179 Example 1: Postpartum breast subinvolution There are two concepts in SNOMED CT-AU that describe postpartum breast subinvolution: Breast subinvolution, postpartum (disorder) ( ) Postpartum subinvolution of breast (finding) ( ) SNOMED CT concept Clinical finding Finding by site Finding of body region Disease Finding of trunk structure Disorder by body site Breast finding Finding of region of thorax Disorder of trunk Finding of lactation Breast finding Disorder of body system Disorder of thorax Subinvolution of breast Disorder of breast Postpartum subinvolution of breast Breast subinvolution, postpartum Box 6.2: Hierarchical representation of Postpartum subinvolution of breast (finding) and Breast subinvolution, postpartum (disorder) 157

180 These two concepts appear to be semantically identical. The one difference I can identify between them is lexical word order and term structures differ. However, the concepts have different semantic tags, with one being a finding, and the other a disorder. Although these concepts appear to represent an identical clinical expression, they are located in two parts of the Clinical Findings hierarchy, as shown in Box 6.2. A second example is found in the clinical expression Tic of organic origin. Example 2: Tic of organic origin There are two nearly identical concepts in SNOMED CT-AU to describe tics of organic origin: Tic of organic origin (disorder) ( ) Tics of organic origin (finding) ( ) The two concepts that refer to tic of organic origin also appear to be semantically identical. The differences between them are slight: one fully specified name is expressed in singular form, and the other in plural form. The concepts also have different semantic tags, with one tagged as a disorder and the other as a finding. Example 3: Food intolerance This issue is represented in SNOMED CT-AU as the following two concepts: Food intolerance (finding) ( ) Food intolerance (disorder) ( ) In Example 3, the two concepts are identical, except for the semantic tag. The SNOMED CT editorial documentation states that the inclusion of clinical conditions as both findings and disorders is valid. This is due to the difference in the definitions of finding and disorder. In SNOMED CT (and therefore in SNOMED CT-AU), disorders are always and necessarily abnormal clinical states, 102 however findings may be normal and may exist only at a single point in time

181 Although this distinction seems clear according to the definitions above, it may not be clear to an end user using the terminology in clinical practice. According to the SNOMED CT Technical Implementation Guide, the preferred term is the term that is deemed to be the most clinically appropriate way of expressing a concept in a clinical record. 100 In contrast the fully specified name is expressed in a way that should remove or reduce any possible ambiguity. 100 It can therefore be assumed that the preferred term is the description that should be presented to the user, rather than the fully specified name. However, preferred terms are not unique, and in the two examples provided above clinical users would be presented with seemingly identical descriptions if they were presented as preferred terms only. The ambiguity created by the preferred term in this example therefore violates the principle of mutual exclusivity. Suggested solution In the examples provided above, one of the concepts could be retired as an active concept to remove ambiguity. A link should be created from the retired concept to the alternate concept which retained its current status. If terminology developers decide that both concepts are valid in SNOMED CT-AU, the differences between findings and disorders must be clearly explained to users. SNOMED CT-AU developers should re-assess their statement that the preferred term is the term that is deemed to be the most clinically appropriate way of expressing a concept in a clinical record 100 in light of potential ambiguity in some preferred terms. ii. Identical concept labels in different foundation RefSets Example and discussion There are a number of examples in SNOMED CT-AU where descriptions that are lexically and semantically identical are included in different foundation RefSets, distinguishable only through the semantic tag and a different concept or description identifier. 159

182 Example 1: Chondromatosis This issue is represented in SNOMED CT-AU as the following two concepts: Chondromatosis (morphologic abnormality) ( ) with preferred term Chondromatosis Chondromatosis (disorder) ( ) with preferred term Chondromatosis This is an example where attaching different semantic tags to apparently identical concepts may have a purpose. There are many other examples where disorders are included as either findings or disorders, and morphologic abnormalities. In SNOMED CT and SNOMED CT-AU, morphologic abnormalities are defined as morphologic alterations from normal body structures. 100 SNOMED CT documentation also states that codes from the morphological abnormality hierarchy should not be used in place of codes from the clinical findings hierarchy, even though they appear to refer to similar clinical situations and that findings should be used to represent the combination of a morphology in a location. 102 As such, it is probably reasonable that the kernel concept is included twice in these examples. However, both concepts have the preferred term Chondromatosis, which may lead to confusion for the end user. Suggested solution Semantic tags are intended to reduce the ambiguity within descriptions, 100 however this only applies if the user is presented with the fully specified name rather than the preferred term. Possible solutions for this issue therefore include the display of the fully specified name in conjunction with the preferred term, or alternatively, limiting the hierarchies available to the user for particular clinical contexts. For example, concepts in the morphologic abnormality hierarchy could be excluded in a problem/diagnosis field in an EHR. Guidance for EHR software vendors implementing SNOMED CT-AU will be required to ensure they understand the differences between hierarchies in 160

183 SNOMED CT-AU and the consequences of implementing the preferred term as the concept description viewed by the user Inconsistent/incorrect inclusion of synonyms Issue As described in Section there are three types of descriptions in SNOMED CT: fully specified names, preferred terms and synonyms. In the SNOMED CT technical documentation, synonyms are defined as: A term that is an acceptable alternative to the preferred term as a way of expressing a concept. Synonyms allow representations of the various ways a concept may be described. 99 In the mapping of ICPC-2 PLUS terms to SNOMED CT-AU, I discovered some concepts where synonym descriptions were included incorrectly or inconsistently when compared with the concept s fully specified name. These fell into the two categories described below. i. Descriptions that are not synonyms Examples and discussion I identified some synonym descriptions linked to SNOMED CT-AU concepts that are not regarded as synonymous with the listed preferred term in the Australian context. Example 1: Malnutrition The SNOMED CT-AU concept Nutritional disorder (disorder) ( ) contains the synonym Malnutrition. Dorland s Medical Dictionary (published in the United States) defines malnutrition as any disorder of nutrition; it may be due to unbalanced or insufficient diet or to defective assimilation of utilisation of foods. 191 This is not the definition used in Australia. The Australian Macquarie Dictionary states that malnutrition is imperfect nutrition; lack of proper nutrition resulting from deficiencies in the diet or the process of assimilation. 192 Dorland s definition any disorder of nutrition is consistent with the inclusion of malnutrition in the SNOMED CT-AU concept Nutritional disorder. However, in Australia, malnutrition is regarded as a type of nutritional disorder due to 161

184 deficiencies in nutrition, but is not a term that is used to group all nutritional disorders. I found other examples where the synonyms in the descriptions list of a SNOMED CT-AU concept were not synonymous with the concept s fully specified name. Example 2: Infertility The SNOMED CT-AU concept Infertile (finding) ( ) contains the synonym Difficulty conceiving. The definition of infertility in Dorland s Medical Dictionary is diminished or absent capacity to produce offspring; the term does not denote complete inability to produce offspring as does sterility. 191 The inclusion of Difficulty conceiving as a synonym of infertility in SNOMED CT-AU is therefore consistent with the definition in Dorland s. However, this synonym may not represent the common usage interpretation of the term. Although Dorland s states that the term infertility does not mean a patient cannot ever conceive (i.e. sterility), in common parlance infertility and sterility are usually regarded as synonyms. In ICPC-2, the definition given for the ICPC-2 codes W15 Infertility/subfertility female and Y10 Infertility/subfertility male is failure to conceive after two years of trying. 40 This definition provides a clear time period for the definition of a conception problem as infertility. A clinician may enter the label Difficulty conceiving into the patient s medical record to indicate their current understanding of the patient s problem, prior to the two year criterion provided in ICPC-2. The label Difficulty conceiving should therefore be the term presented on the health summary or problem list. However, if a clinician is using SNOMED CT-AU the preferred term or the fully specified name of Infertile may also be automatically included on the summary/problem list, together with the synonym chosen by the clinician. The inclusion of the description Infertile, even in conjunction with the clinician s entry of Difficulty conceiving, may not be representative of the clinical concept the clinician 162

185 wishes to record. In addition, such an entry could be misinterpreted by the patient in a way that is highly emotive, and may not be consistent with the message the clinician wishes to convey to the patient. Suggested solution In SNOMED CT-AU, there is a separate concept for malnutrition called Malnutrition (calorie) (disorder) ( ). In the mapping from ICPC-2 PLUS to SNOMED CT-AU, I mapped the ICPC-2 PLUS term for malnutrition to this concept, rather than to Nutritional disorder. However, the use of the concept Nutritional disorder (disorder) ( ) should be viewed with caution in Australia, as it may be incorrectly used to represent malnutrition. Although SNOMED CT is an international clinical terminology, the ability to create national extensions of SNOMED CT for national use cases allows additional flexibility in the use of SNOMED CT for an individual country s purposes. SNOMED CT-AU is one example of a national version of the international SNOMED CT clinical terminology. As such, within SNOMED CT-AU, the synonym Malnutrition should be removed from the concept Nutritional disorder (disorder) ( ). Example 2 represents an issue that is not an easy one to solve. According to the definition given in Dorland s Medical Dictionary, the inclusion of this synonym is accurate. However, the potential for misinterpretation of the synonym is very high. I have not been able to find any indication as to whether SNOMED CT intends to capture terms from the patient s perspective, nor the validity of SNOMED CT for this purpose. I would therefore recommend that Difficulty conceiving ought to be a separate concept from Infertility to incorporate its broader meaning, particularly in general practice. ii. Synonyms that represent characteristic symptoms or signs of a disorder Example 1 and discussion Some synonym descriptions in SNOMED CT-AU represent symptoms of the disorder that is described by the fully specified name and preferred term. 163

186 However, the term used to describe a symptom or sign, and the term used to describe the resulting disorder, are not always synonymous. Example 1: Elevated blood pressure and hypertension The SNOMED CT-AU concept Hypertensive disorder, systemic arterial (disorder) ( ) has the preferred term Hypertensive disorder. This concept also has descriptions including High blood pressure, Elevated blood pressure and Raised blood pressure. In ICPC-2, there are three separate rubrics related to blood pressure problems: K85 Elevated blood pressure, K86 Hypertension, uncomplicated and K87 Hypertension, complicated. The distinction between the rubrics is that K86 and K87 are defined as a finding of high blood pressure on either two or more readings per encounter, taken at two or more encounters, while the inclusion for K85 is elevated blood pressure not meeting criteria for K86 and K These definitions imply that the diagnosis of hypertension requires sustained elevated blood pressure over multiple encounters, rather than a single or isolated instance of elevated blood pressure. The definitions in ICPC-2 are supported by hypertension guidelines developed by the Australian National Heart Foundation, who report that: The diagnosis of hypertension should be based on multiple BP measurements taken on several separate occasions, e.g. at least twice, one or more weeks apart. 193 Although the correct labelling of an incidental or single finding of elevated blood pressure is not discussed in this guideline, it can be implied that a single occasion of elevated blood pressure is not synonymous with the term hypertensive disorder or hypertension. As such, the inclusion of the descriptions High blood pressure, Elevated blood pressure and Raised blood pressure in the SNOMED CT-AU concept for hypertension is incorrect. Suggested solution SNOMED CT-AU does contain a concept called Finding of increased blood pressure (finding) ( ), which already contains the descriptions Elevated blood pressure and Raised blood pressure. An additional description of high blood pressure should be added to this concept. In addition, the 164

187 following descriptions should be removed from the concept Hypertensive disorder, systemic arterial (disorder) ( ): High blood pressure Elevated blood pressure Raised blood pressure. Example 2 and discussion A similar example is found in the concept Gastroesophageal reflux disease. Example 2: Acid reflux and gastro-oesophageal reflux disease Acid reflux is included as a synonym to SNOMED CT-AU concept Gastroesophageal reflux disease (disorder) ( ). The Gastroenterological Society of Australia guidelines for the diagnosis and management of Gastroesophageal Reflux Disease (GORD) state that reflux of gastric contents into the oesophagus is a normal physiological event. 194 A diagnosis of GORD is made when reflux causes complications which impair a patient s wellbeing, and usually occurs when symptoms are present on two or more days a week. 194 As such, a diagnosis is made on the recurrence of reflux symptoms. Suggested solution SNOMED CT does contain alternative concepts for reflux, called Oesophageal reflux finding (finding) ( ) and Reflux (finding) ( ). Clinicians using SNOMED CT-AU should use these concepts to record the symptom of reflux. In addition, the following synonyms should be removed as descriptions from the concept Gastroesophageal reflux disease (disorder) ( ): Gastroesophageal reflux Oesophageal reflux Acid reflux. 165

188 6.5.3 Inconsistent concept labels Issue I observed a lack of standardisation in the labels used to describe related concepts in SNOMED CT-AU. Examples and discussion When navigating the SNOMED CT-AU hierarchies, I identified sibling concepts (i.e. located in the same level of the hierarchy) that were structured or labelled differently. Example 1: Abscesses The SNOMED CT-AU concept Abscess (disorder) ( ) is a parent concept for abscesses of different sites. However, the concepts sitting below Abscess (disorder) have varying term structures. This example is explained further in Box 6.3, which contains the hierarchical representation of these concepts. Abscess of trachea Abscess Muscle abscess Box 6.3: Representation of Abscess and its descendents Examples 2 and 3 provide instances of SNOMED CT-AU hierarchies where concepts that are hierarchically related in a parent child relationship are labelled inconsistently. Example 2: Mountain sickness Mountain sickness is a description for the concept Anoxia due to high altitude (disorder) ( ) for which the preferred term is Anoxia due to high altitude. Its child concepts are called: Acute mountain sickness (disorder) ( ) Subacute mountain sickness (disorder) ( ). 166

189 Example 3: Irritable colon The concept Irritable colon (disorder) ( ) has the preferred term Irritable colon. It also includes, as one of many, the synonym Irritable bowel syndrome. However, the children of this concept are labelled with the kernel concept Irritable bowel syndrome. Example 3 is detailed further in Box 6.4. Irritable bowel syndrome characterised by constipation Irritable colon Irritable bowel syndrome characterised by alternating bowel habit Irritable bowel syndrome variant of childhood Irritable bowel syndrome with diarrhoea Box 6.4: Representation of Irritable colon and its descendents It should also be noted that none of the child concepts representing types of irritable colon/bowel syndrome, which are represented by the kernel label irritable bowel syndrome have an associated description that includes a kernel label of irritable colon. The primary implication for this issue relates to the ease with which users can access these concepts through the user interface where descriptions (including preferred terms and synonyms) are not structured consistently. Suggested solution Labels of similar preferred terms need to be made consistent, whether between different levels of the SNOMED CT hierarchy, or within the same level of the hierarchy. This issue, like many others, has been inherited from the SNOMED CT international release, and needs to be corrected at the international level. It is evident in SNOMED CT that there have been no strict editorial rules applied in the past when structuring the labels used to describe concepts and descriptions in SNOMED CT. The IHTSDO publishes editorial guidelines. 102 Guidance is provided about the issue outlined in Example 1, and 167

190 states that the morphologic abnormality should be named before the site (that is, abscess of muscle rather than muscle abscess ). It is reasonable to expect that the implementation of these guidelines will rectify this issue over time. However, no specific guidance is provided in relation to Examples 2 and Precoordination in SNOMED CT-AU Issue Precoordination is the combination of two or more clinical statements into a single clinical concept. As previously described in Section 6.3.3, there are two types of precoordination: precoordination by combination, where two or more kernel terms are combined to create a single concept precoordination by qualification, where a qualifier (e.g. severity) is combined with a kernel concept. I identified some concepts in SNOMED CT-AU that contained multiple clinical statements in the same concept, and thus could be regarded as excessively precoordinated. i. Precoordination by combination Example and discussion Some precoordinated concepts in SNOMED CT-AU contain more than one clinical condition. Example 1: Lumbar and/or sacral arthritis SNOMED CT-AU contains a concept called Lumbar AND/OR sacral arthritis (disorder) ( ). 168

191 This precoordinated SNOMED CT-AU concept contains three individual clinical conditions: lumbar arthritis sacral arthritis lumbosacral arthritis. As such, this concept is ambiguous in its current form, and is not sufficiently explicit to be clinically meaningful for the user. Sacral arthritis is included as its own concept in SNOMED CT-AU Sacral arthritis (disorder) ( ). The closest match for lumbar arthritis is Arthropathy of lumbar facet joint (disorder) ( ). Lumbosacral arthritis does not appear to be represented separately in SNOMED CT-AU. I examined the use of the expression and/or in SNOMED CT concepts, and discovered that there were 1,085 current concepts in the Clinical Findings foundation RefSet that contained the and/or expression. Some of these included the expression multiple times, for example Open wound of knee and/or leg and/or ankle (disorder) ( ). The inclusion of the expression and/or indicates that there are three clinical expressions included in each clinical concept that contains the expression. One could validly argue separating these expressions into their three constituent parts has the potential to increase the size of SNOMED CT-AU exponentially. However, the precoordinated concept as it is currently expressed is clinically ambiguous. A clinician managing or referring a patient must specify whether the patient has lumbar arthritis, sacral arthritis or lumbosacral arthritis, otherwise there will be insufficient detail to guide management of the condition. Proposed solution The concept Lumbar AND/OR sacral arthritis (disorder) ( ) should be retired from SNOMED CT-AU. A new concept should be created in SNOMED CT-AU for Lumbosacral arthritis. In addition, any concepts that contain the phraseology and/or should be reviewed for potential retirement. 169

192 SNOMED CT supports the creation of postcoordinated expressions as an alternative to precoordinated concepts. The IHTSDO unsuccessfully attempted to define excessive precoordination, and now relies on the usefulness of a clinical concept in the health care domain to determine its appropriateness in SNOMED CT. 102 The examples demonstrated above suggest that this criterion is not adequate, and stricter processes need to be implemented to remove existing excessive precoordination. ii. Precoordination by qualification Examples and discussion There are also examples of precoordination by qualification found in SNOMED CT-AU. Each of these examples represents a precoordinated concept containing contextual information in addition to a kernel concept. Example 1: Complaining of There is a concept in SNOMED CT-AU called Complaining of cough (finding) ( ). This concept is a separate entity to the concept Cough (finding) ( ). The concept Complaining of cough contains two parts: the kernel concept of cough and the contextual information complaining of. The contextual information indicates that the problem is a symptom described by the patient. However, this information could be represented in other ways, using either postcoordination, or using the information model. If using an information model, this concept could be represented by including the concept Cough in a reason for encounter or reason for presentation field. If postcooordination was implemented in the EHR, a postcoordinated expression containing the concept Cough (finding) ( ) and a concept representing complaining of could be created. However, a concept called complaining of on its own is not currently included in SNOMED CT-AU. There are 65 concepts in SNOMED CT-AU that contain the expression complaining of, so this is a common problem. A different type of precoordination by qualification is shown in Example

193 Example 2: Family history of There is a concept in SNOMED CT-AU called Family history: Depression (situation) ( ). The contextual information in Example 2 (that is, family history) is important because it indicates that the patient does not have the condition that represents the kernel concept. Instead, they have a family history of the condition, which may put them at risk of developing the condition. Example 3: Finding of There is a concept in SNOMED CT-AU called Finding of size of ear canal (finding) ( ). Example 3 illustrates a trend in SNOMED CT-AU to include placemarker concepts, that is, concepts that mark a place in the SNOMED CT hierarchies for the child concepts that sit below. Box 6.5 shows that the concept Finding of size of ear canal is the parent of three concepts, each of which represents the clinical finding relating to the size of the ear canal (small, normal or large). This demonstrates that this concept containing Finding of is not a clinical condition in itself, and therefore should not be available for use in clinical settings. Some concepts that begin with the expression Finding of do contain values that have clinical meaning, for example, Finding of increased blood pressure (finding) ( ). However, the inclusion of finding of is redundant in this situation and this concept could alternately be labelled Increased blood pressure. The aim of a concept in a clinical terminology is to capture clinical meaning in a manner that is not vague or ambiguous. 171

194 Large ear canal Finding of size of ear canal Normal sized ear canal Small ear canal Box 6.5: Representation of hierarchy below Finding of size of ear canal SNOMED CT-AU contains a hierarchy/foundation RefSet to incorporate concepts that contain context the Situation with Explicit Context hierarchy. All concepts that relate to history (including, but not limited to family history) belong in this hierarchy. As such, the SNOMED CT developers acknowledge that some concepts must contain contextual information and have managed the issues that result from this by grouping them together in the Situation with Explicit Context hierarchy. Suggested solutions Examples 1 and 2 can alternately be captured using other means, including postcoordination or through use of the information model. As such, the inclusion of these concepts in SNOMED CT-AU is unnecessary and they should be removed. Prior to removal, SNOMED CT-AU editors must ensure that the kernel concept (containing the core component of the concept, such as cough in Example 1) is included in SNOMED CT-AU as a separate concept, so that this clinical meaning can be recorded. Many of the concepts containing finding of do not contain clinical meaning, and should not be included in a clinical record without further specification Content missing in SNOMED CT-AU Issue While preparing the map from ICPC-2 PLUS to SNOMED CT-AU, I identified that there was no equivalent match in SNOMED CT-AU for 8.5% of ICPC-2 PLUS terms (Section 6.4.1). This section describes some areas of missing content in SNOMED CT-AU identified during the mapping process. Three types of concepts were missing in SNOMED CT-AU: content areas that were 172

195 incomplete; whole content areas absent from the terminology; and concepts that had been included in inappropriate hierarchies. i. Incomplete content areas Examples and discussion Example 1: Lesion of cervical spine There are concepts in SNOMED CT-AU for lesions of some parts of the spine, as follow: Lesion of thoracic spine (finding) ( ) Lesion of lumbar spine (finding) ( ) However, concepts do not exist for lesions of the cervical, sacral or lumbosacral spine. As SNOMED CT-AU contains some concepts that include the anatomical level of the spine affected (which could be regarded as precoordinated content), there is a case to include the additional concepts that represent other anatomical levels of the spine. This ensures that the terminology is complete in the sense that all options for any given concept area are included. Advocates of postcoordination might argue that the inclusion of site is not necessary, and that the concept should read Lesion of spine, with subsequent postcoordination of anatomical site if necessary. However, if the SNOMED CT editors feel that some concepts should include anatomical site, all related concepts should be included to ensure this content area is complete. In addition, it should be noted that there is currently no parent concept called Lesion of spine, so there is no concept that can be used when the desired site is not available in SNOMED CT-AU. Suggested solution Each of the terms not available in SNOMED CT-AU will be sent to NEHTA as request submissions for future inclusion in SNOMED CT-AU. 173

196 ii. Concepts absent in SNOMED CT-AU Examples and discussion In Figure 6.1, the highest number of no matches were found in the social chapter of ICPC-2. Some of the terms from the social chapter of ICPC-2 that were not found in SNOMED CT-AU included: Fear of job loss Harassment. SNOMED CT-AU contains concepts for Sexual harassment (finding) ( ), Victim of sexual harassment (finding) ( ) and Harassment by landlord (finding) ( ). However, other types of harassment are not available, including harassment as a concept on its own, other types of harassment, or other perpetrators. Discrimination that is related to age, sex or disability. SNOMED CT-AU contains concepts for Religious discrimination (finding) ( ) and Political discrimination (finding) ( ), but not for other forms of discrimination, or general discrimination. Social problems account for 0.5% of all problems managed in Australian general practice. 9 This is not a high proportion, but does demonstrate that social problems are managed by Australian GPs. Suggested solution Each of the clinical terms not found in SNOMED CT-AU will be forwarded to NEHTA through the request submission process for inclusion in future SNOMED CT-AU releases. iii. Concepts included in inappropriate hierarchies Examples and discussion When searching for concepts in SNOMED CT-AU for the mapping from ICPC-2 PLUS, the matches found for some concepts were only in inappropriate hierarchies. 174

197 Example 1: Infectious jaundice The only match found in SNOMED CT-AU for the ICPC-2 PLUS term Infectious jaundice was to the concept Canine leptospirosis (disorder) ( ) which included a synonym of Infectious jaundice. The concept Canine leptospirosis is a subtype of Non-human disorder (disorder) ( ) and is a veterinary disease affecting dogs. As this condition can be present in humans as well as animals, it should be added to SNOMED CT-AU as a concept that affects humans. I also identified some clinical conditions in SNOMED CT-AU that were available in the Morphologic abnormality foundation RefSet only, and not in the Clinical findings foundation RefSet. Example 2: Haemorrhages The concept Haemorrhage by itself was only included in SNOMED CT-AU as Hemorrhage (morphologic abnormality) ( ). I have earlier referred to the SNOMED CT documentation stating that the combination of a morphologic abnormality and a site is represented in the clinical findings hierarchy, and that morphologic abnormality concepts on their own should not be used as clinical findings. 102 This becomes an issue when clinical expressions such as haemorrhage (with no further specification of site) are not available in the Clinical findings hierarchy. Suggested solutions A new concept needs to be added to SNOMED CT-AU for Infectious jaundice that is not linked to canine leptospirosis. It is difficult to suggest a solution for the issue presented in Example 2. A pragmatic solution is to allow users to enter Haemorrhage from the morphologic abnormality hierarchy, however the IHTSDO has stated that morphologic abnormalities are not acceptable as clinical findings. Precoordinated concepts containing haemorrhage in combination with the site of the haemorrhage are included in the Clinical findings hierarchy, so forcing users to specify the site of the haemorrhage is another practical solution. 175

198 iv. Synonyms not included as descriptions Examples and discussion I identified some gaps in SNOMED CT-AU, where commonly accepted synonyms were not included as descriptions within SNOMED CT-AU concepts. Example 1: Perspiration Perspiration is only included as a synonym for the concept Sweating, function (observable entity) ( ). Perspiration (or variants such as perspiring) is not included as a synonym for other concepts related to sweating, such as Sweating (finding) ( ) and most, if not all, of its children such as Sweating problem (finding) ( ) and Excessive sweating (finding) ( ). Example 2: Pain in limbs Pain in upper limb (finding) ( ) does not include an associated synonym for Pain in arm. Likewise, Pain in lower limb (finding) ( ) does not have an associated synonym for Pain in leg. The implications of these issues largely relate to the interface used to access SNOMED CT-AU, as they minimise the chance of a clinical user finding an appropriate concept unless an alternative thesaurus or logical searching mechanism is implemented. Suggested solutions The SNOMED CT technical documentation states that an additional file containing synonyms can be used to enhance the search mechanisms, 100 and the presence of perspiration (and its variants) in the synonym file may overcome this issue. Alternatively, perspiration (or variants such as perspiring) should be added as a synonym for all concepts related to sweating, including the concept Sweating (finding) ( ) and most, if not all, of its subtypes such as Sweating problem (finding) ( ) and Excessive sweating (finding) ( ). Example 2 could be rectified very easily with the addition of the descriptions listed in the example. 176

199 v. Missing Australian content Examples and discussion During the mapping I identified some Australian terms currently available in ICPC-2 PLUS that could be added as synonyms to SNOMED CT-AU to assist users with the identification of concepts. Example 1: Vomiting SNOMED CT-AU does not contain a synonym linked to the concept Vomiting (disorder) ( ) for the Australian colloquialism Throwing up. Example 2: Diarrhoea In Australia, Diarrhoea (finding) ( ) is colloquially known as having the runs. This synonym is not present in SNOMED CT-AU. Overall I found the inclusion in SNOMED CT-AU of conditions found primarily in Australia to be very good, with clinical conditions such as Barmah Forest Virus disease and Murray Valley encephalitis available as concepts. The use of Australian colloquialisms by patients is well researched in clinical encounters, particularly among those with international medical graduates. 195,196 ICPC-2 PLUS includes some colloquial terms to assist GP users to record RFEs in particular, which are meant to reflect the patient s language. These terms were requested by end users to aid their use of the terminology. However, the inclusion of such colloquialisms is limited in SNOMED CT-AU. Suggested solution The internationalisation of SNOMED CT and the creation of the Australian version of SNOMED CT (SNOMED CT-AU) mean that medical terms with a basis in the Australian idiom can be included in SNOMED CT-AU. Therefore terms such as those identified above could be added as synonyms to their respective concepts. 177

200 6.5.6 Issues when choosing a map of best fit Issue During this project I attempted to find a match of best fit in SNOMED CT-AU for the source ICPC-2 PLUS terms. In other words, I was trying to locate a single concept in SNOMED CT-AU that best represented the term included in ICPC-2 PLUS. Occasionally I found situations in which there were multiple SNOMED CT-AU concepts which could equally be regarded as the target concept in the map. Example and discussion Example 1: Flatulence I located three potential target concepts for the ICPC-2 PLUS term D08008 Flatulence. These target concepts were: Flatulence/wind (finding) ( ) Flatulence symptom (finding) ( ) Excessive upper gastrointestinal gas (finding) ( ) The concepts in Example 1 are all located in the Clinical findings foundation RefSet, and have the finding semantic tag, so the allocation of a map of best fit cannot use the semantic tag to assist the decision making process. The hierarchical representation of these concepts is shown in Box 6.6, with the possible target concepts highlighted. The box shows that there is no direct parent-child relationship between any of the three concepts to assist with the assignation of a map of best fit. I selected the concept Excessive upper gastrointestinal gas (finding) ( ) for the map from ICPC-2 PLUS to SNOMED CT-AU, however any of the three listed concepts could have been chosen as the best fit target concept. 178

201 SNOMED CT concept Clinical finding Finding by site Digestive system finding Finding reported by subject or history provider Gastrointestinal tract finding Digestive symptom Finding of gastrointestinal tract gas Flatulence symptom Flatulence/wind Finding of upper gastrointestinal gas Excessive upper gastrointestinal gas D: Flatulence Box 6.6: Hierarchical representation of concepts related to flatulence Suggested solution These concepts are too similar to be unambiguous, and essentially they all relate to the same clinical concept. The concept Flatulence/wind (finding) ( ) should be removed, with all related descriptions linked to Excessive upper gastrointestinal gas (finding) ( ). The concept Flatulence symptom (finding) ( ) could remain, although this would raise a separate issue about the use of context in SNOMED CT-AU concept labels, already discussed in Section (ii. Precoordination by qualification) Issues with the terminology used in SNOMED CT-AU descriptions Issue Occasionally I was unable to find matches in SNOMED CT-AU for some terms due to the words used within a label differing from the words used in ICPC-2 PLUS. 179

202 Examples and discussion Example 1: Neoplasm of uncertain behaviour The equivalent phrase in SNOMED CT-AU for any ICPC-2 PLUS terms that include the label Neoplasm of uncertain nature, is Neoplasm of uncertain behaviour. Similarly, ICPC-2 PLUS uses the phrase malignant neoplasm whereas SNOMED CT-AU uses the term malignant tumour. Example 2: Tumour versus neoplasm SNOMED CT-AU contains a concept Malignant tumor of colon (disorder) ( ). This concept contains a synonym Cancer of colon but does not contain the synonym malignant neoplasm of colon. Suggested solution The SNOMED CT editorial guidance states that the term neoplasm should be used in preference to the term tumour because a tumour may or may not be neoplastic. 102 This indicates that concepts with preferred terms including the expression malignant tumour should be replaced by preferred terms that include malignant neoplasm. In addition, some of the concepts labelled as malignant tumour have a preferred term that contains the US spelling of tumour (i.e. tumor) rather than the Australian spelling (as shown in the example above). Although my recommendation is to replace malignant tumor/tumour with malignant neoplasm, preferred terms in SNOMED CT-AU should always contain Australian spelling SNOMED CT hierarchies Issue SNOMED CT-AU has a multi-hierarchical structure. Hierarchies are designed to contain many levels of granularity, including both general concepts (low granularity), and very specific concepts (high granularity). Hierarchies are structured in a step-wise manner, where each concept in SNOMED CT-AU is linked to one or more parent concepts using is-a relationships, as described 180

203 in Section 3.3. During the mapping from ICPC-2 PLUS terms to SNOMED CT-AU, I identified some examples where the is-a relationships were incorrectly modelled. Examples and discussion Example 1: Tenosynovitis The hierarchies around tenosynovitis of the hand and its constituent parts (the fingers and thumbs) have not been modelled in a standardised manner. Box 6.7 explains the problem further. Tenosynovitis of fingers (disorder) ( ) is a child of Tenosynovitis of hand (disorder) ( ), which is a child of Tenosynovitis (disorder) ( ). However, types of tenosynovitis of the finger or thumb, including Extensor tenosynovitis of finger (disorder) ( ), Extensor tenosynovitis of thumb (disorder) ( ), Flexor tenosynovitis of finger (disorder) ( ) and Flexor tenosynovitis of thumb (disorder) ( ) are direct children of Tenosynovitis, and are not hierarchically linked to Tenosynovitis of finger. Extensor tenosynovitis of finger Extensor tenosynovitis of thumb Flexor tenosynovitis of finger Tenosynovitis Flexor tenosynovitis of thumb Tenosynovitis of hand Tenosynovitis of fingers Box 6.7: Representation of Tenosynovitis and its descendents 181

204 Example 1 demonstrates a situation where the is-a relationships in SNOMED CT-AU have not been modelled correctly. When considering anatomical specificity, fingers and thumbs are parts of the hand. Therefore it would be expected that all clinical conditions that relate to fingers or thumbs would be hierarchically placed below the concept for the hand, and specific types of conditions of the finger or thumb should be hierarchically placed below the general concept for the finger. Example 2 demonstrates an issue relating to a hierarchical linkage that is missing. The current hierarchy in this example is shown in Box 6.8. Example 2: Thyroid disease in pregnancy The SNOMED CT-AU concept Thyroid disease in pregnancy (disorder) ( ) does not contain hierarchical linkages to disorders of pregnancy. As shown in Box 6.8, the concept Thyroid disease in pregnancy (disorder) ( ) is hierarchically linked to the concept Disorder of thyroid gland (disorder) ( ) but does not have any hierarchical linkages to disorders during pregnancy. One of the assumed advantages of a multi-hierarchical terminology is that the hierarchies may be used for navigation; to identify related concepts of greater or lesser specificity. However, this can only occur if the hierarchical linkages are complete. The examples provided above demonstrate that additional work needs to be undertaken to improve the quality of the SNOMED CT-AU hierarchies for this purpose. 182

205 SNOMED CT concept Clinical finding Finding by site Disease Finding of body region Disorder by body site Finding of head and neck region Disorder of body system Finding of neck region Disorder of endocrine system Disorder of neck Disorder of thyroid gland Thyroid disease in pregnancy Thyrotoxicosis in pregnancy Box 6.8: Hierarchical representation of Thyroid disease in pregnancy Suggested solutions From Example 1, the concepts Extensor tenosynovitis of finger (disorder) ( ), Extensor tenosynovitis of thumb (disorder) ( ), Flexor tenosynovitis of finger (disorder) ( ) and Flexor tenosynovitis of thumb (disorder) ( ) should be moved to sit below the concept Tenosynovitis of fingers (disorder) ( ). 183

206 From Example 2, an additional hierarchical linkage should be made, so that Thyroid disease in pregnancy (disorder) ( ) is linked through an isa relationship to the concept Disorder of pregnancy (disorder) ( ) Ambiguity in SNOMED CT-AU Issue I found some examples where the label of the SNOMED CT-AU concept, as represented by the fully specified name and/or preferred term, was ambiguous. Examples and discussion Example 1: Loss of feeling When mapping the ICPC-2 PLUS term Loss of feeling, a neurological problem, the closest match in SNOMED CT-AU was Feeling of loss of feeling (finding) ( ). The concept Feeling of loss of feeling (finding) is a clinical finding, with parents as shown in Box 6.9. Feeling of loss of feeling (finding) is a psychological concept. In contrast, the ICPC-2 PLUS term is neurological in nature. The Macquarie Dictionary provides 31 definitions of the word feel, including to perceive or examine by touch and to have mental sensations or emotions. 192 The word feel therefore has both neurological and psychological connotations. As such, the phrase loss of feeling can be interpreted in two different ways. In both ICPC-2 PLUS and SNOMED CT-AU, the term loss of feeling can be correctly interpreted only in conjunction with its structural components, for example the ICPC-2 rubric or the SNOMED CT-AU hierarchy. 184

207 SNOMED CT concept Clinical finding Clinical history and observations finding Functional finding Psychological finding Mental state, behaviour, and/or psychosocial function finding Mental state finding Emotional state finding Mood finding Feeling of loss of feeling Box 6.9: Hierarchical representation of Feeling of loss of feeling (finding) Example 2: Drunk The description Drunk is a synonym for two SNOMED CT-AU concepts: Alcohol intoxication (disorder) ( ) and Heavy drinker (finding) ( ). Example 2 demonstrates an issue with ambiguity in the SNOMED CT-AU descriptions. In the concept Alcohol intoxication (disorder) ( ) the synonym Drunk means that the patient is acutely intoxicated by alcohol, more colloquially known as being drunk. In contrast, the synonym Drunk linked to the concept Heavy drinker (finding) ( ) means that the patient drinks excessive amounts of alcohol, but may not currently be acutely intoxicated by alcohol. In colloquial terms, this patient is a drunk. 185

208 The difference between these two interpretations is subtle but important. In both situations the inclusion of the synonym Drunk is correct. However, without the accompanying fully specified name or preferred term, this term is ambiguous. Suggested solutions In both examples provided, the ambiguity described needs to be corrected if the expressions are to be used in isolation, without accompanying information such as the hierarchical placement of the concept in Example 1 or the fully specified name or preferred term in Example

209 6.6 Discussion The purpose of mapping the ICPC-2 PLUS terms to SNOMED CT-AU was primarily to determine whether SNOMED CT-AU is suitable for implementation in Australian general practice. The findings presented in this study raise many issues relating to this aim. This study has used a combination of quantitative and qualitative techniques to demonstrate the scale and types of issues identified during the mapping process. There is very little literature available that refers specifically to SNOMED CT-AU, which has been in existence only since The vast majority of the content in SNOMED CT-AU comes from SNOMED CT. Most research studies about SNOMED CT discussed in this section therefore refer to the SNOMED CT international version, which was released in One of the most important results from this study was the high degree of concordance between frequency of term use (from ICPC-2 PLUS terms used in the BEACH study) and the acceptability of mapping results. This indicated that SNOMED CT-AU contains clinical terms frequently used in general practice, but does not contain some of the terms used less often. Very few researchers have investigated the linkage between the content of clinical terminologies and term usage, so it is difficult to compare these findings with the work of others. The results of Fung et al concur with my findings. They demonstrated that terms that were used frequently were also more mappable, 125 concluding that this result was not surprising because frequently used terms were more likely to be included in a clinical terminology, and this argument has some weight. Overall, the content coverage in SNOMED CT-AU of general practice concepts was disappointing. The ancestry of SNOMED CT includes CTV-3, a version of the Read codes. The Read codes were originally designed for use in general practice, so I expected that the general practice content in SNOMED CT would be better than these results have shown. However, there have been numerous changes to the Read codes over time which may have reduced the content coverage for general practice. These include the development of CTV-3, which widened the scope of the original Read codes 187

210 designed for general practice for use throughout health care, 48 the amalgamation of CTV-3 and SNOMED RT to create SNOMED CT, 197 and the removal of some UK-specific concepts from SNOMED CT to the UK NHS extension Limitations of this study There are a number of limitations to this study. Ideally, the mapping from ICPC-2 PLUS terms to SNOMED CT-AU would have been independently undertaken by two individuals, with conflicting mapping results being resolved by an independent third party. Due to time constraints this did not occur, and I undertook all mapping. To account for this limitation, I consulted my GP supervisor as soon as an issue arose to obtain external input from a clinical perspective. If the mapping results prepared were ever to be released for other purposes, a second independent mapper must review the results prior to use. As well as including terms relating to symptoms and complaints and diagnoses and diseases, ICPC-2 PLUS contains terms for processes of care (covering referrals, clinical and procedural treatments, pathology and imaging orders, and examinations). I did not map any terms from the process components of ICPC-2 PLUS to SNOMED CT-AU, so did not assess the coverage of processes of care in SNOMED CT-AU. My assessment of content coverage in SNOMED CT-AU has been based on the terms in ICPC-2 PLUS being used as a benchmark. I acknowledge that ICPC-2 PLUS is not perfect. I did identify a few issues in ICPC-2 PLUS content during this study, and these have since been rectified. In addition, the version of SNOMED CT-AU used in this mapping was the November 2010 release. It is possible that some of the issues identified in SNOMED CT-AU during this study have been identified and rectified since the November 2010 release. Finally, the mapping methods I used during this study did not include postcoordination. This limitation will be discussed in further detail in Section

211 6.6.2 Mapping terminologies to SNOMED CT-AU The transition process from pre-existing clinical terminologies (sometimes known as legacy terminologies) to SNOMED CT-AU inevitably requires the mapping of content from the legacy terminology to SNOMED CT-AU. The mapping of different types of clinical terms to SNOMED CT has been researched extensively, where the clinical terms used for mapping (that is, the source terminologies) are derived from local termsets, 188,198,199, clinical terminologies, or classifications. 201,205,206 The majority of mapping results reported relate to the mapping of existing problem list terminologies to SNOMED CT, from a variety of settings. Results from mapping projects vary widely, depending on the methods used to perform mapping, the level of specificity contained in the source terminology, and the definition of semantic equivalence used in each study. Mapping from local terminologies to SNOMED CT has also been undertaken in the areas of intensive care, 201 nursing, 207,208 and palliative care. 209 Others have investigated the coverage of specific conditions in SNOMED CT, such as multiple chemical sensitivity 210 and mild traumatic brain injury. 211 Few studies report results from mapping general practice terminologies to SNOMED CT or SNOMED CT-AU. A Canadian study mapped free text problem list terms from EHRs in a single general practice to SNOMED CT, finding matches for approximately 92% of the problem list terms. 188 However, these matches varied in quality, and the authors acknowledged that the representativeness of the results may be limited by the inclusion of one general practice only. In Sweden, 99% of categories (equivalent to rubrics) in the primary care version of ICD-10 (called KSH97-P) were mapped to one or more SNOMED CT concepts. 205 Given the low granularity included in ICD-10 rubric labels, the high proportion of matches in this study is not surprising. If ICD-10 rubric labels were of sufficient specificity for clinical care, clinical terminologies would not be needed. Studies that report the results of maps to SNOMED CT often allowed the creation of postcoordinated expressions as a mapping technique. 188,198,199,204,209,210,212 In 1999 Rector acknowledged the potential 189

212 power of postcoordination to create clinical expressions that allow for breadth, depth and complexity beyond those available in precoordinated clinical terminologies. 50 SNOMED CT was designed to facilitate the creation of postcoordinated expressions, adding to the expressive nature of the terminology while minimising combinatorial explosion. I did not create postcoordinated expressions during the mapping undertaken in this study. At present, no Australian general practice vendor has implemented software capable of creating postcoordination expressions, or expressed interest in modifying their software to support postcoordination. I have no doubt that extending the scope of my study to include postcoordination would have increased the proportion of ICPC-2 PLUS terms covered by SNOMED CT-AU. However, at present the implementation of postcoordination is not planned in Australian general practice, and therefore its inclusion in determining the coverage of SNOMED CT-AU has limited validity from a pragmatic perspective. Given the reluctance of Australian vendors to implement postcoordination, I am somewhat surprised it features so prominently internationally in publications about mapping. The creation of postcoordinated expressions during mapping often increases the coverage of SNOMED CT enormously. For example, Rosenbloom et al reported that more than 90% of a sample of terms from two terminologies (500 terms from each terminology) were mappable to SNOMED CT, and concluded that this represented high coverage of SNOMED CT for the two terminologies tested. The two terminologies in question, MEDCIN and CHISL are interface terminologies used in the US: MEDCIN in several commercial EHR systems; and CHISL at the Vanderbilt University Medical Centre. However, this study included postcoordinated expressions in its assessment of SNOMED CT coverage. The inclusion of postcoordination undoubtedly inflated its results. Postcoordination was not required in only 3.2% of matches in MEDCIN, and 49.1% of matches in CHISL. 203 I have highlighted this study to demonstrate that the presence or absence of postcoordination has a significant impact on the results of studies that assess the coverage of SNOMED CT. It also highlights the heavy reliance on postcoordination needed for the successful and effective implementation of SNOMED CT. 190

213 Although the capacity for postcoordination is generally seen as one of SNOMED CT s primary strengths, it is not without fault. The SNOMED CT concept model includes a set of permitted attributes for each concept which can be used to facilitate postcoordination. Rector and Iannone recently published a study investigating the modelling of the qualifiers acute and chronic in SNOMED CT, reporting that precoordinated concepts containing either of these qualifiers were incorrectly modelled in 20% of concepts containing the word acute and 28% of concepts containing the word chronic. The authors conclude that the modelling of qualifiers such as acute and chronic needs to be standardised. 213 A number of mapping projects used the Unified Medical Language System (UMLS) to facilitate the mapping process. 125,207 The UMLS is a repository that links clinical terminologies and other clinical termsets to enable interoperability, 214 but is not used for clinical care. We have previously used the UMLS as a tool to facilitate mapping, 81 and due to the low number of matches found in that earlier project I decided against its use in developing this map. Although the UMLS is a convenient tool to facilitate mapping, it also introduces a third element into the mapping, which may lead to more information loss than necessary, compared with directly mapping between the source and target terminologies. Matney et al states that it is important for users of the UMLS Metathesaurus to understand that the UMLS Metathesaurus adds its own view of synonymy which may differ from that of an individual UMLS source vocabulary such as SNOMED CT. 207 In other words, an indication of synonymy in the UMLS may not represent true synonymy in a clinical sense, and therefore may not be clinically safe. One important outcome of mapping legacy terminologies to SNOMED CT is the identification of concepts or synonyms that are not included in SNOMED CT. 188,210 This provides a feedback mechanism to the IHTSDO for the inclusion of new content. The IHTSDO has developed an online request submission process for users to submit requests for new SNOMED CT content. 103 The extent to which those involved with mapping projects do provide feedback to either their national release centre or the IHTSDO is not known, however the IHTSDO also reviews published literature to identify 191

214 recommendations for SNOMED CT content from the academic community. 215 This indicates that the IHTSDO is committed to identifying problems in SNOMED CT to assist with improving the quality of the terminology. Assessing both the quantitative and qualitative findings from this study, the content of SNOMED CT-AU in its current form as a precoordinated terminology is not suitable for clinical use in Australian general practice. This could be rectified in two steps: 1. Creating additional precoordinated content. 2. Encouraging EHR software vendors to modify their EHRs to support postcoordination, and changing the culture of data entry to facilitate the entry of postcoordinated expressions. From a pragmatic perspective, significant resources would be needed by vendors to change their EHR database structures to allow the creation of postcoordinated expressions, and then convince users to utilise this functionality. It is unlikely this would be achieved without financial incentives. I suspect that the size of SNOMED CT-AU, with approximately 300,000 concepts, provides a false sense of completeness for EHR vendors who are used to local terminologies or termsets, often containing less than 10,000 terms or concepts. Without a thorough investigation of the content coverage of SNOMED CT-AU, developers could easily assume that SNOMED CT-AU contains the necessary content for their users without the need to include functionality to support postcoordinated expressions Use of SNOMED CT-AU as an interface terminology The significance of the findings from this study are somewhat dependent on the proposed methods of implementing SNOMED CT-AU. NEHTA s guidelines for mapping clinical terminologies to SNOMED CT-AU state that vendors can implement SNOMED CT-AU using one of three methods: 1. Native implementation: where SNOMED CT-AU is implemented at the user interface, and clinical terminologies or termsets previously used (legacy terminologies/termsets) are retired. 192

215 2. Migration from a legacy terminology/termset: where the legacy terminology/termset is mapped to SNOMED CT-AU, and SNOMED CT-AU is subsequently implemented at the user interface. 3. Implementation of a map to SNOMED CT-AU: where the legacy terminology/termset is mapped to SNOMED CT-AU, the legacy terminology/termset continues to be used at the user interface, and the map to SNOMED CT-AU is utilised for interoperability. 216 Terminologies are categorised according to their purpose. Interface terminologies are designed to be implemented at the user interface, and are defined as a maintained set of unique identified terms designed to be compatible with the natural language of the user 16 while reference terminologies are designed to uniquely represent concepts, 16 providing references to alternate terms. SNOMED CT is a reference terminology 100 yet two of the three implementation options listed above (native implementation and migration) involve the native use of SNOMED CT-AU at the user interface, or as an interface terminology. One feature of interface terminologies is the inclusion of synonyms, 217 allowing users to enter information using the term(s) they prefer. Findings from this study do not support the native implementation of SNOMED CT-AU in general practice due to the high proportion of matches that were assessed as either synonyms or best fit matches. The assessment of matches as synonyms in this study indicated that SNOMED CT-AU contained the relevant concept, but did not contain the same phrase as that used in the ICPC-2 PLUS term. For example, the ICPC-2 PLUS term Pain in arm mapped to Pain in upper limb in SNOMED CT-AU. These concepts are equivalent (or synonymous), but Pain in arm is not explicitly included as a synonym in SNOMED CT-AU. There are two ways this problem can be viewed. From the reference terminology perspective, the explicit inclusion of complete synonymy is not required, and it is the responsibility of a separate user interface to ensure that these concepts can be found. However, if SNOMED CT-AU is to be implemented at the user interface, missing synonyms have a significant impact on the usability of the terminology. A few 193

216 studies have tested the validity of SNOMED CT in its current form as an interface terminology. Brown et al found that SNOMED CT performed better as a reference terminology than as an interface terminology, and questioned its usefulness as an interface terminology. 200 Rosenbloom et al concluded that additional work needed to be undertaken, particularly relating to the addition of synonyms, before it could be used in this manner. 203 Internationally, a number of sites have implemented SNOMED CT as a reference terminology, with an associated interface terminology ,201 Osornio et al state that the use of a local interface terminology has two advantages over the use of SNOMED CT natively: it is easier to extend the terminology for local usage, and it isolates the terminology from changes to the international release of SNOMED CT, providing greater stability. They also acknowledge that this approach requires additional work at the local level to develop the links from the local terminology to SNOMED CT, and maintain these links with each six monthly update of SNOMED CT. 187 The three options for SNOMED CT-AU implementation provided by NEHTA indicate that they regard native implementation as a viable option. Native implementation implies that termsets or legacy terminologies currently used would be retired, and SNOMED CT-AU would be used alone, both at the user interface, and as a reference terminology for interoperability. SNOMED CT-AU must be suitable for use at the user interface for this scenario to be feasible, and this would require increasing the number of synonym descriptions available. It is concerning that NEHTA s description of native SNOMED CT-AU implementation refers to the conversion from legacy termsets to SNOMED CT-AU without explicitly stating that a map needs to be developed. 216 There must be a transition process from legacy termsets or clinical terminologies in current use to SNOMED CT-AU to ensure data is linked for historical purposes. Many years of clinical data have been collected using legacy termsets. Implementation of SNOMED CT-AU without developing a map from terms in legacy termsets will limit users ability to easily extract historical data from EHRs. 194

217 Rosenbloom et al state that if SNOMED CT is not to be used as the user interface, the insufficient synonymy is not a significant issue, nor a reason to prevent SNOMED CT implementation. 203 However, if SNOMED CT is to be implemented in native form without an associated interface terminology, the findings of my study have greater implications. My results indicate that there is significant work to be done to SNOMED CT-AU to make it suitable for native implementation in Australian general practice EHRs. On the basis of results presented in this chapter, I do not support the native implementation of SNOMED CT-AU in Australian general practice. It is notable that the IHTSDO website does not state that SNOMED CT is a reference terminology. Instead, the terms clinical terminology and comprehensive terminology for healthcare are used. 218 This may be purely coincidental, or it may represent a future desire that SNOMED CT be promoted as a terminology for the user interface. Until the issue of synonymy has been resolved legacy terminologies must continue to be used, with maps to SNOMED CT-AU created to provide a bridge to SNOMED CT-AU as the common reference point for information reuse and interoperability between systems. There is another separate, but related issue about the use of SNOMED CT-AU at the user interface. As previously stated in Box 3.3, preferred terms are regarded as the most clinically appropriate way of expressing a concept in a clinical record. 100 However, documentation provided with SNOMED CT does not explicitly tell vendors how to display SNOMED CT descriptions (fully specified names, preferred terms and synonyms) at the user interface. Similarly, there is no explicit instruction stating which type of description(s) should be used for different purposes, such as the problem list or for transfer of SNOMED CT-AU concepts for semantic interoperability. This type of instruction is needed by vendors when planning their SNOMED CT implementation, to ensure that SNOMED CT-AU data are stored and transferred in a manner that promotes semantic interoperability. 195

218 6.6.4 Editorial consistency in SNOMED CT-AU content A number of the issues identified in Section 6.5 pertained to editorial issues in SNOMED CT-AU. It is reasonable to assume that the majority of these have been inherited from the SNOMED CT international release. SNOMED CT development is guided by four principles: the involvement of both clinical and informatics professionals; the inclusion of high quality clinical content; an open quality improvement process; and ensuring there are minimal barriers to its adoption. 100 For some time SNOMED CT editorial staff have acknowledged that the achievement of these principles requires delicate balance, and that measuring their achievement is hard to quantify. 219 The SNOMED CT editorial guide states that the most basic principle to determine whether a clinical concept is valid for inclusion in SNOMED CT is whether the concept is valid for semantic interoperability, using the criteria of Understandable, Reproducible and Useful (URU). 102 In SNOMED CT and SNOMED CT-AU, the fully specified name (FSN) is meant to be explicit, that is, the name represents the clinical concept without ambiguity, and includes a semantic tag to indicate the concept s location in the SNOMED CT hierarchies. Therefore, during the mapping I chose the FSN as the benchmark against which I compared the source ICPC-2 PLUS term. However, the preferred term is the SNOMED CT description that is meant to be implemented for clinical use, and the preferred term is not unique. In an implementation scenario, this may become problematic. As I described in Section 6.5.1, the only way to differentiate between some concepts is through the FSN, with some concepts seemingly identical if shown as the preferred term. If preferred terms are shown to users in a clinical setting, some clinical concepts will appear twice, for example Neutrophilia which is both a finding and a disorder. The user could easily assume that the same concept was incorrectly duplicated. The potential implications are substantial, including the loss of semantic interoperability, and the inability to accurately count the frequency of some conditions. In a 2004 paper discussing the design of SNOMED CT, Spackman and Reynoso stated that some of the issues regarded as errors in SNOMED CT are intentional design characteristics, referencing the seemingly identical 196

219 entries in the morphologic abnormality and disorders hierarchies. 219 Although this statement acknowledges that such issues may be perceived as a problem by the user, I have not been able to find any recommendations in the IHTSDO documentation about how this should be managed at the user interface to remove ambiguity. It is not appropriate to use the preferred term as the primary description viewed by the user. The only unambiguous description in SNOMED CT-AU is the FSN, and therefore this should be the description at the user interface. However, I acknowledge that use of the FSN requires the user to have greater understanding of SNOMED CT s structure. This issue will be discussed further in Section Another issue identified was the occasional tendency to include the symptoms of disease as a synonym of the disease/disorder (e.g. GORD and Hypertensive disorder). A single episode of a symptom does not necessarily equate to a disease label; symptom labels and disease labels are not synonymous. This has serious implications for clinical care, with the possibility of the concept s preferred term being added to a patient s problem list instead of the selected synonym, whereby SNOMED CT essentially diagnoses the patient s problem. Overdiagnosis is a growing concern in the field of medicine, but is normally attributed to the broadening of disease definitions 220,221 or identification of subclinical disease through screening. 221 While the use of clinical terminologies has not been discussed as a cause of overdiagnosis in the literature, my findings suggest that overdiagnosis could occur in this manner. The reliability of data derived from SNOMED CT-AU concepts for epidemiological purposes may also be affected, with research studies potentially reporting higher levels of disease/disorders in the population than actually exist, leading to overestimates of disease incidence and prevalence. This issue also has significant clinical safety consequences, particularly if the information is transferred to other health care professionals. For example, a patient could be given medication for a clinical condition they do not have, which may result in adverse events. ICPC provides specific guidance on this issue, suggesting that health problems in general practice cannot always be labelled as diseases and (when appropriate) should be labelled instead as symptoms (symptom diagnoses). ICPC states that recording should be at the 197

220 highest level of diagnostic refinement for which the user can be confident. 40 I therefore strongly recommend that symptoms of disease are removed as synonyms of disease labels in SNOMED CT-AU. In Section I described an issue relating to inconsistent labelling of concepts at different levels of the SNOMED CT-AU hierarchies. Such inconsistency brings into question the accuracy and importance of preferred terms in SNOMED CT-AU, and is confusing for end users. Baud et al assert that the construct of preferred terms has historically been developed by experts rather than users, and they may not suit the needs of users in different contexts, or with different local habits. They support the notion of replacing preferred terms with a bag of terms that are representative of the clinical concept. 222 Moerkerke and Ceusters concur with this assessment, and suggest that the trend towards preferred terms is derived from a desire to overcome ambiguity in term use. 223 However as described in Chapter 4, GPs develop preferences for the terms they use, which indicates that the presence of a pre-defined preferred term may not be clinically useful for individual GPs, and their inclusion is probably not necessary for GP users. Regardless of this, labels of preferred terms must be made consistent to improve their value to users Hierarchical issues Since the first release of SNOMED CT in 2002, a number of authors have criticised the structural, or ontological, framework on which SNOMED CT is based. In particular, many of these criticisms relate to the assumptions created by the SNOMED CT hierarchies, and both the defining and attribute relationships. In this study I did not specifically aim to evaluate the SNOMED CT-AU hierarchies, as this is an area that has been well researched by others with far greater understanding of the underlying description logics in SNOMED CT than I possess, for example Schulz et al, 224,225 Bodenreider et al 226 and Rector et al. 227,228 However, during the mapping process, I did identify problems with the hierarchical structure of SNOMED CT-AU, as described in Section

221 In a recent (2010) study of the application of SNOMED CT by Rector et al, a number of systematic issues with the description logic modelling in the SNOMED CT hierarchies were identified, most of which had widespread consequences. The issues identified were grouped into seven categories, including: omission; incomplete modelling; errors of site; incorrect modelling of anatomy; overgeneralisation; lack of distinction of structure and inconsistent modelling of complications. The authors concluded that the SNOMED CT hierarchies were not suitable for use within their applications (a clinical information system and a system for the ontological development of ICD-11). The authors estimated that a full quality assurance review of one subset of SNOMED CT would take under two years, probably less. 228 Given that the subset used contains less that 10,000 concepts, this implies that a full quality assurance review of all concepts in SNOMED CT would take decades. The conclusions in this paper are strongly worded: Without further quality assurance, clinicians may not realise the implications of what they are saying; researchers may not realise what their queries should retrieve, and postcoordination cannot be expected to be reliable. Interoperability, and therefore meaningful use, will be limited. 228 Rector et al published a second related paper in 2011 in which they described additional problems with the SNOMED CT hierarchies. The authors state that incomplete modelling is found throughout SNOMED CT, and that fixes of the modelling have been undertaken in a haphazard manner, rather than systematically. 229 Based on findings in my study and the examples provided in Section 6.5.8, I agree with the authors that the identification and resolution of issues in SNOMED CT need to be undertaken in a systematic manner Education and user satisfaction Vendor education The usability of SNOMED CT (and SNOMED CT-AU) is heavily reliant on the quality of its implementation. Software vendors who plan to implement SNOMED CT-AU therefore need to have a detailed understanding of the technical structure and content of SNOMED CT-AU. There is a multitude of documentation available about SNOMED CT, including a User Guide (106 pages January 2011 edition), a Technical Implementation Guide (746 pages January 2011 edition) and a Technical Reference Guide (239 pages January 199

222 2011 edition). Additionally there are editorial guidelines, technical documents (for example, specifications for RefSets) and other information including release notes. Although there is significant overlap between the documents, when considered as a whole the documentation would be overwhelming for a vendor, yet a high level of understanding is necessary for successful implementation. For example, vendors need an understanding of the differences between the clinical findings and morphologic abnormality hierarchies, and between findings and disorders, some of which have seemingly identical descriptions. However, the information overload potentially caused by over 1,000 pages of technical documentation may become a barrier to vendors demonstrating a willingness to implement SNOMED CT-AU. Despite the physical amount of technical documentation available, Lee et al stated in 2010 that the guidance provided for the implementation of SNOMED CT is not adequate to be useful to potential future implementers. They specified that more detail is needed in the technical documentation, including details of tools and examples of actual use of SNOMED CT. 209 These conclusions are supported by the work reported by Conley and Benson, who surveyed the health informatics community to determine the type and depth of SNOMED CT knowledge needed by different types of users. Many of the barriers they reported prohibiting the adoption of SNOMED CT related to the need to train health information technology personnel, such as software programmers. More than one-third of respondents totally agreed with the statement the documentation is too long and complex. 230 I agree with the assessments of both Lee at al, and Conley and Benson. As a result of my personal experience evaluating SNOMED CT-AU during this study, I observed that wading through the SNOMED CT documentation is labour intensive. Most of the information needed is included in the documentation, but it is sometimes hard to find relevant information unless you know what you are looking for. Implementation guidance needs to be simplified, in a form that is independent of the current technical 200

223 documentation, with references to the appropriate documentation for more detail. Conley and Benson s survey asked respondents to nominate the amount of formal training received in SNOMED CT according to a scale that included the maximum number of formal training days as three or more, and only one-quarter of respondents selected this category. 230 I assert that three days formal training is not sufficient for a developer to competently understand the complexity involved in implementing SNOMED CT. Detailed formal training programs are needed to educate EHR vendors about the content, structure and implementation of SNOMED CT. The authors of a study describing the creation of a palliative care set of terms coded to SNOMED CT reported that if they could find a match for their source term only in the wrong hierarchy, for example the morphologic abnormality hierarchy rather than the clinical findings hierarchy, they postcoordinated the morphologic abnormality concept with a focus concept of Disease. 209 This demonstrates that vendors will develop workarounds to make SNOMED CT-AU fit their needs. This is potentially dangerous, as the development of ad-hoc workarounds by each vendor would reduce the opportunity for semantic interoperability. The IHTSDO needs to be made aware of the development of workarounds, monitor their existence, and possibly provide clearer implementation instructions to ensure that SNOMED CT can continue to be seen as a tool that promotes semantic interoperability. End user education and satisfaction User satisfaction is paramount to successful implementation of SNOMED CT-AU. Few published studies have examined user satisfaction in relation to the use of SNOMED CT. Elhanan et al reported that 58% of respondents to a survey of SNOMED CT users indicated that they were satisfied with the coverage of SNOMED CT. No more detail about the level or type of satisfaction was given in this paper. 182 An evaluation of an interface terminology based on SNOMED CT found that the primary reason for usability issues in their interface terminology was the 201

224 complexity of the SNOMED CT hierarchies, with users saying they did not understand the hierarchies. 231 This leads to a question of the extent to which users need to be educated in the use of SNOMED CT. Conley and Benson state that clinicians need only a basic understanding of terminology, which does not include technical detail. 230 For most terminologies this is true, but is heavily dependent on the quality of the implementation. At the very least, clinicians need to have an understanding of the SNOMED CT hierarchies and the differences between them. For example, users must understand that the same clinical term may be present as both a finding and a disorder, and understand the definitions of each Maintenance It is well known that clinical terminologies change over time, in response to change in other factors including: medical knowledge; terminology used by clinicians; and the technical structure of terminologies. The need to carefully manage change was acknowledged by both Cimino (who used the term graceful evolution ) 49 and Rector 50 in their seminal papers published in the late 1990s. There are two aspects to maintenance: maintenance of the clinical terminologies that are used to create a map, and maintenance of a map that has previously been created. The clinical terminologies used in the map created during this study are released at different times of the year, and have different update cycles. ICPC-2 PLUS is updated 3 times per year, in April, July and October (with an additional release in January if required). SNOMED CT-AU is updated twice a year in May and November. 232 Three years after the introduction of SNOMED CT, Spackman reported the numbers and types of changes within the terminology since its first release, concluding that the scale of changes required was decreasing. 158 Another seven years on, it could be assumed that SNOMED CT has stabilised further. However, some authors have reported this is not the case. A review of the Clinical Observations Recording and Encoding (CORE) problem list subset revealed that 41% of concepts included in this subset had been modified in some way from one release (January 2009) to the next (July 2009)

225 As previously described in Section 6.6.3, there are three options for SNOMED CT-AU implementation, including one that involves implementation of a map from a legacy terminology to SNOMED CT-AU. If this option is chosen, maintenance of the maps is imperative. The requirement for map maintenance has advantages some of the deficiencies identified during this study may be rectified during the course of SNOMED CT-AU updates, creating a more accurate map over time. However, there are also disadvantages. Map maintenance has the capacity to be a resource intensive task. 187 Wade and Rosenbloom (2009) reported that 71% of legacy concepts mapped to SNOMED CT required maintenance due to changes in SNOMED CT over two releases 18 months apart. 164 This result appears alarming. However, during the time period reported in Wade and Rosenbloom s study, extensive maintenance occurred in one hierarchy (where the Context dependent category hierarchy was modified and re-named the Situation with Explicit Context hierarchy). Such major modifications to SNOMED CT are not a usual occurrence, and as such, I do not believe these results represent the usual changes between SNOMED CT releases. More research is needed about the maintenance burden created by mapping legacy terminologies to SNOMED CT (and SNOMED CT-AU), as the costs of maintenance must be planned for if this path is chosen Conclusion Results presented in this chapter indicate that the content of SNOMED CT-AU does not sufficiently represent the symptoms and complaints, and diseases and diagnoses, recorded in Australian general practice, and thus is not currently suitable for this setting. This conclusion must be considered in light of the methodology used. I restricted mapping to precoordinated SNOMED CT-AU concepts to reflect likely implementation scenarios in Australian general practice. This led to disappointing results, particularly when results were divided into acceptable and unacceptable matches, and indicates that the coverage of SNOMED CT-AU as a precoordinated clinical terminology is not currently sufficient for clinical use in Australian general practice. 203

226 A more promising result from this study was the relationship between the validity of the match from ICPC-2 PLUS to SNOMED CT-AU, and the frequency with which the source terms were used in ICPC-2 PLUS. Frequently used terms were more often acceptably matched to SNOMED CT-AU, indicating that SNOMED CT-AU contains the clinical concepts used most often in general practice, but contains fewer of the terms used less frequently. Examples of issues identified during the mapping process can be used to prioritise content modifications to SNOMED CT-AU, and if all the issues identified can be rectified, SNOMED CT-AU would be suitable for use in Australian general practice. 204

227 7 Discussion This research represents the most comprehensive review undertaken of the use of clinical terminologies in Australian general practice. I have used a bottom-up approach in this research, investigating the patterns of current use and identifying user issues in relation to clinical terminologies. This provided a framework for my examination of the extent to which SNOMED CT-AU contained the clinical terms used in general practice, and subsequently, the suitability of its content for implementation in this setting. I chose this pragmatic approach to enhance the likelihood that any recommendations for implementation will meet the content needs of users. I have identified a number of important issues affecting the use of clinical terminologies in Australian general practice. Findings from the GP and vendor workshops indicated that individual GPs have different attitudes towards clinical terminologies, and different GPs like to use the terms they want to use in clinical situations, rather than a term forced upon them by their system. GPs also expressed concerns about the completeness of clinical terminologies, and the speed with which they could find an appropriate term to enter in their EHR. Vendors stated that they regard the needs of their users as paramount when planning modifications to the structure or content of their EHRs, and indicated that any change in the clinical terminology used must not adversely affect the overall usability of the EHR. After the workshops I investigated the current use of clinical terminology in Australian general practice. I found that 60% of free text terms recorded by GPs were represented in the ICPC-2 PLUS terminology as an exact or synonymous match. However, in 25% of the free text terms recorded by GPs some information was lost in the transition to a coded term, the majority of the lost information being attributes such as laterality or severity. When I assessed the frequency of use of ICPC-2 PLUS terms as either RFEs or problems managed in the BEACH study of general practice, results showed that GPs use a few clinical terms a lot of the time to describe RFEs and problems managed, and that the vast majority of terms recorded by GPs are used infrequently. I also found that GPs record processes of care as problems 205

228 managed when they are permitted to do so. The maintenance patterns for ICPC-2 PLUS indicated that clinical terminologies need regular maintenance, even many years after their introduction. Finally, results of the map from ICPC-2 PLUS terms in Components 1 (symptoms and complaints) and 7 (diagnoses and diseases) to SNOMED CT-AU indicated that only 70% of ICPC-2 PLUS terms were acceptably matched to SNOMED CT-AU concepts, but less than 50% of ICPC-2 PLUS terms were explicitly included in SNOMED CT-AU. A higher proportion of matches were acceptable for diagnosis/disease terms than for symptom/complaint terms, indicating that SNOMED CT-AU represents diagnosis/disease labels used by Australian GPs more comprehensively than it does symptoms and complaints. Mapping results were combined with the previously reported results about frequency of ICPC-2 PLUS term use, showing that terms used frequently were also more often mappable to SNOMED CT-AU. In its present form, my research has indicated that SNOMED CT-AU is not currently suitable for implementation in Australian general practice. SNOMED CT has existed for 10 years, and SNOMED CT-AU for three years. As a result, both terminologies are still maturing. At present, literature specifically about SNOMED CT-AU is sparse. Literature about SNOMED CT is increasing internationally in the form of evaluations and implementation reports, so the body of knowledge about SNOMED CT is growing rapidly. In addition, regular (six monthly) updates improve the quality of SNOMED CT and its functionality, and these improvements are reflected in SNOMED CT-AU revisions. The IHTSDO has stated that they review the literature regularly to identify recommendations for changes to SNOMED CT content from the wider community. 215 For example, a high profile, critical review of SNOMED CT was published by Rector in The IHTSDO has publicly stated that it is working to overcome the issues identified in this review. 215 Further, the IHTSDO Annual Quality Report for 2011 states that the IHTSDO had commissioned a review of content, 215 although the results of this review are not publically available. 206

229 The primary purpose of clinical terminologies is to label clinical information in a standardised manner to facilitate re-use, 60 either by the clinician who entered the information, or by other health care providers. It is therefore imperative that clinical terminologies contain sufficient content to allow users to record information at the level of detail required, so that it is captured accurately and can be subsequently re-used in a reliable manner. Some researchers suggest that coded information should represent clinical information in a form that is as close as possible to common parlance. 125,126 When GPs free text descriptions were compared with their equivalent coded term (Section 5.2), a relatively high proportion of terms were assessed as a best fit, that is, the coded ICPC-2 PLUS term was broadly comparable to the free text, but there was no lexical or semantically equivalent match between the free text and the coded ICPC-2 PLUS term. This suggests that some information included in GPs free text descriptions is not easily captured in a clinical terminology. Defining the amount of information that should be explicitly included, or encapsulated 129 in a coded term is a subjective exercise. The Australian GP Data Model and Core Data Set project, completed in 1999, advised the use of a free text field in conjunction with every codeable data field in an EHR, 12 to capture information not required for reporting or decision support, but regarded by the GP as important for individual patient care. The accurate capture of clinical information (either through a terminology or free text) is crucial for high quality care. The POMR has already been discussed a number of times in this thesis. It supports the notion of a problem label changing over time in light of increased understanding of a problem (e.g. a first presentation of abdominal pain may, after investigation, be diagnosed as appendicitis), thus supporting continuity of care for individual patients. Continuity of care is specifically included in the RACGP standards for general practice. One important aspect of continuity of care is informational continuity, 233 which focuses on access to clinical information for re-use, both for individual patients over time (as discussed above in relation to the POMR) and when care (and thus the clinical information needed for clinical care) is transferred between providers. 207

230 There is a strong relationship between high quality clinical care and clinical safety. The RACGP states that inadequate or inaccurate recording of information in a medical record is one factor that affects patient safety. 95 The Australian Commission on Safety and Quality in Healthcare has released a set of guidelines for clinical handover to aid clinicians when transferring the care of patients, either within or outside a clinical setting (e.g. general practice), which include guidance on the exchange of clinical information during handover. 234 Some of my results have important clinical safety implications. The mapping of ICPC-2 PLUS terms to SNOMED CT-AU in Chapter 6 demonstrated that some descriptions regarded as synonyms in SNOMED CT-AU were not clinically synonymous with the concept s fully specified name, which is meant to unambiguously represent the clinical meaning of the concept (for example, acid reflux, a symptom of GORD, was included as a synonym for GORD). The importance of ensuring that all descriptions associated with a concept are synonymous with the concept s fully specified name cannot be underestimated. Mislabelling of health problems creates the potential for serious harm to patients, particularly if information is transferred to other health care providers who may receive a problem label and advise treatment based on the label without further clarification. For example, patients could be inappropriately prescribed a medication which has the potential to cause adverse outcomes. Clinicians have both an ethical and medicolegal responsibility for the information in their medical records. Thus clinicians who use clinical terminologies must review the terms they record, to ensure that the clinical information entered is accurate. The relationships inherent within SNOMED CT-AU concepts mean that data entered as a synonym (e.g. acid reflux) could be converted to a preferred term or fully specified name (e.g. GORD) for data extraction without the user s knowledge. Developers of clinical terminologies have a responsibility to their users (i.e. clinicians) to ensure that the information contained in terminologies can be recorded and interpreted unambiguously. Any non-synonymous descriptions in SNOMED CT-AU must be immediately removed, and safeguards introduced to mitigate the associated clinical risks. 208

231 The accurate labelling of clinical information is not only reliant on clinical terminologies. Some information written in free text by GPs but not captured using ICPC-2 PLUS (described in Section 5.2) could be captured using information models rather than clinical terminologies. In terminological circles, the information model refers to the database structures that act as the bucket into which coded data are entered. 129 Information models include data elements, which can be structured to guide the user to enter particular types of data for specific purposes. For example, it is important to differentiate between a patient s current problem and their family history of a problem, so EHR developers could create separate data elements for Problem/diagnosis (or equivalent) and Family history of disease. Terms/concepts from clinical terminologies can then be bound or attached to these data elements to create expressivity that contains the selected term/concept and the data element. 124 For example, a patient presents with a family history of heart disease. This information could be explicitly included in the clinical terminology as the term Family history of heart disease. Alternatively, a data element called Family history could be included in the EHR and populated with a term from the clinical terminology (e.g. Heart disease ). When the data element and the coded term are used in combination, the expression Family history of heart disease can be obtained. Although conceptually this appears relatively simple, there is uncertainty about whether particular types of information (including family history) are best captured using the information model or the terminology model. 235 In my opinion, contextual information should not be included within precoordinated concepts in a clinical terminology. Using the example above, the inclusion of different types of history (e.g. personal history, family history) as precoordinated concepts for every possible clinical condition would lead to combinatorial explosion, 50 which I have already discussed in Chapter 5. Combinatorial explosion should be avoided where possible, as the increase in terminology size reduces the usability of the terminology. If contextual information can be alternately captured using the information model without loss of semantic meaning, this option should be adopted. 209

232 One of the overarching themes emerging through this thesis is the need for standards for EHRs, including both technical and content standards. The need for standardisation in EHRs is often espoused. Standards Australia defines standards as published documents setting out specifications and procedures designed to ensure products, services and systems are safe, reliable and consistently perform the way they were intended to. 236 I have no doubt that standards are underutilised in Australian general practice EHRs. A number of technical standards for GP EHRs exist, including the functional requirements specification for general practice computer systems prepared by IBM Consulting Group in 1997, 237 and the General Practice Core Data Model and Core Data Set Project, completed in However, these technical specifications were never implemented in Australian general practice EHRs, primarily because there was no incentive for, nor obligation on, GP EHR vendors to modify their systems to include these standards. My research has primarily focussed on the need for EHR content standards through the use of clinical terminologies, such as SNOMED CT-AU. No content standards (in the form of standardised clinical terminologies) have yet been adopted in Australian GP EHRs. I have previously commented that all GP EHR systems currently use proprietary terminologies or homegrown termsets. This is partially because none of the terminologies available were labelled as a standard until the purchase of a national licence for SNOMED CT. An attempt was made to find a standard clinical terminology in the early 1990s, which led to a field test of the Read codes in Australian general practice (the Aus-Read trial). 61 However, poor results from this trial meant that a licence for the Read codes was not purchased by the Australian Government. Thus, there was no guidance for vendors about the implementation of clinical terminologies, leading to the current situation in which each GP EHR vendor has either adopted a pre-existing terminology such as ICPC-2 PLUS, or developed their own termset (e.g. the Medical Director termset). The purchase of a national licence for SNOMED CT and the subsequent development of SNOMED CT-AU mean that a standard for clinical terminologies in Australia now exists. However, this has not been 210

233 accompanied by the development of technical standards for GP EHRs. The two technical specifications documents for GP EHRs listed above were created up to 15 years ago. Although the principles espoused in these specifications remain valid, technology has improved so markedly over this period that new specifications are needed, but work on such specifications has not yet commenced. The development of new specifications for Australian GP EHRs must be based on international developments currently occurring. ISO is updating the European standard CEN 13940: Health informatics - System of concepts to support continuity of care. The acceptance of this standard by ISO will provide an international benchmark for the future development of Australian GP EHR specifications. My findings suggest that labelling of a product as a standard (such as a clinical terminology or a technical specification) does not necessarily mean that the standard is suitable for implementation in particular use cases. SNOMED CT-AU is regarded as a standard, and according to its title the IHTSDO is a standards development organisation. My research indicates that the content in SNOMED CT-AU must be enhanced and modified prior to its implementation in Australian general practice before it meets the criteria of being safe and reliable. This demonstrates that products promoted as standards are not perfect, and need to be updated regularly in light of new knowledge and technology. Although one of the principles of SNOMED CT development is to minimise barriers to its adoption, 100 there are barriers to the adoption of clinical terminologies in general, 238 and SNOMED CT in particular, that need to be addressed. Some barriers relate directly to the terminology itself, while others relate to implementation. One of these barriers is the lack of incentives for vendors to implement standards such as clinical terminologies. Lack of market demand is also a barrier to the adoption of clinical terminologies. 181 This was indirectly mentioned during the workshops described in Chapter 4, when vendors stated that users do not ask specifically for terminologies. However, this comment was clarified by a separate statement that users do ask for mechanisms to facilitate data extraction. It can 211

234 be inferred therefore that users demand specific outcomes from their EHRs, but do not directly ask for the processes needed to derive these outcomes. Essentially, the use of clinical terminologies in a clinical setting can be seen as a process or a means to an end, with the end point representing a desired outcome, for example data extraction or transmitting clinical concepts to another clinician via a referral letter. It is not surprising that users do not put a name to this process (that is, they do not specifically ask for clinical terminologies), as only clinicians with a high level of knowledge about information technology (IT) or clinical terminologies will realise that data extraction can be facilitated through the use of clinical terminologies. It is therefore the responsibility of vendors to assess feedback from their users, and to identify the processes that need to be implemented to address user requests. Another potential barrier to the introduction of SNOMED CT-AU is that users are satisfied with the proprietary termset or clinical terminology they are currently using and do not feel the need to change, particularly if it involves change to the user interface, or to the methods used to enter data. Perceived barriers to the implementation of SNOMED CT were described in the results of an online survey by Conley and Benson. The majority of respondents in this study were from the UK. The barrier rated highest by respondents (primarily health IT professionals and clinicians) was the financial and time investment needed to train staff (59.4%), followed by the complexity of SNOMED CT (57.5% of respondents). 230 Overcoming barriers, including those described here, is essential to ensuring the successful implementation of SNOMED CT-AU in Australian general practice. Incentives or drivers may be needed to overcome some barriers. Often, these incentives are external, in the form of additional funding from governments to EHR vendors to offset the cost to vendors of implementing new standards, or incentives to users who adopt a new system. However, not all incentives are financial. A number of successful initiatives in Australian general practice have been introduced through their inclusion in practice 212

235 accreditation standards, whereby practices must adopt the new initiative if they are to be an accredited practice. A clear finding from both GPs and vendors at the workshops was that the interface used in conjunction with clinical terminologies must facilitate ease of use. A user interface that facilitates ease of use can be achieved through the design of search mechanisms within the EHR; and the usability of the clinical terminology installed in the EHR. The coverage of a terminology, that is, the completeness of its content, is one of the primary factors that facilitates ease of use. GPs described their frustration when they could not find the term they wanted to use, and stated that they give up when they cannot find an appropriate term. From this perspective, the results from the mapping of ICPC-2 PLUS terms to SNOMED CT-AU are alarming, with less than half the ICPC-2 PLUS terms having an explicit match to a SNOMED CT-AU concept (including exact matches and description matches). Although another 22% of ICPC-2 PLUS terms were synonymous with SNOMED CT-AU concepts, searching for synonyms requires additional effort by the user, and it is this process that causes frustration. My findings in Chapter 5 that GPs use a few terms a lot of the time could be seen to contradict GPs statements in Chapter 4 that they regularly cannot find terms they wish to use. Collectively, these results create a quandary for terminology developers it is easy to include high frequency terms/concepts in a clinical terminology, but as demonstrated in Figures and 5.3.2, the tail of terminology use is long, and determining the appropriate size for a clinical terminology is highly subjective. I hypothesise that the terms GPs cannot find in their current clinical terminologies are not used often in Australian general practice, perhaps because these labels represent rare diseases some of which may have been diagnosed by other clinicians (e.g. medical specialists). Although rare diseases may not be seen often in general practice, every patient with a rare disease will see their GP, and GPs must be able to enter all problems on a problem list, regardless of the overall frequency of the disease in the population. 213

236 Use of the RefSet mechanism in SNOMED CT-AU can be used to overcome this problem, by creating a subset of SNOMED CT-AU content for particular use cases (such as general practice). Creation of a general practice RefSet can facilitate use of SNOMED CT-AU by limiting SNOMED CT-AU concepts immediately visible, to those concepts most often used in general practice. This allows users to search for SNOMED CT-AU concepts faster and more efficiently, increasing the likelihood of finding an appropriate concept quickly. Although such RefSets restrict the SNOMED CT-AU content available to the user, it is possible to install all the SNOMED CT-AU content in the background of the EHR (not initially visible to the user), and create a mechanism for users to secondarily search this additional content if they cannot initially find an appropriate concept in the RefSet. The RefSet mechanism has the potential to enhance the use of SNOMED CT-AU as a clinical terminology across all clinical specialties, and may provide a solution to allow users to find content beyond the usual scope of their clinical specialty, on the proviso this content exists in SNOMED CT-AU. In the UK, attempts have been made to standardise the interface used by clinicians through the Microsoft Common User Interface. 116 In recent years, the increasing use and efficiency of internet searching has improved user interfaces in general terms. The interface used by Google has become ubiquitous, and as time progresses users will demand interfaces in EHRs that have faster and more accurate search results to match those in Google. Results presented in this thesis have demonstrated that the interface used to access clinical terminologies is of vital importance when determining acceptability or suitability of a clinical terminology. Vendors at the workshops spoke of user acceptability in terms of the number of mouse clicks, and made it very clear that terminology interfaces that are slower than current interfaces are not acceptable. Therefore, those implementing clinical terminologies must consider, in combination, the terminology s content and the utility of the user interface, to ensure that the implementation will be acceptable to users. When deciding on a methodology to create the map from ICPC-2 PLUS terms to SNOMED CT-AU, I decided to limit the map targets to precoordinated SNOMED CT-AU concepts, and did not test the extent to which 214

237 postcoordinated expressions could be created when there was no explicit match in SNOMED CT-AU. Postcoordination is the function by which two or more individual SNOMED CT-AU concepts are combined to create a new expression. 100 Australian GP EHR vendors have not shown interest in implementing postcoordinated expressions, and, as the focus of my research was to test the suitability of SNOMED CT-AU for Australian general practice, I limited the target matches to precoordinated concepts to imitate the most likely implementation scenario. The mapping results presented in this thesis must be considered in light of this limitation. I demonstrated that less than 50% of ICPC-2 PLUS terms covering symptoms and complaints, and diagnoses and diseases, were explicitly included as precoordinated concepts in SNOMED CT-AU. Postcoordination is one method by which the coverage of SNOMED CT-AU could be improved for Australian general practice, however the extent to which the use of postcoordination could overcome some of the content deficiencies identified in this study depends on the type of SNOMED CT-AU implementation chosen. It is possible that the inclusion of postcoordination as a mapping technique may raise the proportion of acceptable matches to a level deemed suitable for the implementation of SNOMED CT-AU in Australian general practice. This is an area for future research. The easiest way to implement postcoordination would be by way of a map from an existing terminology to SNOMED CT-AU that includes postcoordinated expressions. In this scenario, postcoordinated SNOMED CT-AU expressions sit in the back-end of the EHR and user interfaces are not modified. This would not require much change to existing EHRs, at least from the users perspective. In contrast, provision of the facility to create SNOMED CT-AU postcoordinated expressions at the user interface requires change to both the structure of the EHR and the user interface. Users are not currently familiar with postcoordination, so they would also need to modify the processes they use to enter clinical data into their EHRs, and would need education about how to do so. If users needed to create a postcoordinated expression, they would need to enter multiple concepts to create this expression. Given the statements made by GPs in Chapter 4 about the need for speed and 215

238 efficiency of data entry, users may not (initially at least) easily accept the introduction of a clinical terminology that requires more complex data entry. However, results in Section 5.2 clearly demonstrated that ICPC-2 PLUS could not capture some data GPs wrote in free text, such as attributes including severity and laterality. This suggests that in the long term, data collected using a terminology that permits postcoordination may be more complete than data captured using termsets and clinical terminologies currently available. The benefits and costs of implementing SNOMED CT-AU with postcoordination must be carefully considered by vendors. The market for GP EHRs in Australia is relatively small and extremely competitive, with approximately ten EHR products used across general practice. Vendors may be reluctant to change their user interface for fear of losing customers to a competitor. At this stage it appears the risk of modifying the user interface may be too high for vendors, and significant incentives would be needed before this could become a viable option. The presence or absence of postcoordination during SNOMED CT-AU implementation has consequences for the amount of work needed to improve SNOMED CT-AU. If postcoordination is to be implemented, creating some additional content in SNOMED CT-AU may not be necessary, as some clinical concepts identified as missing in my study could be constructed using postcoordination. Therefore, one of the first steps needed is a national implementation plan for SNOMED CT-AU that includes a decision about the support or rejection of postcoordination. If postcoordination is supported, significant effort will be needed to define the development work required of EHR vendors to facilitate its implementation, including details of incentives needed by vendors for adoption. Rules for postcoordination must be developed, to specify the extent to which postcoordination is permissible. Users may be allowed to create expressions including separate concepts (e.g. vomiting and diarrhoea), or postcoordination may be limited to expressions containing attributes such as laterality and severity (e.g. asthma with severity levels of mild, moderate or acute). SNOMED CT (and thus SNOMED CT-AU) contains sets of permissible values 216

239 for attributes that ensure postcoordinated expressions are useful and clinically meaningful. 100 Hypothetically, expressions that are nonsensical such as left asthma cannot be created if these constrained attribute sets are used during the implementation of postcoordination. This could be achieved through the use of templates that link SNOMED CT-AU concepts with their permissible attributes. However, the validity of the SNOMED CT attribute sets has not yet been tested in general practice, and this must occur prior to their introduction. If a decision is made to not adopt postcoordination, attention must focus on adding new precoordinated content to support the clinical terminology needs of users in Australia. Postcoordination does have a role in overcoming some of the deficiencies identified during the mapping of ICPC-2 PLUS terms to SNOMED CT-AU. This is an area of further research I intend to pursue, to determine the extent to which postcoordination can rectify deficiencies identified during the mapping undertaken in this thesis, and the deficiencies that remain. Regardless of whether postcoordination is supported during implementation, some additions to SNOMED CT-AU content must occur, particularly for clinical concept areas I found to be missing when mapping from ICPC-2 PLUS terms to SNOMED CT-AU. Above, I stated the need for a national implementation plan for the introduction of SNOMED CT-AU in Australian general practice. I have already discussed the current lack of standards in Australian GP EHRs, and believe that there must be a proactive approach towards SNOMED CT-AU implementation to ensure that this standard is implemented in a standardised manner across all GP EHR vendors. This approach will promote semantic interoperability, and will enhance the likelihood of successful implementation. A plan for the introduction of SNOMED CT-AU into Australian general practice must be driven by the needs of GPs. Professional organisations representing GPs, including the RACGP, ACRRM, the AMA, the Rural Doctors Association of Australia, the Australian Medicare Local Alliance and General Practice 217

240 Registrars Australia must therefore take a leadership role in the development of such a plan to ensure the process is underpinned by clinical governance. As the body responsible for practice accreditation standards, the RACGP also has a role in determining the extent to which SNOMED CT-AU implementation can be used as an accreditation standard. Results from the workshops indicated that the early involvement of GP EHR vendors is vital to gaining their co-operation. SNOMED CT-AU cannot be implemented in any form without the support and involvement of vendors. Their early engagement and input will result in greater pragmatism in the resulting implementation plan, as vendors are best suited to identify potential barriers to implementation, from both a technical perspective, and from the perspective of their customers. As the organisation responsible for SNOMED CT-AU development in Australia, NEHTA must be heavily involved in planning the implementation of SNOMED CT-AU in Australian general practice. NEHTA is also directly responsible for the content additions and modifications specified in the findings of this thesis. Policy makers must be heavily involved in the development of a SNOMED CT-AU implementation plan. The Federal Government is responsible for funding Australian general practice. If financial incentives are needed they are likely to be provided by the Federal Government, perhaps in the form of funding to GP EHR vendors to implement SNOMED CT-AU. Such incentives could act as an enabler to offset the considerable costs involved with major software modifications. Those developing the plan for SNOMED CT-AU implementation in general practice should carefully consider the results about the current use of clinical terminology presented in Chapter 5. The results in Chapter 5 were based on terms used in the ICPC-2 PLUS terminology. A substudy reported in Section 5.2 demonstrated ICPC-2 PLUS adequately represented the free text terms used by Australian GPs, although GPs recorded some information in free text (such as severity and laterality) that could not be captured using ICPC-2 PLUS. ICPC-2 PLUS has evolved over more than 15 years. It is 218

241 regularly updated based on feedback from GP EHR developers and end users, and through the free text terms recorded in the BEACH study of general practice, which are secondarily coded using ICPC-2 PLUS. This bottom-up approach, and ongoing evolution indicate that it is a valid benchmark against which clinical terminology use in general practice can be assessed. I have identified two streams of work needed to enhance the content coverage of SNOMED CT-AU for use in Australian general practice. Firstly, areas of missing content (identified by the mapping undertaken in this thesis) must be addressed by creating new content in SNOMED CT-AU. ICPC-2 PLUS terms that were used frequently but not present in SNOMED CT-AU (i.e. with a map result category of no match ) must be prioritised for addition as new concepts. Also, ICPC-2 PLUS terms that were not acceptably mapped to concepts in SNOMED CT-AU (i.e. those assigned to the map result category of best fit ) must be examined to determine how they can be represented more accurately in SNOMED CT-AU. The addition of new synonyms in SNOMED CT-AU should be considered by reviewing all ICPC-2 PLUS terms with a map result category of synonym, to improve the coverage of synonyms in SNOMED CT-AU. Secondly, some existing content in SNOMED CT-AU must be modified. The issues outlined in Chapter 6 indicate there are a number of areas that need modification to remove ambiguity, or to remove inconsistencies in the labels used to describe SNOMED CT-AU concepts. My research focussed on two types of clinical information commonly recorded in general practice symptoms and complaints, and diagnoses/diseases, as recorded in the RFE and problem/diagnosis data elements. The map from ICPC-2 PLUS terms to SNOMED CT-AU was limited to terms from the symptoms and complaints, and diagnoses/diseases components of ICPC-2 PLUS. I did not assess the suitability of SNOMED CT-AU in other clinical areas (e.g. clinical treatments or surgical procedures). Further research is needed to determine the suitability of SNOMED CT-AU to capture other types of information recorded in Australian GP EHRs. 219

242 Although more related to the structure of SNOMED CT-AU than its content, I also identified inconsistencies in SNOMED CT-AU hierarchies that must be rectified. The implications of the hierarchical issues identified focus on data extraction rather than on data entry. A considerable amount of work is being undertaken at an international level to map SNOMED CT concepts to international classifications such as ICD and ICPC to assist data extraction for epidemiological purposes. The hierarchies in SNOMED CT-AU are also promoted as a tool to aggregate (or group) concepts for data extraction. 100 However, my research indicates that SNOMED CT-AU hierarchies are not complete for this purpose. One example provided in Section 6.5.8, Thyroid disease in pregnancy, showed that this concept was hierarchically linked to thyroid disorders only, and was not hierarchically linked to the concept Disorder of pregnancy. As a result, this concept would not appear if users utilised only the SNOMED CT-AU hierarchies related to Disorder of pregnancy to extract information about disorders experienced during pregnancy. Inconsistencies such as these must be resolved if SNOMED CT-AU is to be used for data extraction. ICPC-2 PLUS was chosen as the source for the mapping to SNOMED CT-AU because it is a clinical terminology currently used in Australian general practice, and the implementation of SNOMED CT-AU necessitates transition from existing clinical terminologies to SNOMED CT-AU through the development of a map. I examined the extent to which ICPC-2 PLUS represents the free text used by Australian GPs to validate its use as a source terminology to use for the mapping to SNOMED CT-AU, determining that it was fit for this purpose. However, the research presented in this thesis did not map the free text terms written by GPs in BEACH directly to SNOMED CT-AU. It is possible that some information recorded in free text but not captured in ICPC-2 PLUS may be explicitly included in SNOMED CT-AU, and this should be investigated in the future. Based on the mapping from ICPC-2 PLUS terms to SNOMED CT-AU, in its current form SNOMED CT-AU does not include the content needed by Australian general practice and is not suitable for implementation in this setting. Prior to any implementation, I therefore strongly recommend that the 220

243 content additions and modifications outlined in Chapter 6 are considered for inclusion in SNOMED CT-AU. ICPC-2 PLUS terms that were unacceptably matched to SNOMED CT-AU should be prioritised for the addition of new concepts or modification of existing concepts above those with low usage. I will provide details of the issues identified in this thesis to NEHTA and the IHTSDO through their request submission processes. Until these modifications have been finalised, GPs should continue using existing proprietary termsets and clinical terminologies. However, if the additions and modifications outlined in this thesis are addressed, SNOMED CT-AU will be suitable for implementation in Australian general practice in the future, and recommend that in the short term, SNOMED CT-AU be implemented in the form of maps from existing (or legacy) terminologies to SNOMED CT-AU concepts to minimise the impact of implementation. When introduced in the form of maps, SNOMED CT-AU is not implemented at the user interface. Users continue to enter clinical information using their current termset or clinical terminology, and a map from the current terminology to SNOMED CT-AU is implemented at the back end of the software. In this way, there is no demonstrable change to the methods used for data entry. However, data can be converted to SNOMED CT-AU and extracted in the form of SNOMED CT-AU concepts for transfer to other health care providers (e.g. through referral letters). This creates a layer of semantic interoperability that is not currently available using proprietary terminologies, and negates the need for change at the user interface. Maps must be updated each time either the proprietary terminology or SNOMED CT-AU is updated. Although in my opinion the implementation of maps to SNOMED CT-AU is currently the most pragmatic short-term implementation solution, ongoing updating of these maps will be resource intensive in the long term. Maps from legacy termsets and terminologies to SNOMED CT-AU must be created and updated in a manner that is reproducible. An ISO international standard outlining the principles for mapping between terminologies (including classifications) is currently under development. 241 On its release, this standard 221

244 will create the benchmark for all mapping projects. In the meantime, NEHTA has created a guidance and requirements document that outlines the processes, tools and quality controls that should be used when mapping to SNOMED CT-AU. 216 In the longer term, the use of SNOMED CT-AU at the user interface could be considered, although I have considerable reservations about this form of implementation. The issues identified in my research suggest that many changes are needed before SNOMED CT-AU can be considered clinically safe and suitable in this form. Studies that have investigated the use of SNOMED CT as an interface terminology have not found it appropriate for this use (as discussed in Chapter 6). As such, further research about the use of SNOMED CT-AU as an interface terminology in Australian general practice is needed before this implementation option should be considered. Implementation of SNOMED CT-AU at the user interface will be resource intensive, as vendors must change both the back end structure and the user interface to enable SNOMED CT-AU concepts to be recorded. In addition, end users must be informed about the change in approach, and education must be provided to ensure that users can accurately capture clinical information. Regardless of the implementation type chosen, the use of SNOMED CT-AU must be field tested by GPs to ensure that it is suitable for use in Australian general practice prior to large scale implementation. 222

245 8 Conclusion In 1999, when considering the reasons why development of clinical terminologies is hard, Rector concluded: We will know we have succeeded when clinical terminologies in software are used and re-used, and when multiple independently developed medical records, decision support, and clinical information retrieval systems sharing the same information using the same terminology are in routine use. 50 SNOMED CT was first released three years after this statement was published. Promoted as the most comprehensive clinical terminology in the world, 43 its release raised expectations that the theorised benefits of clinical terminologies (as summarised by Rector) may be realised on a global scale. In this context I evaluated the content of the Australian version of SNOMED CT (SNOMED CT-AU) to determine its suitability for use in Australian general practice. I found that its current content needs significant alterations (including new content and modifications to existing content), and concluded that SNOMED CT-AU is not currently suitable for implementation in Australian general practice. The importance of this finding is reflected in statements from GPs that clinical terminologies must contain sufficient content, and also be fast and easy to use. GPs use a few clinical terms most of the time to describe RFEs and problems managed, while many terms are used infrequently. The high association between clinical terms used frequently and those acceptably mapped to SNOMED CT-AU suggests that the work needed prior to SNOMED CT-AU implementation is achievable, but some time will be needed to resolve the issues identified. It is imperative that modifications are made prior to the start of any implementation process, regardless of the type of implementation proposed. Due to the high proportion of general practice terms not explicitly included in SNOMED CT-AU I have serious reservations about the use of SNOMED CT-AU at the user interface, and believe that it needs considerable modification before it can be considered for implementation in this form. I conclude that the adoption of SNOMED CT-AU in Australian general practice 223

246 is likely to be most successful in the short term if implemented in the form of a map from existing clinical terminologies to SNOMED CT-AU. An important first step is a national plan for the implementation of SNOMED CT-AU in general practice. This plan must involve all stakeholders and be driven by the needs of clinical users. Many decisions about SNOMED CT-AU implementation are yet to be made the use of postcoordinated expressions and the development of RefSets to facilitate search and display of SNOMED CT-AU concepts are examples of issues yet to be resolved. It is imperative that such decisions are made nationally to ensure that SNOMED CT-AU is introduced in a standardised manner across all Australian GP EHR vendors. The field of clinical terminology emerged more than 500 years ago, and in some way all clinical terminologies are influenced by this history. The desire to wipe the slate clean and start afresh may be tempting. However this is simply not practical. We therefore must acknowledge the limitations of terminologies currently in existence, continue to identify and rectify their deficiencies and work towards making the development and implementation of clinical terminologies as easy as possible. 224

247 Postscript SNOMED CT-AU has not yet been implemented in Australian general practice. The GPRS project, described in Chapter 4, was designed to promote the introduction of SNOMED CT-AU in this setting through the development of a general practice SNOMED CT-AU RefSet. In 2010 NEHTA cancelled the GPRS project after the requirements and methods were developed, but before mapping commenced. To date, the GPRS project has not been completed, and to my knowledge no further work has been undertaken to actively support the implementation of SNOMED CT-AU in Australian general practice. However, a number of initiatives have run in parallel to the research undertaken in this thesis that may influence the future introduction of SNOMED CT-AU in Australian general practice. In theory, the development of the PCEHR provided an ideal opportunity to introduce standards in EHRs that would promote semantic interoperability. As stated in Chapter 5, the PCEHR will act as an online summary of patients health information but is not a full EHR for use at the point of care. Information contained in EHRs used at the point of care will be uploaded into the PCEHR to populate the repository. However, the specifications for the PCEHR do not promote standardisation for GP EHRs used at the point of care. In the PCEHR the data fields for problem/diagnosis in shared health summaries and event summaries are labelled as codeable text, with the value set (clinical terminology) listed as SNOMED CT-AU. In the context of the specifications, this means that SNOMED CT-AU concepts should be used to populate the problem/diagnosis data element but are not mandatory. Other clinical terminologies/termsets or free text can be used as alternatives. 148,149 In other words, there is no restriction on what can be entered as a problem/diagnosis in the PCEHR, and thus, no standardisation. This is perhaps not surprising. The initial timeframe for the entire PCEHR project was two years. 242 Patients were able to register for a PCEHR from July 1, 2012, however at this stage the amount of information included is minimal and will be increased incrementally. 143 Currently, data from Medicare is being added to the system, 243 and it was expected that the functionality to 225

248 support the uploading of shared health summaries into the PCEHR from GP EHRs would commence in September Although the potential for standardisation of clinical terminology through the use of SNOMED CT-AU in the PCEHR still exists, the fact it was not included at the start removes any immediate impetus for its introduction. NEHTA has stated that it will support the PCEHR specifications for at least two years, and the specifications will remain stable unless issues arise relating to patient safety, legislation, or changes to national infrastructure. 245 This implies that the use of SNOMED CT-AU in the PCEHR will not be mandated for some time. However, the mapping results presented in this thesis suggest that content enhancements are needed in SNOMED CT-AU prior to its introduction. An opportunity therefore exists to resolve the issues identified in my research prior to SNOMED CT-AU implementation in the PCEHR. A harmonisation agreement between the IHTSDO and Wonca for the harmonisation and conjoint use of SNOMED CT and ICPC-2 in general practice internationally was signed in 2009, 246 indicating that the use of SNOMED CT in general practice is a high priority for the IHTSDO. The IHTSDO has given responsibility for the general practice content of SNOMED CT to the International Family/General Practice Special Interest Group (IFP/GP SIG) (chaired by a Wonca nominee), 246 of which I am a member. As part of the harmonisation agreement, I am the Project Manager of a project to create an international SNOMED CT RefSet for use in general practice, and map this RefSet to ICPC-2 to facilitate international comparisons of data for epidemiological purposes. 240 Similar to the approach I used in this thesis, the international RefSet is being developed using a bottom-up approach, with general practice termsets and terminologies currently used in different countries forming the basis of the RefSet content. ICPC-2 PLUS is one of the terminologies being used in this work. Development of the international RefSet is also hampered by the issues identified in this thesis. A process to regularly inform the IHTSDO about issues identified by the Project Group has been developed and will commence shortly. At present (September 2012) the international general practice RefSet is under construction, and the map from concepts in the RefSet to ICPC-2 is 226

249 about to commence. Numerous delays have occurred in the project due to the time needed to map each source termset to SNOMED CT, the need to consult Project Group members in various countries, and delays in access to specialised software needed for the project. On completion of the RefSet and map development, both products will be field tested by GPs in multiple countries before their release. I anticipate that the international general practice RefSet will be ready for technical release by the end of The international general practice RefSet will not include content that is specific to Australian usage (e.g. Ross River fever). However, this RefSet could be used as a starting point for the development of an Australian SNOMED CT-AU general practice RefSet. If this was to occur, the international RefSet must be compared with the clinical terms used in Australia and extended to include those specific for Australia. The use of administrative terms differs in each country, so content will also have be added to account for Australian administrative terms. This approach minimises the work needed to create an Australian GP RefSet by using the international work as a baseline, and, once available, may represent an ideal opportunity to promote the introduction of SNOMED CT-AU in Australian general practice. 227

250 References in citation order 1. Royal Australian College of General Practitioners. What is general practice? Viewed 24 February 2011, 2. Australian Government Department of Health and Ageing. Medicare Statistics - March Quarter 2011, Group BC tables. Canberra: DoHA, Viewed 26 September 2011, 3. Australian Government Department of Human Services. Medicare Viewed 19 July 2012, e. 4. White KL. The ecology of medical care: origins and implications for population-based healthcare research. Health Serv Res 1997;32(1): Backer LA. The medical home: an idea whose time has come... again. Fam Pract Manag 2007;14(8): Royal Australian College of General Practitioners. Greater access to a 'medical home' key to improved health outcomes for Australia's elderly Viewed 8 August 2012, 7. Richards B, Bolton P, Veale B, Quinlan F. Information technology in general practice: a monograph commissioned by the General Practice Branch of the Commonwealth Department of Health and Aged Care. Publications Production Unit (Public Affairs, Parliamentary and Access Branch), Commonwealth Department of Aged Care. Monograph series No ACNeilsen. A study into levels of, and attitudes towards information technology in general practice. North Sydney: ACNeilsen, Britt H, Miller G, Charles J, Henderson J, Bayram C, Valenti L et al. General practice activity in Australia General Practice Series no. 29. Sydney: Sydney University Press,

251 10. Auricht C & Goulston R (eds) Medical Record Systems in Primary Health Care in Australia. Sydney: Royal Australian College of General Practitioners, Weed LL. Medical records, medical education and patient care. Chicago: Year Book Medical Publishers, Inc., Thiedeman B, Miller GC, Britt H, Rowed D. General Practice Data Model and Core Data Set Project: final project report Viewed 9 October 2012, del/final_project_report.pdf. 13. Gibbons P, Arzt N, Surke-Beebe S, Chute C, Dickinson G, Flewelling T et al. Coming to Terms: Scoping Interoperability for Health Care. Health Level Seven EHR Interoperability Work Group, Viewed 10 August 2012, February-2007.pdf. 14. Stroetmann VN, Kalra D, Lewalle P, Rector A, Rodrigues JM, Stroetmann KA et al. Semantic Interoperability for Better Health and Safer Healthcare: deployment and research roadmap for Europe Viewed 23 August 2012, ns/2009/2009semantic-health-report.pdf. 15. Bentzen N. Wonca Dictionary of general/family practice. Copenhagen: Manedsskrift for Praktisk Laegegerning, Standards Australia. The language of health concept representation (AS ). Sydney: Standards Australia, Susser M & Adelstein A (eds). Vital statistics: a memorial volume of selections from the reports and writings of William Farr, with an introduction by Mervyn Susser and Abraham Adelstein. Metuchen, New Jersey: The Scarecrow Press, Inc., Graunt J. Natural and political observations mentioned in a following index, and made upon the Bills of Mortality. London: The Roycroft, for John Martin, James Allestry, The Dicas, at the Sign of the Bell in St Pauls Church-yard, delacy M. Nosology, mortality, and disease theory in the eighteenth century. J Hist Med Allied Sci 1999;54(2): Sauvages F. Nosologica Methodica Sistens Morborum Classes. Amsterdam: Fratum de Tournes,

252 21. Egdahl A. Linnaeus' "Genera Morborum," and Some of His Other Medical Works. Med Library Hist J 1907;5(3): Latham RG. The works of Thomas Sydenham, M.D. London: The Sydenham Society, American Public Health Association. The Bertillon Classification of Causes of Death Michigan, Lansing. 24. World Health Organization. International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Volume 2. Geneva: World Health Organization, Israel RA. The history of the International Classification of Diseases. Health Bulletin 1991;49(1): The Research Committee of the Council of the College of General Practitioners. A Classification of Disease. J Coll Gen Pract Res Newsl 1959;2(2): Eimerl TS. The E Book system for record-keeping in general practice. Med Care 1973;11(2):Suppl Bentsen BG. Classifying of health problems in primary medical care. J R Coll Gen Pract Occas Pap 1976;1: Westbury RC & Tarrant M. Classification of disease in general practice: a comparative study. Can Med Assoc J 1969;101(10): Westbury RC. A plan to develop an international classification of diseases in family medicine. Melbourne: The Organising Committee of the Fifth World Conference on General Practice, Royal Australian College of General Practitioners; 1972; World Organization of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians. Fifth world conference on general practice proceedings. Melbourne: Royal Australian College of General Practitioners, Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICHPPC-2-defined (International Classification of Health Problems in Primary Care). Oxford Medical Publications. 3rd ed. Oxford: Oxford University Press,

253 33. World Organisation of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians. International Classification of Health Problems in Primary Care (ICHPPC). Chicago: American Hospital Association, Froom J. The International Classification of Health Problems for Primary Care. Med Care 1976;14(5): Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICHPPC-2 (International Classification of Health Problems in Primary Care) (1979 revision). 2nd ed. Oxford: Oxford University Press, Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICHPPC-2-defined (International Classification of Health Problems in Primary Care). Oxford Medical Publications. 3rd ed. Oxford: Oxford University Press, Lamberts H, Meads S, Wood M. Classification of reasons why persons seek primary care: pilot study of a new system. Public Health Rep 1984;99: Classification Committee of the World Organization of Family Doctors (WICC).Lamberts H & Wood M (eds) International Classification of Primary Care (ICPC). Oxford: Oxford University Press, Lamberts H, Meads S, Wood M. Results of the international field trial with the Reason for Encounter Classification. Soz Praventivmed 1985;30: Classification Committee of the World Organization of Family Doctors (WICC). International Classification of Primary Care, Second Edition (ICPC-2). Oxford: Oxford University Press, Thomson J. The works of William Cullen, M.D., Volume One. Edinburgh: William Blackwood, Edinburgh and T & G Underwood, London, McKusick VA. On lumpers and splitters, or the nosology of genetic disease. Perspect Biol Med 1969;12(2): International Health Terminology Standards Development Organisation. SNOMED CT value proposition Viewed 14 July 2012, 231

254 44. College of American Pathologists. SNOMED historical perspectives Viewed 14 July 2012, ide=%2fportlets%2fcontentviewer%2fshow&_windowlabel=cntvwrpt lt&cntvwrptlt%7bactionform.contentreference%7d=snomed%2fhistp ersp.html&_state=maximized&_pagelabel=cntvwr. 45. Benson T. Principles of health interoperability: HL7 and SNOMED. Health Informatics Series. London: Springer, Benson T. The history of the Read Codes: the inaugural James Read Memorial Lecture Inform Prim Care 2011;19(3): Booth N. What are the Read Codes? Health Libr Rev 1994;11(3): NHS Information Authority. The Clinical Terms Version 3 (The Read Codes) Viewed 5 October 2012, docs/chap1.pdf. 49. Cimino JJ. Desiderata for controlled medical vocabularies in the twenty-first century. Methods Inf Med 1998;37(4-5): Rector AL. Clinical terminology: why is it so hard? Methods Inf Med 1999;38(4-5): Stearns MQ, Price C, Spackman KA, Wang AY. SNOMED Clinical Terms: Overview of the development process and project status. JAMIA Symposium Supplement 2001: National E-Health Transition Authority. NEHTA welcomes new standards development organisation Viewed 19 July 2012, National E-Health Transition Authority. NEHTA secures international health terminology for Australia Viewed 19 July 2012, College of American Pathologists. IHTSDO and SNOMED CT support Viewed 9 January 2012, 232

255 55. International Health Terminology Standards Development Organisation. Who are the IHTSDO Charter Members and what does this mean? Viewed 20 July 2012, International Health Terminology Standards Development Organisation. Members of IHTSDO Viewed 20 July 2012, International Health Terminology Standards Development Organisation. Governance and Advisory Viewed 20 July 2012, International Health Terminology Standards Development Organisation. Special Interest Groups Viewed 20 July 2012, The Royal Australian College of General Practitioners. Computers as Medical Tools: Conference report from the RACGP 5th Computer Conference Britt H, Beaton N, Miller G. General practice medical records: why code? Why classify? Aust Fam Physician 1995;24(4):612, 614-2, Miller GC, Britt H, Jurd E. The Aus-Read Trial. N Z Fam Physician 1994;21(4): Bridges-Webb C, Britt H, Miles DA, Neary S, Charles J, Traynor V. Morbidity and treatment in general practice in Australia Med J Aust 1992;157(19 Oct Spec Sup):S1-S Britt H. A new coding tool for computerised clinical systems in primary care - ICPC plus. Aust Fam Physician 1997;26(Suppl 2): Family Medicine Research Centre. Current uses of ICPC-2 PLUS Viewed 27 July 2010, Oon YK. Docle coding and classification system browser Viewed 20 July 2012, Letrilliart L, Bacis A-K, Mennerat F, Colin C. Interface terminologies: a case study on the International Classification of Primary Care. World Academy of Science, Engineering and Technology 2009;59:

256 67. General Practice Computing Group. Final report of the General Practice Coding Jury. Barton, ACT: General Practice Computing Group, Viewed 9 October 2012, ng_jury.pdf. 68. Docle Systems. Docle Systems Viewed 5 July 2010, Britt H & Miller G. A critical review of the Final Report of the General Practice Coding Jury Viewed 14 August 2012, w%20of%20jury%20report.pdf. 70. Scott P, Walker D, McIsaac P. The Australian General Practice Vocabulary and the ICD-10-AM index: parsing and classification issues Viewed 26 April 2012, 54.pdf. 71. Walker D. Stage 2 of the GP vocabulary project Viewed 26 April 2012, DH4 Pty Ltd. Development of a Business Care for National Clinical Terminologies: needs, requirements and options. Discussion Document for Consultation. DH4 Pty Ltd. Jan National E-Health Transition Authority. Delivering E-Health's Foundations Viewed 19 July 2012, National E-Health Transition Authority. New national entity to drive e- health Viewed 5 October 2012, National E-Health Transition Authority. Key foundation of electronic health communication in place Viewed 28 September 2007, =136&Itemid= National E-Health Transition Authority. SNOMED CT-AU Viewed 10 May 2011, 234

257 77. Stanek J & Simon B. SNOMED-CT Coding in General Practice - Barriers and Opportunities. Brunswick East, Vic: Health Informatics Society of Australia; 2008; Family Medicine Research Centre. Theses and treaties - completed Viewed 29 May 2012, O'Halloran J, Miller GC, Britt H. Defining chronic conditions for primary care with ICPC-2. Fam Pract 2004;21(4): Miller G, O'Halloran J, Bayram C. General practice EHR and data query minimum data set development Viewed 9 October 2012, ects/gpcg_sp1_%20gpracehrmindatasetfinalreport.pdf. 81. Wang Y, Patrick J, Miller G, O'Hallaran J. A computational linguistics motivated mapping of ICPC-2 PLUS to SNOMED CT. BMC Med Inform Decis Mak 2008;8 Suppl 1:S Britt H, Sayer GP, Miller GC, Charles J, Scahill S, Horn F et al. General practice activity in Australia General Practice Series no. 2. AIHW Cat. no. GEP 2. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Charles J, Knox S, Sayer GP, Valenti L et al. General practice activity in Australia General Practice Series No. 5. AIHW Cat. no. GEP 5. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General Practice Series No. 8. AIHW Cat. no. GEP 8. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 10. AIHW Cat. no. GEP 10. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 14. AIHW Cat. no. GEP 14. Canberra: Australian Institute of Health and Welfare,

258 87. Britt H, Miller GC, Knox S, Charles J, Valenti L, Pan Y et al. General Practice Activity in Australia General Practice Series No. 16. AIHW Cat. no. GEP 16. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Knox S, Charles J, Pan Y, Henderson J et al. General practice activity in Australia General practice series no. 18. AIHW Cat. no. GEP 18. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Charles J, Pan Y, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 19. AIHW Cat. no. GEP 19. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Charles J, Henderson J, Bayram C, Harrison C et al. General practice activity in Australia General practice series no. 22. AIHW Cat. no. GEP 22. Canberra: Australian Institute of Health and Welfare, Britt H, Miller GC, Charles J, Bayram C, Pan Y, Henderson J et al. General practice activity in Australia General practice series no. 21. AIHW Cat. no. GEP 21. Canberra: Australian Institute of Health and Welfare, Britt H, Miller G, Charles J, Henderson J, Bayram C, Pan Y et al. General practice activity in Australia General Practice Series no. 25. AIHW Cat. no. GEP 25. Canberra: Australian Institute of Health and Welfare, Britt H, Miller G, Charles J, Henderson J, Bayram C, Pan Y et al. General practice activity in Australia General Practice Series no. 27. AIHW Cat. no. GEP 27. Canberra: Australian Institute of Health and Welfare, General Practice Registrars Australia. General Practice Registrars Australia: About us Viewed 15 June 2012, The Royal Australian College of General Practitioners. Standards for general practices: 4th edition Viewed 10 May 2011, Britt H, Miller GC, Charles J, Henderson J, Bayram C, Valenti L et al. A decade of Australian general practice activity to General Practice Series no. 30. Sydney: Sydney University Press,

259 97. Family Medicine Research Centre. ICPC-2 PLUS: the BEACH coding system Viewed 15 June 2012, Family Medicine Research Centre. ICPC-2 PLUS functionality requirements for developers Viewed 18 October 2010, International Health Terminology Standards Development Organisation. SNOMED CT Technical Reference Guide: January 2011 international release. Jan International Health Terminology Standards Development Organisation. SNOMED CT Technical Implementation Guide: January International Health Terminology Standards Development Organisation. Jan Spackman KA. Normal forms for description logic expressions of clinical concepts in SNOMED RT. Proc AMIA Symp 2001: International Health Terminology Standards Development Organisation. SNOMED CT Editorial Guide: July 2011 international release (GB English) Viewed 9 January 2012, alguide_current-en-gb_int_ pdf International Health Terminology Standards Development Organisation. Online Request Submission System Viewed 9 October 2012, National E-Health Transition Authority. NEHTA launches the first release of SNOMED CT-AU National E-Health Transition Authority. SNOMED CT-AU release note. Nov International Health Terminology Standards Development Organisation. SNOMED CT Release Format 2.0 Reference Set Specifications Viewed 6 July 2011, s/snomed_ct_release_format_2_-_refset_specifications_v1.0.pdf Hansen DP, Kemp ML, Mills SR, Mercer MA, Frosdick PA, Lawley MJ. Developing a national emergency department data reference set based on SNOMED CT. Med J Aust 2011;194(4):S

260 108. National E-Health Transition Authority. SNOMED CT-AU Release Note (Document ID NEHTA 666:2010) Viewed 10 May 2011, Henderson J, Britt H, Miller G. Extent and utilisation of computerisation in Australian general practice. Med J Aust 2006;185(2): National E-Health Transition Authority. Clinical Leaders Viewed 1 February 2011, National Centre for Classification in Health. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian modification (ICD-10-AM) - sixth edition. Sydney: National Centre for Classification in Health, The University of Sydney, Roberts RF, Innes KC, Walker SM. Introducing ICD-10-AM in Australian hospitals. Med J Aust 1998;169 Suppl:S32-S Medical Objects. General practitioners: secure messaging Viewed 28 April 2011, International Organization for Standardization (ISO). Information technology - Vocabulary - part 1: Fundamental terms Viewed 8 July 2011, Joss MW & Wszola S. The engines that can: text search and retrieval software, their strategies, and vendors. CD ROM Professional 1996;9(6): Bainbridge M. The Common User Interface (CUI) project Viewed 10 May 2011, Microsoft Corporation. What is Microsoft Health CUI? Viewed 9 September 2012, Khan A & Western M. Does attitude matter in computer use in Australian general practice? A zero-inflated Poisson regression analysis. HIM J 2011;40(2):

261 119. Australian Primary Care Collaboratives. Objective and aims of the Australian Primary Care Collaboratives (APCC) Program Viewed 23 June 2011, CC.pdf Improvement Foundation (Australia). APCC Program Topic Measures and Other Improvement Measures Viewed 23 June 2011, pdf Australian Government Department of Health and Ageing. Medicare Benefits Schedule Book. Canberra: DoHA, /Downloads Pearce C, Shearer M, Gardner K, Kelly J. A division's worth of data. Aust Fam Physician 2011;40(3): Field D, Ward AM, Lopez DG. Apart from Medicare, what other work do GPs do? Aust Fam Physician 2003;32(6): Rector AL, Qamar R, Marley T. Binding ontologies and coding systems to electronic health records and messages. Applied Ontology 2009;4(1): Fung KW, McDonald C, Srinivasan S. The UMLS-CORE project: a study of the problem list terminologies used in large healthcare institutions. J Am Med Inform Assoc 2010;17(6): Cimino JJ. In defense of the Desiderata. J Biomed Inform 2006;39(3): Elkin PL, Brown SH, Carter J, Bauer BA, Wahner-Roedler D, Bergstrom L et al. Guideline and quality indicators for development, purchase and use of controlled health vocabularies. Int J Med Inform 2002;68(1-3): International Health Terminology Standards Development Organisation. SNOMED CT User Guide: July 2011 international release Viewed 18 June 2012, Rector AL. The interface between information, terminology, and inference models. Stud Health Technol Inform 2001;84(Pt 1):

262 130. Patrick J, Wang Y, Budd P. Developing SNOMED CT subsets from clinical notes for intensive care service. Health Care and Informatics Review Online 2008;12(3): Rogers J. Terminologies: Getting clinical data in; getting useful data out Viewed 30 September 2011, Vikstrom A, Nystrom M, Ahlfeldt H, Strender LE, Nilsson GH. Views of diagnosis distribution in primary care in 2.5 million encounters in Stockholm: a comparison between ICD-10 and SNOMED CT. Inform Prim Care 2010;18(1): National E-Health Transition Authority. NEHTA data group specifications: problem/diagnosis Viewed 23 January 2012, Comité Européen de Normalisation. Health informatics: System of concepts to support continuity of care (Part 1: Basic concepts) (CEN ) ASTM International. Standard specification for the continuity of care record (CCR) HL7. Electronic health record - system functional model, release 1. Feb Britt H, Miller GC, Charles J, Henderson J, Bayram C, Valenti L et al. General practice activity in Australia to : 10 year data tables. General Practice Series No. 28. AIHW Cat. no. GEP 28. Canberra: Australian Institute of Health and Welfare, Okkes IM & Lamberts H. Classification and the domain of family practice. In: Jones R (ed). The Oxford Textbook of Primary Medical Care (Volume 1). Oxford: Oxford University Press, 2003; Lauteslager M, Brouwer HJ, Mohrs J, Bindels PJ, Grundmeijer HG. The patient as a source to improve the medical record. Fam Pract 2002;19(2): Van Vleck TT, Wilcox A, Stetson PD, Johnson SB, Elhadad N. Content and structure of clinical problem lists: a corpus analysis. AMIA Annu Symp Proc 2008: AAP. Human error caused needless death. Perth Now. May 12,

263 142. Royal Australian College of General Practitioners. General practice fact sheet: Medical history Viewed 10 December 2011, ealth/gpehealthsummary/fact-sheet-medical-history.pdf McDonald K & James S. NEHTA outlines PCEHR roadmap for coming months. Pulse+IT magazine Viewed 24 July 2012, ew=article&id=1079:nehta-outlines-pcehr-roadmap-for-comingmonths&catid=16:australian-ehealth&itemid= National E-Health Transition Authority. Clinical information systems and ehealth records Viewed 25 June 2012, National E-Health Transition Authority. What is a PCEHR? Viewed 25 June 2012, Australian Government Department of Health and Ageing. Concept of operations: Relating to the introduction of a Personally Controlled Electronic Health Record System (Section Event summaries) Viewed 23 August 2012, t/pcehrs-intro-toc~ch4~4_3~4_3_ Australian Government Department of Health and Ageing. Concept of operations: Relating to the introduction of a Personally Controlled Electronic Health Record System (Section Shared Health Summaries) Viewed 23 August 2012, t/pcehrs-intro-toc~ch4~4_3~4_3_ National E-Health Transition Authority. Event summary: structured content specification (Version 1.1) [registration required] Viewed 23 August 2012, egoryid=308# National E-Health Transition Authority. Shared health summary: structured content specification (Version 1.1) [registration required] Viewed 23 August 2012, egoryid=312#0. 241

264 150. The Royal Australian College of General Practitioners. Guidelines for preventive activities in general practice (The Red Book) Viewed 15 December 2011, Harris MF, Bailey L, Snowdon T, Litt J, Smith JW, Joyner B et al. Developing the guidelines for preventive care - two decades of experience. Aust Fam Physician 2010;39(1-2): Australian Government Department of Health and Ageing. The November 2010 Medicare Benefits Schedule. Canberra: DoHA, Viewed 5 October 2012, /Downloads Classification Committee of WONCA (World Organization of National Colleges,Academies,and Academic Associations of General Practitioners/Family Physicians). ICPC-2: International Classification for Primary Care. 2nd ed. Oxford: Oxford Medical Publications, Oxford University Press, European Community for Standardization (CEN). CEN/TC Standards under development Viewed 17 December 2011, TechnicalCommittees/Pages/WP.aspx?param=6232&title=CEN%2FTC Baorto D, Li L, Cimino JJ. Practical experience with the maintenance and auditing of a large medical ontology. J Biomed Inform 2009;42(3): Cimino JJ, Clayton PD, Hripcsak G, Johnson SB. Knowledge-based approaches to the maintenance of a large controlled medical terminology. J Am Med Inform Assoc 1994;1(1): Robinson D, Schulz E, Brown P, Price C. Updating the Read Codes: user-interactive maintenance of a dynamic clinical vocabulary. J Am Med Inform Assoc 1997;4(6): Spackman KA. Rates of change in a large clinical terminology: three years experience with SNOMED Clinical Terms. AMIA Annu Symp Proc 2005: Oliver DE, Shahar Y, Shortliffe EH, Musen MA. Representation of change in controlled medical terminologies. Artif Intell Med 1999;15(1):

265 160. Raiez F, Arts D, Cornet R. Terminological system maintenance: a procedures framework and an exploration of current practice. Stud Health Technol Inform 2005;116: Bakhshi-Raiez F, Cornet R, de Keizer NF. Development and application of a framework for maintenance of medical terminological systems. J Am Med Inform Assoc 2008;15(5): Lee D, Cornet R, Lau F. Implications of SNOMED CT versioning. Int J Med Inform 2011;80(6): Lai C, Przychodzin D, Bunch C, Morrison T, Chaney R. The resource impact of supporting SNOMED CT updates in a clinical decision support system Wade G & Rosenbloom ST. The impact of SNOMED CT revisions on a mapped interface terminology: terminology development and implementation issues. J Biomed Inform 2009;42(3): Nachimuthu SK & Lau LM. Practical issues in using SNOMED CT as a reference terminology. Stud Health Technol Inform 2007;129(Pt 1): Family Medicine Research Centre. Past commissioned research Viewed 7 March 2012, Miller GC & Britt H. The development of a code set for pathology orders by general practitioners. The University of Sydney Britt H, Miller GC, Scahill S, Sayer G. The development of a code set for imaging orders by general practitioners and a study into diagnostic imaging requests by general practitioners. The University of Sydney Jones R, Barraclough K, Dowrick C. When no diagnostic label is applied. BMJ 2010;340:c Howie JG. Diagnosis--the Achilles heel? J R Coll Gen Pract 1972;22(118): Heneghan C, Glasziou P, Thompson M, Rose P, Balla J, Lasserson D et al. Diagnostic strategies used in primary care. BMJ 2009;338:b Baerheim A. The diagnostic process in general practice: has it a twophase structure? Fam Pract 2001;18(3):

266 173. World Health Organization. International Classification of Primary Care, Second edition (ICPC-2) Viewed 17 December 2008, Andre M, Anden A, Borgquist L, Rudebeck CE. GPs' decision-making - perceiving the patient as a person or a disease running title GPs' decision-making. BMC Fam Pract 2012;13(1): Thomas KB. Temporarily dependent patient in general practice. Br Med J 1974;1(5908): Oxford Dictionaries. Definition of fit Viewed 4 August 2012, NSW Health. SNOMED CT within the NSW Public Health System Viewed 4 August 2012, Chicco J & Pegler G. SNOMED CT pioneers: the NSW experience Viewed 4 August 2012, 0&%20Pegler/Chicco%20&%20Pegler.swf Patrick J. A Study of a Health Enterprise Information System. School of Information Technologies, University of Sydney Ryan A, Patrick J, Herkes R. Introduction of Enhancement Technologies into the Intensive Care Service, Royal Prince Alfred Hospital, Sydney. Health Information Management Journal 2008;37(1): Giannangelo K & Fenton SH. SNOMED CT survey: an assessment of implementation in EMR/EHR applications. Perspect Health Inf Manag 2008;5: Elhanan G, Perl Y, Geller J. A Survey of Direct Users and Uses of SNOMED CT: 2010 Status. AMIA Annu Symp Proc 2010;2010: Dolin RH, Mattison JE, Cohn S, Campbell KE, Wiesenthal AM, Hochhalter B et al. Kaiser Permanente's Convergent Medical Terminology. Medinfo 2004;11(Pt 1): Mantena S & Schadow G. Evaluation of the VA/KP problem list subset of SNOMED as a clinical terminology for electronic prescription clinical decision support. AMIA Annu Symp Proc 2007:

267 185. St.Sauver J, Buntrock J, Rademacher D, Albrecht D, Gregg M, Ihrke D et al. Comparison of Mayo Clinical Coding Systems Viewed 4 August 2012, mayoresearch.mayo.edu/mayo/research/biostat/upload/83.pdf Campbell JR, Lopez Osornio A, de Quiros F, Luna D, Reynoso G. Semantic Interoperability and SNOMED CT: A Case Study in Clinical Problem Lists. Medinfo Osornio AL, Luna D, Gambarte ML, Gomez A, Reynoso G, de Quiros FG. Creation of a local interface terminology to SNOMED CT. Stud Health Technol Inform 2007;129(Pt 1): Lau F, Simkus R, Lee D. A methodology for encoding problem lists with SNOMED CT in general practice. Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008): May 31-June 2, 2008, Arizona, USA Epub 2008 May Commonwealth Scientific and Industrial Research Organisation (CSIRO). Snapper Viewed 17 February 2012, Commonwealth Scientific and Industrial Research Organisation (CSIRO). CSIRO software to fast-track adoption of ehealth Viewed 18 April 2012, Dorland's Illustrated Medical Dictionary, 32nd edition. Philadelphia: Elsevier Saunders, The Macquarie Concise Dictionary (revised 3rd edition). Macquarie University: The Macquarie Library Pty Ltd, National Heart Foundation of Australia (National blood pressure and vascular disease advisory committee). Guide to management of hypertension 2008: assessing and managing raised blood pressure in adults (Updated December 2010) Viewed 5 October 2012, onguidelines2008to2010update.pdf Gastroenterological Society of Australia. Reflux disease: Gastrooesophageal Reflux Disease in adults (clinical update) Viewed 16 March 2012, 245

268 195. McDonnell L & Usherwood T. International medical graduates - challenges faced in the Australian training program. Aust Fam Physician 2008;37(6): Sommer J, Macdonald W, Bulsara C, Lim D. Grunt language versus accent: the perceived communication barriers between international medical graduates and patients in Central Wheatbelt catchments. Aust J Primary Health Wang AY, Barrett JW, Bentley T, Markwell D, Price C, Spackman KA et al. Mapping between SNOMED RT and Clinical terms version 3: a key component of the SNOMED CT development process. Proc AMIA Symp 2001: Richesson RL, Andrews JE, Krischer JP. Use of SNOMED CT to represent clinical research data: a semantic characterization of data items on case report forms in vasculitis research. J Am Med Inform Assoc 2006;13(5): Elkin PL, Brown SH, Husser CS, Bauer BA, Wahner-Roedler D, Rosenbloom ST et al. Evaluation of the content coverage of SNOMED CT: ability of SNOMED clinical terms to represent clinical problem lists. Mayo Clin Proc 2006;81(6): Brown SH, Elkin PL, Bauer BA, Wahner-Roedler D, Husser CS, Temesgen Z et al. SNOMED CT: utility for a general medical evaluation template. AMIA Annu Symp Proc 2006: Bakhshi-Raiez F, Ahmadian L, Cornet R, de JE, de Keizer NF. Construction of an interface terminology on SNOMED CT. Generic approach and its application in intensive care. Methods Inf Med 2010;49(4): Penz JF, Brown SH, Carter JS, Elkin PL, Nguyen VN, Sims SA et al. Evaluation of SNOMED coverage of Veterans Health Administration terms. Medinfo 2004;11(Pt 1): Rosenbloom ST, Brown SH, Froehling D, Bauer BA, Wahner-Roedler DL, Gregg WM et al. Using SNOMED CT to represent two interface terminologies. J Am Med Inform Assoc 2009;16(1): Wade G & Rosenbloom ST. Experiences mapping a legacy interface terminology to SNOMED CT. BMC Med Inform Decis Mak 2008;8 Suppl 1:S3. 246

269 205. Nystrom M, Vikstrom A, Nilsson GH, Ahlfeldt H, Orman H. Enriching a primary health care version of ICD-10 using SNOMED CT mapping. J Biomed Semantics 2010;1(1): Nadkarni PM & Darer JA. Migrating existing clinical content from ICD-9 to SNOMED. J Am Med Inform Assoc 2010;17(5): Matney SA, Warren JJ, Evans JL, Kim TY, Coenen A, Auld VA. Development of the nursing problem list subset of SNOMED CT(R). J Biomed Inform Westra BL, Bauman R, Delaney CW, Lundberg CB, Petersen C. Validation of concept mapping between PNDS and SNOMED CT. AORN J 2008;87(6): Lee DH, Lau FY, Quan H. A method for encoding clinical datasets with SNOMED CT. BMC Med Inform Decis Mak 2010;10: Sampalli T, Shepherd M, Duffy J, Fox R. An evaluation of SNOMED CT in the domain of complex chronic conditions. Int J Integr Care 2010;10:e Montella D, Brown SH, Elkin PL, Jackson JC, Rosenbloom ST, Wahner-Roedler D et al. Comparison of SNOMED CT versus Medcin terminology concept coverage for mild Traumatic Brain Injury. AMIA Annu Symp Proc 2011;2011: Brown SH, Husser CS, Wahner-Roedler D, Bailey S, Nugent L, Porter K et al. Using SNOMED CT as a reference terminology to cross map two highly pre-coordinated classification systems. Stud Health Technol Inform 2007;129(Pt 1): Rector A & Iannone L. Lexically suggest, logically define: quality assurance of the use of qualifiers and expected results of postcoordination in SNOMED CT. J Biomed Inform 2012;45(2): U.S.National Library of Medicine. UMLS Quick Start Guide Viewed 30 August 2012, International Health Terminology Standards Development Organisation. IHTSDO Annual Quality Report Viewed 27 August 2012, lity_assurance/ihtsdo_annual_quality_report_2011_ _v1_ 00.pdf. 247

270 216. National E-Health Transition Authority. Mapping to SNOMED CT-AU: guidelines and requirements (Version 1.0) [registration required] Viewed 16 August 2012, Rosenbloom ST, Miller RA, Johnson KB, Elkin PL, Brown SH. Interface terminologies: facilitating direct entry of clinical data into electronic health record systems. J Am Med Inform Assoc 2006;13(3): International Health Terminology Standards Development Organisation. About SNOMED CT Viewed 16 August 2012, Spackman KA & Reynoso G. Examining SNOMED from the Perspective of Formal Ontological Principles: Some Preliminary Analysis and Observations. 2004; Moynihan R. Who benefits from treating prehypertension? BMJ 2010;341:c Moynihan R, Doust J, Henry D. Preventing overdiagnosis: how to stop harming the healthy. BMJ 2012;344:e Baud RH, Ceusters W, Ruch P, Rassinoux AM, Lovis C, Geissbuhler A. Reconciliation of ontology and terminology to cope with linguistics. Stud Health Technol Inform 2007;129(Pt 1): Moerkerke C & Ceusters W. The myth of preferred terms in medical sublanguage and its impact on natural language understanding applications: an empirical study. 2000; Schulz S, Suntisrivaraporn B, Baader F. SNOMED CT's problem list: ontologists' and logicians' therapy suggestions. Stud Health Technol Inform 2007;129(Pt 1): Schulz S, Suntisrivaraporn B, Baader F, Boeker M. SNOMED reaching its adolescence: ontologists' and logicians' health check. Int J Med Inform 2009;78 Suppl 1:S86-S Bodenreider O, Smith B, Kumar A, Burgun A. Investigating subsumption in SNOMED CT: An exploration into large description logic-based biomedical terminologies. Artif Intell Med Rector AL & Brandt S. Why do it the hard way? The case for an expressive description logic for SNOMED. J Am Med Inform Assoc 2008;15(6):

271 228. Rector AL, Brandt S, Schneider T. Getting the foot out of the pelvis: modeling problems affecting use of SNOMED CT hierarchies in practical applications. J Am Med Inform Assoc 2011;18(4): Rector A, Iannone L, Stevens R. Quality assurance of the content of a large DL-based terminology using mixed lexical and semantic criteria: experience with SNOMED CT. Presented at the K-CAP'11 Proceedings of the sixth international conference on Knowledge capture; 25 Jun 2011; Banff, Alberta: Conley E & Benson T. SNOMED CT: who needs to know what? European Journal for Biomedical Informatics 2011;7(2): Bakhshi-Raiez F, de Keizer NF, Cornet R, Dorrepaal M, Dongelmans D, Jaspers MW. A usability evaluation of a SNOMED CT based compositional interface terminology for intensive care. Int J Med Inform 2012;81(5): National E-Health Transition Authority. NCTIS Terminology Product Releases: Release Cycles and Versioning Viewed 17 August 2012, nctis-release-cycle-and-versioning Hennen BK. Continuity of care in family practice. Part 1: dimensions of continuity. J Fam Pract 1975;2(5): Australian Commission on Safety and Quality in Healthcare. OSSIE Guide to Clinical Handover Improvement Viewed 29 September 2012, Markwell D, Sato L, Cheetham E. Representing clinical information using SNOMED Clinical Terms with different structural information models. 2008; Standards Australia. What is a standard? Viewed 22 September 2012, ard/pages/default.aspx IBM Consulting Group Health Practice. Clinical & Administrative General Practice Computer Systems Consultancy: Functional Requirements Specification Report Viewed 22 September 2012,

272 238. De Lusignan S. The barriers to clinical coding in general practice: a literature review. Med Inform Internet Med 2005;30(2): International Health Terminology Standards Development Organisation. Mapping SNOMED CT to ICD-10 Project [registration required] Viewed 27 September 2012, 10_proj/ International Health Terminology Standards Development Organisation & Intern. Family Practice/General Practice RefSet and ICPC mapping project [registration required] Viewed 27 September 2012, tice_2/ International Organization for Standardization (ISO) & International. ISO/DTR 12300: Health informatics - Principles of mapping between terminological systems Viewed 27 August 2012, snumber= Australian Government Department of Health and Ageing. Personally Controlled Electronic Health Records for all Australians Viewed 27 August 2012, hmedia09.htm Australian Government Department of Health and Ageing. Welcome to ehealth.gov.au Viewed 24 August 2012, McDonald K. September the month for PCEHR-enabled GP software. Pulse+IT magazine Viewed 24 August 2012, ew=article&id=1017:september-the-month-for-pcehr-enabled-gpsoftware&catid=16:australian-ehealth&itemid= National E-Health Transition Authority. PCEHR Specification & Standards Development: Upcoming collaboration opportunities on the development of the Personally Controlled Electronic Health Record (PCEHR) specifications Viewed 27 September 2012, 7_OCT_1700.wmv. 250

273 246. International Health Terminology Standards Development Organisation. Wonca and IHTSDO seek to strengthen informatics in primary care Viewed 27 September 2012, Wonca_IHTSDO_agreement_announcement.pdf. 251

274 References in alphabetical order (citation order in parentheses) AAP. Human error caused needless death. Perth Now. May 12, (141) ACNeilsen. A study into levels of, and attitudes towards information technology in general practice. North Sydney: ACNeilsen, (8) American Public Health Association. The Bertillon Classification of Causes of Death Michigan, Lansing. (23) Andre M, Anden A, Borgquist L, Rudebeck CE. GPs' decision-making - perceiving the patient as a person or a disease running title GPs' decisionmaking. BMC Fam Pract 2012;13(1):38. (174) ASTM International. Standard specification for the continuity of care record (CCR) (135) Auricht C & Goulston R (eds) Medical Record Systems in Primary Health Care in Australia. Sydney: Royal Australian College of General Practitioners, (10) Australian Commission on Safety and Quality in Healthcare. OSSIE Guide to Clinical Handover Improvement Viewed 29 September 2012, (234) Australian Government Department of Health and Ageing. Concept of operations: Relating to the introduction of a Personally Controlled Electronic Health Record System (Section Event summaries) Viewed 23 August 2012, HRS-Intro-toc~ch4~4_3~4_3_2. (146) Australian Government Department of Health and Ageing. Concept of operations: Relating to the introduction of a Personally Controlled Electronic Health Record System (Section Shared Health Summaries) Viewed 23 August 2012, HRS-Intro-toc~ch4~4_3~4_3_1. (147) 252

275 Australian Government Department of Health and Ageing. Medicare Benefits Schedule Book. Canberra: DoHA, (121) Australian Government Department of Health and Ageing. Medicare Statistics - March Quarter 2011, Group BC tables. Canberra: DoHA, Viewed 26 September 2011, (2) Australian Government Department of Health and Ageing. Personally Controlled Electronic Health Records for all Australians Viewed 27 August 2012, hmedia09.htm. (242) Australian Government Department of Health and Ageing. The November 2010 Medicare Benefits Schedule. Canberra: DoHA, Viewed 5 October 2012, loads (152) Australian Government Department of Health and Ageing. Welcome to ehealth.gov.au Viewed 24 August 2012, (243) Australian Government Department of Human Services. Medicare Viewed 19 July 2012, (3) Australian Primary Care Collaboratives. Objective and aims of the Australian Primary Care Collaboratives (APCC) Program Viewed 23 June 2011, df. (119) Backer LA. The medical home: an idea whose time has come... again. Fam Pract Manag 2007;14(8): (5) Baerheim A. The diagnostic process in general practice: has it a two-phase structure? Fam Pract 2001;18(3): (172) Bainbridge M. The Common User Interface (CUI) project Viewed 10 May 2011, (116) 253

276 Bakhshi-Raiez F, Ahmadian L, Cornet R, de JE, de Keizer NF. Construction of an interface terminology on SNOMED CT. Generic approach and its application in intensive care. Methods Inf Med 2010;49(4): (201) Bakhshi-Raiez F, Cornet R, de Keizer NF. Development and application of a framework for maintenance of medical terminological systems. J Am Med Inform Assoc 2008;15(5): (161) Bakhshi-Raiez F, de Keizer NF, Cornet R, Dorrepaal M, Dongelmans D, Jaspers MW. A usability evaluation of a SNOMED CT based compositional interface terminology for intensive care. Int J Med Inform 2012;81(5): (231) Baorto D, Li L, Cimino JJ. Practical experience with the maintenance and auditing of a large medical ontology. J Biomed Inform 2009;42(3): (155) Baud RH, Ceusters W, Ruch P, Rassinoux AM, Lovis C, Geissbuhler A. Reconciliation of ontology and terminology to cope with linguistics. Stud Health Technol Inform 2007;129(Pt 1): (222) Benson T. Principles of health interoperability: HL7 and SNOMED. Health Informatics Series. London: Springer, (45) Benson T. The history of the Read Codes: the inaugural James Read Memorial Lecture Inform Prim Care 2011;19(3): (46) Bentsen BG. Classifying of health problems in primary medical care. J R Coll Gen Pract Occas Pap 1976;1:1-5. (28) Bentzen N. Wonca Dictionary of general/family practice. Copenhagen: Manedsskrift for Praktisk Laegegerning, (15) Bodenreider O, Smith B, Kumar A, Burgun A. Investigating subsumption in SNOMED CT: An exploration into large description logic-based biomedical terminologies. Artif Intell Med (226) Booth N. What are the Read Codes? Health Libr Rev 1994;11(3): (47) Bridges-Webb C, Britt H, Miles DA, Neary S, Charles J, Traynor V. Morbidity and treatment in general practice in Australia Med J Aust 1992;157(19 Oct Spec Sup):S1-S56. (62) Britt H & Miller G. A critical review of the Final Report of the General Practice 254

277 Coding Jury Viewed 14 August 2012, of%20jury%20report.pdf. (69) Britt H, Beaton N, Miller G. General practice medical records: why code? Why classify? Aust Fam Physician 1995;24(4):612, 614-2, 615. (60) Britt H, Miller G, Charles J, Henderson J, Bayram C, Pan Y et al. General practice activity in Australia General Practice Series no. 25. AIHW Cat. no. GEP 25. Canberra: Australian Institute of Health and Welfare, (92) Britt H, Miller G, Charles J, Henderson J, Bayram C, Pan Y et al. General practice activity in Australia General Practice Series no. 27. AIHW Cat. no. GEP 27. Canberra: Australian Institute of Health and Welfare, (93) Britt H, Miller G, Charles J, Henderson J, Bayram C, Valenti L et al. General practice activity in Australia General Practice Series no. 29. Sydney: Sydney University Press, (9) Britt H, Miller GC, Charles J, Bayram C, Pan Y, Henderson J et al. General practice activity in Australia General practice series no. 21. AIHW Cat. no. GEP 21. Canberra: Australian Institute of Health and Welfare, (91) Britt H, Miller GC, Charles J, Henderson J, Bayram C, Harrison C et al. General practice activity in Australia General practice series no. 22. AIHW Cat. no. GEP 22. Canberra: Australian Institute of Health and Welfare, (90) Britt H, Miller GC, Charles J, Henderson J, Bayram C, Valenti L et al. A decade of Australian general practice activity to General Practice Series no. 30. Sydney: Sydney University Press, (96) Britt H, Miller GC, Charles J, Henderson J, Bayram C, Valenti L et al. General practice activity in Australia to : 10 year data tables. General Practice Series No. 28. AIHW Cat. no. GEP 28. Canberra: Australian Institute of Health and Welfare, (137) Britt H, Miller GC, Charles J, Knox S, Sayer GP, Valenti L et al. General practice activity in Australia General Practice Series No. 5. AIHW Cat. no. GEP 5. Canberra: Australian Institute of Health and Welfare, (83) 255

278 Britt H, Miller GC, Charles J, Pan Y, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 19. AIHW Cat. no. GEP 19. Canberra: Australian Institute of Health and Welfare, (89) Britt H, Miller GC, Knox S, Charles J, Pan Y, Henderson J et al. General practice activity in Australia General practice series no. 18. AIHW Cat. no. GEP 18. Canberra: Australian Institute of Health and Welfare, (88) Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General Practice Series No. 8. AIHW Cat. no. GEP 8. Canberra: Australian Institute of Health and Welfare, (84) Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 10. AIHW Cat. no. GEP 10. Canberra: Australian Institute of Health and Welfare, (85) Britt H, Miller GC, Knox S, Charles J, Valenti L, Henderson J et al. General practice activity in Australia General practice series no. 14. AIHW Cat. no. GEP 14. Canberra: Australian Institute of Health and Welfare, (86) Britt H, Miller GC, Knox S, Charles J, Valenti L, Pan Y et al. General Practice Activity in Australia General Practice Series No. 16. AIHW Cat. no. GEP 16. Canberra: Australian Institute of Health and Welfare, (87) Britt H, Miller GC, Scahill S, Sayer G. The development of a code set for imaging orders by general practitioners and a study into diagnostic imaging requests by general practitioners. The University of Sydney (168) Britt H, Sayer GP, Miller GC, Charles J, Scahill S, Horn F et al. General practice activity in Australia General Practice Series no. 2. AIHW Cat. no. GEP 2. Canberra: Australian Institute of Health and Welfare, (82) Britt H. A new coding tool for computerised clinical systems in primary care - ICPC plus. Aust Fam Physician 1997;26(Suppl 2): (63) 256

279 Brown SH, Elkin PL, Bauer BA, Wahner-Roedler D, Husser CS, Temesgen Z et al. SNOMED CT: utility for a general medical evaluation template. AMIA Annu Symp Proc 2006: (200) Brown SH, Husser CS, Wahner-Roedler D, Bailey S, Nugent L, Porter K et al. Using SNOMED CT as a reference terminology to cross map two highly pre-coordinated classification systems. Stud Health Technol Inform 2007;129(Pt 1): (212) Campbell JR, Lopez Osornio A, de Quiros F, Luna D, Reynoso G. Semantic Interoperability and SNOMED CT: A Case Study in Clinical Problem Lists. Medinfo (186) Chicco J & Pegler G. SNOMED CT pioneers: the NSW experience Viewed 4 August 2012, 0Pegler/Chicco%20&%20Pegler.swf. (178) Cimino JJ, Clayton PD, Hripcsak G, Johnson SB. Knowledge-based approaches to the maintenance of a large controlled medical terminology. J Am Med Inform Assoc 1994;1(1): (156) Cimino JJ. Desiderata for controlled medical vocabularies in the twenty-first century. Methods Inf Med 1998;37(4-5): (49) Cimino JJ. In defense of the Desiderata. J Biomed Inform 2006;39(3): (126) Classification Committee of the World Organization of Family Doctors (WICC).Lamberts H & Wood M (eds) International Classification of Primary Care (ICPC). Oxford: Oxford University Press, (38) Classification Committee of the World Organization of Family Doctors (WICC). International Classification of Primary Care, Second Edition (ICPC- 2). Oxford: Oxford University Press, (40) Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICHPPC-2-defined (International Classification of Health Problems in Primary Care). Oxford Medical Publications. 3rd ed. Oxford: Oxford University Press, (32) 257

280 Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICHPPC-2 (International Classification of Health Problems in Primary Care) (1979 revision). 2nd ed. Oxford: Oxford University Press, (35) Classification Committee of WONCA (World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians). ICPC-2: International Classification of Primary Care. 2nd ed. Oxford: Oxford Medical Publications, Oxford University Press, (153) College of American Pathologists. IHTSDO and SNOMED CT support Viewed 9 January 2012, (54) College of American Pathologists. SNOMED historical perspectives Viewed 14 July 2012, %2Fportlets%2FcontentViewer%2Fshow&_windowLabel=cntvwrPtlt&cntvwr Ptlt%7BactionForm.contentReference%7D=snomed%2FhistPersp.html&_sta te=maximized&_pagelabel=cntvwr. (44) Comité Européen de Normalisation. Health informatics: System of concepts to support continuity of care (Part 1: Basic concepts) (CEN ) (134) Commonwealth Scientific and Industrial Research Organisation (CSIRO). Snapper Viewed 17 February 2012, (189) Commonwealth Scientific and Industrial Research Organisation (CSIRO). CSIRO software to fast-track adoption of ehealth Viewed 18 April 2012, (190) Conley E & Benson T. SNOMED CT: who needs to know what? European Journal for Biomedical Informatics 2011;7(2):40-7. (230) De Lusignan S. The barriers to clinical coding in general practice: a literature review. Med Inform Internet Med 2005;30(2): (238) delacy M. Nosology, mortality, and disease theory in the eighteenth century. J Hist Med Allied Sci 1999;54(2): (19) 258

281 DH4 Pty Ltd. Development of a Business Care for National Clinical Terminologies: needs, requirements and options. Discussion Document for Consultation. DH4 Pty Ltd. Jan (72) Docle Systems. Docle Systems Viewed 5 July 2010, (68) Dolin RH, Mattison JE, Cohn S, Campbell KE, Wiesenthal AM, Hochhalter B et al. Kaiser Permanente's Convergent Medical Terminology. Medinfo 2004;11(Pt 1): (183) Dorland's Illustrated Medical Dictionary, 32nd edition. Philadelphia: Elsevier Saunders, (191) Egdahl A. Linnaeus' "Genera Morborum," and Some of His Other Medical Works. Med Library Hist J 1907;5(3): (21) Eimerl TS. The E Book system for record-keeping in general practice. Med Care 1973;11(2):Suppl-44. (27) Elhanan G, Perl Y, Geller J. A Survey of Direct Users and Uses of SNOMED CT: 2010 Status. AMIA Annu Symp Proc 2010;2010: (182) Elkin PL, Brown SH, Carter J, Bauer BA, Wahner-Roedler D, Bergstrom L et al. Guideline and quality indicators for development, purchase and use of controlled health vocabularies. Int J Med Inform 2002;68(1-3): (127) Elkin PL, Brown SH, Husser CS, Bauer BA, Wahner-Roedler D, Rosenbloom ST et al. Evaluation of the content coverage of SNOMED CT: ability of SNOMED clinical terms to represent clinical problem lists. Mayo Clin Proc 2006;81(6): (199) European Community for Standardization (CEN). CEN/TC Standards under development Viewed 17 December 2011, icalcommittees/pages/wp.aspx?param=6232&title=cen%2ftc+251. (154) Family Medicine Research Centre. Current uses of ICPC-2 PLUS Viewed 27 July 2010, (64) Family Medicine Research Centre. ICPC-2 PLUS functionality requirements for developers Viewed 18 October 2010, (98) 259

282 Family Medicine Research Centre. ICPC-2 PLUS: the BEACH coding system Viewed 15 June 2012, (97) Family Medicine Research Centre. Past commissioned research Viewed 7 March 2012, (166) Family Medicine Research Centre. Theses and treaties - completed Viewed 29 May 2012, (78) Field D, Ward AM, Lopez DG. Apart from Medicare, what other work do GPs do? Aust Fam Physician 2003;32(6): (123) Froom J. The International Classification of Health Problems for Primary Care. Med Care 1976;14(5): (34) Fung KW, McDonald C, Srinivasan S. The UMLS-CORE project: a study of the problem list terminologies used in large healthcare institutions. J Am Med Inform Assoc 2010;17(6): (125) Gastroenterological Society of Australia. Reflux disease: Gastro-oesophageal Reflux Disease in adults (clinical update) Viewed 16 March 2012, (194) General Practice Computing Group. Final report of the General Practice Coding Jury. Barton, ACT: General Practice Computing Group, Viewed 9 October 2012, ry.pdf. (67) General Practice Registrars Australia. General Practice Registrars Australia: About us Viewed 15 June 2012, (94) Giannangelo K & Fenton SH. SNOMED CT survey: an assessment of implementation in EMR/EHR applications. Perspect Health Inf Manag 2008;5:7. (181) Gibbons P, Arzt N, Surke-Beebe S, Chute C, Dickinson G, Flewelling T et al. Coming to Terms: Scoping Interoperability for Health Care. Health Level Seven EHR Interoperability Work Group, Viewed 10 August 2012, (13) 260

283 Graunt J. Natural and political observations mentioned in a following index, and made upon the Bills of Mortality. London: The Roycroft, for John Martin, James Allestry, The Dicas, at the Sign of the Bell in St Pauls Church-yard, (18) Hansen DP, Kemp ML, Mills SR, Mercer MA, Frosdick PA, Lawley MJ. Developing a national emergency department data reference set based on SNOMED CT. Med J Aust 2011;194(4):S8-10. (107) Harris MF, Bailey L, Snowdon T, Litt J, Smith JW, Joyner B et al. Developing the guidelines for preventive care - two decades of experience. Aust Fam Physician 2010;39(1-2):63-5. (151) Henderson J, Britt H, Miller G. Extent and utilisation of computerisation in Australian general practice. Med J Aust 2006;185(2):84-7. (109) Heneghan C, Glasziou P, Thompson M, Rose P, Balla J, Lasserson D et al. Diagnostic strategies used in primary care. BMJ 2009;338:b946. (171) Hennen BK. Continuity of care in family practice. Part 1: dimensions of continuity. J Fam Pract 1975;2(5): (233) HL7. Electronic health record - system functional model, release 1. Feb (136) Howie JG. Diagnosis--the Achilles heel? J R Coll Gen Pract 1972;22(118): (170) IBM Consulting Group Health Practice. Clinical & Administrative General Practice Computer Systems Consultancy: Functional Requirements Specification Report Viewed 22 September 2012, (237) Improvement Foundation (Australia). APCC Program Topic Measures and Other Improvement Measures Viewed 23 June 2011, (120) International Health Terminology Standards Development Organisation. SNOMED CT value proposition Viewed 14 July 2012, (43) 261

284 International Health Terminology Standards Development Organisation. Who are the IHTSDO Charter Members and what does this mean? Viewed 20 July 2012, (55) International Health Terminology Standards Development Organisation. Members of IHTSDO Viewed 20 July 2012, (56) International Health Terminology Standards Development Organisation. Governance and Advisory Viewed 20 July 2012, (57) International Health Terminology Standards Development Organisation. Special Interest Groups Viewed 20 July 2012, (58) International Health Terminology Standards Development Organisation. SNOMED CT Technical Reference Guide: January 2011 international release. Jan (99) International Health Terminology Standards Development Organisation. SNOMED CT Technical Implementation Guide: January International Health Terminology Standards Development Organisation. Jan (100) International Health Terminology Standards Development Organisation. SNOMED CT Editorial Guide: July 2011 international release (GB English) Viewed 9 January 2012, e_current-en-gb_int_ pdf. (102) International Health Terminology Standards Development Organisation. Online Request Submission System Viewed 9 October 2012, (103) International Health Terminology Standards Development Organisation. SNOMED CT Release Format 2.0 Reference Set Specifications Viewed 6 July 2011, OMED_CT_Release_Format_2_-_Refset_Specifications_v1.0.pdf. (106) International Health Terminology Standards Development Organisation. SNOMED CT User Guide: July 2011 international release Viewed 18 June 2012, (128) 262

285 International Health Terminology Standards Development Organisation. IHTSDO Annual Quality Report Viewed 27 August 2012, ssurance/ihtsdo_annual_quality_report_2011_ _v1_00.pdf. (215) International Health Terminology Standards Development Organisation. About SNOMED CT Viewed 16 August 2012, (218) International Health Terminology Standards Development Organisation. Mapping SNOMED CT to ICD-10 Project [registration required] Viewed 27 September 2012, oj/. (239) International Health Terminology Standards Development Organisation & Intern. Family Practice/General Practice RefSet and ICPC mapping project [registration required] Viewed 27 September 2012, /. (240) International Health Terminology Standards Development Organisation. Wonca and IHTSDO seek to strengthen informatics in primary care Viewed 27 September 2012, a_ihtsdo_agreement_announcement.pdf. (246) International Organization for Standardization (ISO) & International. ISO/DTR 12300: Health informatics - Principles of mapping between terminological systems Viewed 27 August 2012, ber= (241) International Organization for Standardization (ISO). Information technology - Vocabulary - part 1: Fundamental terms Viewed 8 July 2011, mber=7229. (114) Israel RA. The history of the International Classification of Diseases. Health Bulletin 1991;49(1):62-6. (25) Jones R, Barraclough K, Dowrick C. When no diagnostic label is applied. BMJ 2010;340:c2683. (169) 263

286 Joss MW & Wszola S. The engines that can: text search and retrieval software, their strategies, and vendors. CD ROM Professional 1996;9(6): (115) Khan A & Western M. Does attitude matter in computer use in Australian general practice? A zero-inflated Poisson regression analysis. HIM J 2011;40(2): (118) Lai C, Przychodzin D, Bunch C, Morrison T, Chaney R. The resource impact of supporting SNOMED CT updates in a clinical decision support system (163) Lamberts H, Meads S, Wood M. Classification of reasons why persons seek primary care: pilot study of a new system. Public Health Rep 1984;99: (37) Lamberts H, Meads S, Wood M. Results of the international field trial with the Reason for Encounter Classification. Soz Praventivmed 1985;30:80-7. (39) Latham RG. The works of Thomas Sydenham, M.D. London: The Sydenham Society, (22) Lau F, Simkus R, Lee D. A methodology for encoding problem lists with SNOMED CT in general practice. Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008): May 31-June 2, 2008, Arizona, USA Epub 2008 May 31. (188) Lauteslager M, Brouwer HJ, Mohrs J, Bindels PJ, Grundmeijer HG. The patient as a source to improve the medical record. Fam Pract 2002;19(2): (139) Lee D, Cornet R, Lau F. Implications of SNOMED CT versioning. Int J Med Inform 2011;80(6): (162) Lee DH, Lau FY, Quan H. A method for encoding clinical datasets with SNOMED CT. BMC Med Inform Decis Mak 2010;10:53. (209) Letrilliart L, Bacis A-K, Mennerat F, Colin C. Interface terminologies: a case study on the International Classification of Primary Care. World Academy of Science, Engineering and Technology 2009;59: (66) Mantena S & Schadow G. Evaluation of the VA/KP problem list subset of SNOMED as a clinical terminology for electronic prescription clinical decision support. AMIA Annu Symp Proc 2007: (184) 264

287 Markwell D, Sato L, Cheetham E. Representing clinical information using SNOMED Clinical Terms with different structural information models. 2008;72-9. (235) Matney SA, Warren JJ, Evans JL, Kim TY, Coenen A, Auld VA. Development of the nursing problem list subset of SNOMED CT(R). J Biomed Inform (207) McDonald K & James S. NEHTA outlines PCEHR roadmap for coming months. Pulse+IT magazine Viewed 24 July 2012, ticle&id=1079:nehta-outlines-pcehr-roadmap-for-comingmonths&catid=16:australian-ehealth&itemid=328. (143) McDonald K. September the month for PCEHR-enabled GP software. Pulse+IT magazine Viewed 24 August 2012, ticle&id=1017:september-the-month-for-pcehr-enabled-gpsoftware&catid=16:australian-ehealth&itemid=328. (244) McDonnell L & Usherwood T. International medical graduates - challenges faced in the Australian training program. Aust Fam Physician 2008;37(6): (195) McKusick VA. On lumpers and splitters, or the nosology of genetic disease. Perspect Biol Med 1969;12(2): (42) Medical Objects. General practitioners: secure messaging Viewed 28 April 2011, (113) Microsoft Corporation. What is Microsoft Health CUI? Viewed 9 September 2012, (117) Miller G, O'Halloran J, Bayram C. General practice EHR and data query minimum data set development Viewed 9 October 2012, PCG_SP1_%20GPracEHRMinDataSetFinalReport.pdf. (80) Miller GC & Britt H. The development of a code set for pathology orders by general practitioners. The University of Sydney (167) Miller GC, Britt H, Jurd E. The Aus-Read Trial. N Z Fam Physician 1994;21(4): (61) 265

288 Moerkerke C & Ceusters W. The myth of preferred terms in medical sublanguage and its impact on natural language understanding applications: an empirical study. 2000; (223) Montella D, Brown SH, Elkin PL, Jackson JC, Rosenbloom ST, Wahner- Roedler D et al. Comparison of SNOMED CT versus Medcin terminology concept coverage for mild Traumatic Brain Injury. AMIA Annu Symp Proc 2011;2011: (211) Moynihan R, Doust J, Henry D. Preventing overdiagnosis: how to stop harming the healthy. BMJ 2012;344:e3502. (221) Moynihan R. Who benefits from treating prehypertension? BMJ 2010;341:c4442. (220) Nachimuthu SK & Lau LM. Practical issues in using SNOMED CT as a reference terminology. Stud Health Technol Inform 2007;129(Pt 1): (165) Nadkarni PM & Darer JA. Migrating existing clinical content from ICD-9 to SNOMED. J Am Med Inform Assoc 2010;17(5): (206) National Centre for Classification in Health. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian modification (ICD-10-AM) - sixth edition. Sydney: National Centre for Classification in Health, The University of Sydney, (111) National E-Health Transition Authority. Clinical information systems and ehealth records Viewed 25 June 2012, (144) National E-Health Transition Authority. Clinical Leaders Viewed 1 February 2011, (110) National E-Health Transition Authority. Delivering E-Health's Foundations Viewed 19 July 2012, (73) National E-Health Transition Authority. Event summary: structured content specification (Version 1.1) [registration required] Viewed 23 August 2012, Id=308#0. (148) 266

289 National E-Health Transition Authority. Key foundation of electronic health communication in place Viewed 28 September 2007, &Itemid=144. (75) National E-Health Transition Authority. Mapping to SNOMED CT-AU: guidelines and requirements (Version 1.0) [registration required] Viewed 16 August 2012, (216) National E-Health Transition Authority. NCTIS Terminology Product Releases: Release Cycles and Versioning Viewed 17 August 2012, (232) National E-Health Transition Authority. NEHTA data group specifications: problem/diagnosis Viewed 23 January 2012, (133) National E-Health Transition Authority. NEHTA launches the first release of SNOMED CT-AU (104) National E-Health Transition Authority. NEHTA secures international health terminology for Australia Viewed 19 July 2012, (53) National E-Health Transition Authority. NEHTA welcomes new standards development organisation Viewed 19 July 2012, (52) National E-Health Transition Authority. New national entity to drive e-health Viewed 5 October 2012, (74) National E-Health Transition Authority. PCEHR Specification & Standards Development: Upcoming collaboration opportunities on the development of the Personally Controlled Electronic Health Record (PCEHR) specifications Viewed 27 September 2012, T_1700.wmv. (245) 267

290 National E-Health Transition Authority. Shared health summary: structured content specification (Version 1.1) [registration required] Viewed 23 August 2012, Id=312#0. (149) National E-Health Transition Authority. SNOMED CT-AU release note. Nov (105) National E-Health Transition Authority. SNOMED CT-AU Release Note (Document ID NEHTA 666:2010) Viewed 10 May 2011, (108) National E-Health Transition Authority. SNOMED CT-AU Viewed 10 May 2011, (76) National E-Health Transition Authority. What is a PCEHR? Viewed 25 June 2012, (145) National Heart Foundation of Australia (National blood pressure and vascular disease advisory committee). Guide to management of hypertension 2008: assessing and managing raised blood pressure in adults (Updated December 2010) Viewed 5 October 2012, delines2008to2010update.pdf. (193) NHS Information Authority. The Clinical Terms Version 3 (The Read Codes) Viewed 5 October 2012, chap1.pdf. (48) NSW Health. SNOMED CT within the NSW Public Health System Viewed 4 August 2012, (177) Nystrom M, Vikstrom A, Nilsson GH, Ahlfeldt H, Orman H. Enriching a primary health care version of ICD-10 using SNOMED CT mapping. J Biomed Semantics 2010;1(1):7. (205) O'Halloran J, Miller GC, Britt H. Defining chronic conditions for primary care with ICPC-2. Fam Pract 2004;21(4): (79) 268

291 Okkes IM & Lamberts H. Classification and the domain of family practice. In: Jones R (ed). The Oxford Textbook of Primary Medical Care (Volume 1). Oxford: Oxford University Press, 2003; (138) Oliver DE, Shahar Y, Shortliffe EH, Musen MA. Representation of change in controlled medical terminologies. Artif Intell Med 1999;15(1): (159) Oon YK. Docle coding and classification system browser Viewed 20 July 2012, (65) Osornio AL, Luna D, Gambarte ML, Gomez A, Reynoso G, de Quiros FG. Creation of a local interface terminology to SNOMED CT. Stud Health Technol Inform 2007;129(Pt 1): (187) Oxford Dictionaries. Definition of fit Viewed 4 August 2012, (176) Patrick J, Wang Y, Budd P. Developing SNOMED CT subsets from clinical notes for intensive care service. Health Care and Informatics Review Online 2008;12(3): (130) Patrick J. A Study of a Health Enterprise Information System. School of Information Technologies, University of Sydney (179) Pearce C, Shearer M, Gardner K, Kelly J. A division's worth of data. Aust Fam Physician 2011;40(3): (122) Penz JF, Brown SH, Carter JS, Elkin PL, Nguyen VN, Sims SA et al. Evaluation of SNOMED coverage of Veterans Health Administration terms. Medinfo 2004;11(Pt 1): (202) Raiez F, Arts D, Cornet R. Terminological system maintenance: a procedures framework and an exploration of current practice. Stud Health Technol Inform 2005;116: (160) Rector A & Iannone L. Lexically suggest, logically define: quality assurance of the use of qualifiers and expected results of post-coordination in SNOMED CT. J Biomed Inform 2012;45(2): (213) Rector A, Iannone L, Stevens R. Quality assurance of the content of a large DL-based terminology using mixed lexical and semantic criteria: experience with SNOMED CT. Presented at the K-CAP'11 Proceedings of the sixth international conference on Knowledge capture; 25 Jun 2011; Banff, Alberta: (229) 269

292 Rector AL & Brandt S. Why do it the hard way? The case for an expressive description logic for SNOMED. J Am Med Inform Assoc 2008;15(6): (227) Rector AL, Brandt S, Schneider T. Getting the foot out of the pelvis: modeling problems affecting use of SNOMED CT hierarchies in practical applications. J Am Med Inform Assoc 2011;18(4): (228) Rector AL, Qamar R, Marley T. Binding ontologies and coding systems to electronic health records and messages. Applied Ontology 2009;4(1): (124) Rector AL. Clinical terminology: why is it so hard? Methods Inf Med 1999;38(4-5): (50) Rector AL. The interface between information, terminology, and inference models. Stud Health Technol Inform 2001;84(Pt 1): (129) Richards B, Bolton P, Veale B, Quinlan F. Information technology in general practice: a monograph commissioned by the General Practice Branch of the Commonwealth Department of Health and Aged Care. Publications Production Unit (Public Affairs, Parliamentary and Access Branch), Commonwealth Department of Aged Care. Monograph series No (7) Richesson RL, Andrews JE, Krischer JP. Use of SNOMED CT to represent clinical research data: a semantic characterization of data items on case report forms in vasculitis research. J Am Med Inform Assoc 2006;13(5): (198) Roberts RF, Innes KC, Walker SM. Introducing ICD-10-AM in Australian hospitals. Med J Aust 1998;169 Suppl:S32-S35. (112) Robinson D, Schulz E, Brown P, Price C. Updating the Read Codes: userinteractive maintenance of a dynamic clinical vocabulary. J Am Med Inform Assoc 1997;4(6): (157) Rogers J. Terminologies: Getting clinical data in; getting useful data out Viewed 30 September 2011, SNOMED.pdf. (131) Rosenbloom ST, Brown SH, Froehling D, Bauer BA, Wahner-Roedler DL, Gregg WM et al. Using SNOMED CT to represent two interface terminologies. J Am Med Inform Assoc 2009;16(1):81-8. (203) 270

293 Rosenbloom ST, Miller RA, Johnson KB, Elkin PL, Brown SH. Interface terminologies: facilitating direct entry of clinical data into electronic health record systems. J Am Med Inform Assoc 2006;13(3): (217) Royal Australian College of General Practitioners. General practice fact sheet: Medical history Viewed 10 December 2011, GPehealthsummary/Fact-sheet-medical-history.pdf. (142) Royal Australian College of General Practitioners. Greater access to a 'medical home' key to improved health outcomes for Australia's elderly Viewed 8 August 2012, (6) Royal Australian College of General Practitioners. What is general practice? Viewed 24 February 2011, (1) Ryan A, Patrick J, Herkes R. Introduction of Enhancement Technologies into the Intensive Care Service, Royal Prince Alfred Hospital, Sydney. Health Information Management Journal 2008;37(1):40-5. (180) Sampalli T, Shepherd M, Duffy J, Fox R. An evaluation of SNOMED CT in the domain of complex chronic conditions. Int J Integr Care 2010;10:e038. (210) Sauvages F. Nosologica Methodica Sistens Morborum Classes. Amsterdam: Fratum de Tournes, (20) Schulz S, Suntisrivaraporn B, Baader F, Boeker M. SNOMED reaching its adolescence: ontologists' and logicians' health check. Int J Med Inform 2009;78 Suppl 1:S86-S94. (225) Schulz S, Suntisrivaraporn B, Baader F. SNOMED CT's problem list: ontologists' and logicians' therapy suggestions. Stud Health Technol Inform 2007;129(Pt 1): (224) Scott P, Walker D, McIsaac P. The Australian General Practice Vocabulary and the ICD-10-AM index: parsing and classification issues Viewed 26 April 2012, pdf. (70) 271

294 Sommer J, Macdonald W, Bulsara C, Lim D. Grunt language versus accent: the perceived communication barriers between international medical graduates and patients in Central Wheatbelt catchments. Aust J Primary Health (196) Spackman KA & Reynoso G. Examining SNOMED from the Perspective of Formal Ontological Principles: Some Preliminary Analysis and Observations. 2004; (219) Spackman KA. Normal forms for description logic expressions of clinical concepts in SNOMED RT. Proc AMIA Symp 2001: (101) Spackman KA. Rates of change in a large clinical terminology: three years experience with SNOMED Clinical Terms. AMIA Annu Symp Proc 2005: (158) St.Sauver J, Buntrock J, Rademacher D, Albrecht D, Gregg M, Ihrke D et al. Comparison of Mayo Clinical Coding Systems Viewed 4 August 2012, mayoresearch.mayo.edu/mayo/research/biostat/upload/83.pdf. (185) Standards Australia. The language of health concept representation (AS ). Sydney: Standards Australia, (16) Standards Australia. What is a standard? Viewed 22 September 2012, ages/default.aspx. (236) Stanek J & Simon B. SNOMED-CT Coding in General Practice - Barriers and Opportunities. Brunswick East, Vic: Health Informatics Society of Australia; 2008; (77) Stearns MQ, Price C, Spackman KA, Wang AY. SNOMED Clinical Terms: Overview of the development process and project status. JAMIA Symposium Supplement 2001: (51) Stroetmann VN, Kalra D, Lewalle P, Rector A, Rodrigues JM, Stroetmann KA et al. Semantic Interoperability for Better Health and Safer Healthcare: deployment and research roadmap for Europe Viewed 23 August 2011, 9/2009semantic-health-report.pdf. (14) 272

295 Susser M & Adelstein A (eds) Vital statistics: a memorial volume of selections from the reports and writings of William Farr, with an introduction by Mervyn Susser and Abraham Adelstein. Metuchen, New Jersey: The Scarecrow Press, Inc., (17) The Macquarie Concise Dictionary (revised 3rd edition). Macquarie University: The Macquarie Library Pty Ltd, (192) The Research Committee of the Council of the College of General Practitioners. A Classification of Disease. J Coll Gen Pract Res Newsl 1959;2(2): (26) The Royal Australian College of General Practitioners. Computers as Medical Tools: Conference report from the RACGP 5th Computer Conference (59) The Royal Australian College of General Practitioners. Guidelines for preventive activities in general practice (The Red Book) Viewed 15 December 2011, (150) The Royal Australian College of General Practitioners. Standards for general practices: 4th edition Viewed 10 May 2011, (95) Thiedeman B, Miller GC, Britt H, Rowed D. General Practice Data Model and Core Data Set Project: final project report Viewed 9 October 2012, al_project_report.pdf. (12) Thomas KB. Temporarily dependent patient in general practice. Br Med J 1974;1(5908): (175) Thomson J. The works of William Cullen, M.D., Volume One. Edinburgh: William Blackwood, Edinburgh and T & G Underwood, London, (41) U.S.National Library of Medicine. UMLS Quick Start Guide Viewed 30 August 2012, (214) Van Vleck TT, Wilcox A, Stetson PD, Johnson SB, Elhadad N. Content and structure of clinical problem lists: a corpus analysis. AMIA Annu Symp Proc 2008: (140) 273

296 Vikstrom A, Nystrom M, Ahlfeldt H, Strender LE, Nilsson GH. Views of diagnosis distribution in primary care in 2.5 million encounters in Stockholm: a comparison between ICD-10 and SNOMED CT. Inform Prim Care 2010;18(1): (132) Wade G & Rosenbloom ST. Experiences mapping a legacy interface terminology to SNOMED CT. BMC Med Inform Decis Mak 2008;8 Suppl 1:S3. (204) Wade G & Rosenbloom ST. The impact of SNOMED CT revisions on a mapped interface terminology: terminology development and implementation issues. J Biomed Inform 2009;42(3): (164) Walker D. Stage 2 of the GP vocabulary project Viewed 26 April 2012, (71) AY, Barrett JW, Bentley T, Markwell D, Price C, Spackman KA et al. Mapping between SNOMED RT and Clinical terms version 3: a key component of the SNOMED CT development process. Proc AMIA Symp 2001: (197) Wang Y, Patrick J, Miller G, O'Hallaran J. A computational linguistics motivated mapping of ICPC-2 PLUS to SNOMED CT. BMC Med Inform Decis Mak 2008;8 Suppl 1:S5. (81) Weed LL. Medical records, medical education and patient care. Chicago: Year Book Medical Publishers, Inc., (11) Westbury RC & Tarrant M. Classification of disease in general practice: a comparative study. Can Med Assoc J 1969;101(10):82-7. (29) Westbury RC. A plan to develop an international classification of diseases in family medicine. Melbourne: The Organising Committee of the Fifth World Conference on General Practice, Royal Australian College of General Practitioners; 1972; (30) Westra BL, Bauman R, Delaney CW, Lundberg CB, Petersen C. Validation of concept mapping between PNDS and SNOMED CT. AORN J 2008;87(6): (208) White KL. The ecology of medical care: origins and implications for population-based healthcare research. Health Serv Res 1997;32(1): (4) 274

297 World Health Organization. International Classification of Primary Care, Second edition (ICPC-2) Viewed 17 December 2008, (173) World Health Organization. International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Volume 2. Geneva: World Health Organization, (24) World Organisation of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians. Fifth world conference on general practice proceedings. Melbourne: Royal Australian College of General Practitioners, (31) World Organisation of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians. International Classification of Health Problems in Primary Care (ICHPPC). Chicago: American Hospital Association, (33) 275

298 Appendices Appendix 1: Project plan & requirements specification: General Practice Reference Set Project (reproduced with the permission of the National e-health Transition Authority) 276

299 nehta Project Plan & Requirements Specification General Practice Reference Set Project Version Commercial in Confidence National E-Health Transition Authority

300 Project Plan & Requirements Specification General Practice Reference Set Project National E-Health Transition Authority Ltd Level Pitt Street Sydney, NSW, 2000 Australia. Disclaimer NEHTA makes the information and other material ( Information ) in this document available in good faith but without any representation or warranty as to its accuracy or completeness. NEHTA cannot accept any responsibility for the consequences of any use of the Information. As the Information is of a general nature only, it is up to any person using or relying on the Information to ensure that it is accurate, complete and suitable for the circumstances of its use. Document Control This document is maintained in electronic form. The current revision of this document is located on the NEHTA Web site and is uncontrolled in printed form. It is the responsibility of the user to verify that this copy is of the latest revision. Copyright 2010 NEHTA. This document contains information which is protected by copyright. All Rights Reserved. No part of this work may be reproduced or used in any form or by any means graphic, electronic, or mechanical, including photocopying, recording, taping, or information storage and retrieval systems without the permission of NEHTA. All copies of this document must include the copyright and other information contained on this page. ii

301 nehta Document information Document information Document owner Document owner The National Clinical Terminology and Information Service Change history Version/Release/Date Comments 0.5 / Draft release to GPRS - Support Group for initial meeting and feedback 1.0 / Final - prepared for publication iii

302 Project Plan & Requirements Specification General Practice Reference Set Project This page is blank intentionally iv

303 nehta Table of contents Table of contents 1 Executive summary Introduction Purpose Intended audience Scope Overview of the project About this report Questions and feedback The general practice environment General practice in Australia Use of computers in Australian general practice Project background and rationale Nature and scope of consultation Organisational stakeholders Vendors GPs NEHTA Clinical leads Breadth of consultation Related projects NEHTA Emergency department reference sets (EDRS) IHTSDO General/family practice reference sets National Library of Medicine Clinical Observations Recording and Encoding (CORE) subset Map from Best Practice terms to SNOMED CT Use of SNOMED CT in Houston Medical SNOMED CT UK GP subset Other related projects Results from consultation process Project scope GP recommendations Recommendations for reference sets Mapping recommendations Implementation recommendations Maintenance requirements Other recommendations Preliminary project plan Overview Indicative project delivery time frame Steps for completion of the GPRS Phase 2: Design phase Phase 3: Build phase Phase 4: Validate and test phase Phase 5: Finalise GPRS pack Phase 6: GPRS Early Adopter Program Key deliverables Project governance Resources Assumptions and dependencies Risks Use cases Explanation of use cases v

304 Project Plan & Requirements Specification General Practice Reference Set Project Patient recall Coded data for the Australian Primary Care Collaboratives Program Research purposes Electronic communication using HL7 - referrals Electronic communication using HL7 discharge summaries Decision support Management of legacy data Stakeholders Requirements overview Purpose of providing requirements Scope Context About the requirements Source of requirements Identification Wording Priority Detailed requirements Content requirements BR.GPRS.0001 Scope of content - [HIGH] BR.GPRS.0002 Navigational groupings - [HIGH] BR.GPRS.0003 Exception handling - [HIGH] BR.GPRS.0004 Legacy data - [HIGH] BR.GPRS.0005 Links to classifications - [HIGH] BR.GPRS.0006 Maintenance - [MEDIUM] Implementation requirements BR.GPRS.0007 Release format - [HIGH] BR.GPRS.0008 Terminology viewers - [HIGH] BR.GPRS.0009 Impact of implementation -[MEDIUM] BR.GPRS.0010 Usability - [MEDIUM] BR.GPRS.0011 End user education - [HIGH] BR.GPRS.0012 Technical education - [HIGH] Governance BR.GPRS GPRS licence - [HIGH] BR.GPRS GPRS governance - [HIGH] BR.GPRD GPRS maintenance process - [HIGH] BR.GPRS Change management - [MEDIUM] BR.GPRS.0017 Release cycle - [HIGH] References Appendix A: Definitions, Acronyms and Abbreviations vi

305 nehta Executive summary 1 Executive summary At present, information collected within general practice computer systems is not standardised. Each GP software vendor uses different codesets and terminologies that GPs may choose to use to record their clinical notes. The Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT 1 ) is recognised as a world-leading standardised clinical terminology. The Australian government purchased a licence for SNOMED CT in 2005 and uses it as the basis for its Australian release, SNOMED CT -AU. At present, SNOMED CT has not been implemented in any of the available GP desktop computer systems in Australia, and it is recognised that in its entirety SNOMED CT is too large and complex to be effectively used in clinical subdomains. In late 2009 the Family Medicine Research Centre, University of Sydney was contracted by the National e-health Transition Authority (NEHTA) to gather requirements for a number of reference sets collectively known as the general practice reference set (GPRS) of SNOMED CT AU and prepare a project plan and requirements specification for development of the GPRS. Engagement workshops were held with key stakeholders, including general practitioners (GPs) who currently use computers in their clinical practice, GP software vendors and NEHTA Clinical Leads. The feedback clearly identified that the GPRS must reflect the terms used in Australian general practice, and therefore should be derived from codesets and terminologies already used in this setting. GPs and software vendors indicated that SNOMED CT AU would be most useful for preparing health summaries and problem lists, and for populating referrals. GP software vendors saw linkage to decision support systems as an important outcome. Maps from the GPRS to international classifications were regarded as a useful aid for the ongoing reporting of data from general practice for a variety of purposes - for individual GPs, practices, divisions and across Australia. As a result of the consultation process, this report recommends that a general practice reference set should be created using SNOMED CT AU as its foundation. The data elements that could be populated with the GPRS include reason for encounter/reason for visit, problem/diagnosis/reason for prescription and patient allergies. The GPRS would incorporate symptoms and signs, diagnoses and procedures. The GPRS would also allow personal and family histories to be populated. The termset used in the Best Practice software, the Australian Classification and Terminology of Community Health (CATCH), ICPC-2 PLUS and the University of Adelaide GP termset (derived from Medical Director terms) would be used as important sources of data for the project. Creating links from relevant data sources to SNOMED CT AU will ensure the ongoing usefulness of legacy data collected in GP computer software. Classifying the GPRS to the International Classification of Primary Care (ICPC-2) will be undertaken for data reporting purposes. The development and maintenance of the GPRS and classification linkages will facilitate the uptake of SNOMED CT AU, and encourage the use of SNOMED CT AU as a standardised terminology. The creation of links from codesets and terminologies currently used in general practice to the GPRS will help ensure the maintenance of historic records in existing systems. Adoption of the GPRS will enhance the quality of data entered into general practice clinical software, and enable the sharing of clinical information within the wider Australian healthcare system. 1 SNOMED CT is a registered trademark of the International Health Terminology Standards Development Organisation

306 Project Plan & Requirements Specification General Practice Reference Set Project This report will be presented for endorsement to the GPRS Support Group, consisting of representative general practice associations

307 nehta Introduction 2 Introduction 2.1 Purpose The purpose of this report is to document and report requirements for a group of SNOMED CT-AU reference sets collectively known as the General practice reference set (GPRS). This report along with additional engagement will be used to create design specifications for the development of the GPRS and associated maps based on SNOMED CT-AU. 2.2 Intended audience 2.3 Scope This document has been written for the NEHTA CTI Management Group and GPRS Support Group (GPRS-SG) for endorsement of the requirements for the GPRS and proposed project plan. This report presents a scan of the general practice environment, reports the results of consultation undertaken during the first phase (of six) of this project, and provides a preliminary set of requirements for the development of the GPRS and maps to classifications. 2.4 Overview of the project During 2009, the National Clinical Terminology and Information Service (NCTIS) within NEHTA commissioned the development of a group of reference sets (collectively known as the General practice reference set GPRS) and associated maps to currently used classifications. The objectives of the project are twofold. The first objective is to design and develop the GPRS for the Australian release of the Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT-AU). The second objective is to map the reference sets to appropriate international classifications currently used in the Australian general practice environment. The project has been divided into six phases. This report provides the results of the scoping work undertaken during Phase 1, from September to December About this report This report incorporates feedback gathered from GPs and GP clinical software vendors consulted about the use of medical terminology and clinical coding systems in general practice software programs in general, and specifically about the introduction of SNOMED CT-AU reference sets for general practice. Consultation and requirements gathering occurred between October and December Questions and feedback The NCTIS encourages questions, comments or suggestions about the GPRS project and this document from the GPRS SG

308 Project Plan & Requirements Specification General Practice Reference Set Project 3 The general practice environment 3.1 General practice in Australia General practice is an important provider of health care in Australia. General practitioners (GPs) are the first point of contact in the Australian health system, and facilitate access to other healthcare services and providers. Approximately 88% of the Australian population attended general practice in [KNOX2008], with an average of five visits per patient, per year. It is estimated that people in Australia spend 83 minutes with a GP each year [BIND2007]. In the financial year, 107 million general practice services (excluding practice nurse services) were claimed through Medicare, Australia s universal healthcare scheme [BRIT2009]. Results from the Bettering the Evaluation and Care of Health (BEACH) study of Australian general practice indicate that GPs manage 1.5 problems at each patient encounter. On average, the length of general practice encounters is 14.8 minutes [BRIT2009]. Definitions of general practice are changing. The concept of general practice no longer implies a general practitioner (GP) working alone or in consultation with other GPs. In % of GPs had a practice nurse working at their major practice address [BRIT2009] compared with 60.2% in [BRIT2005a]. The introduction of GP superclinics incorporating GPs, nurses, medical specialists, allied health practitioners and other healthcare providers [AGDH2009] further broadens the scope of general practice. 3.2 Use of computers in Australian general practice In , 96% of GPs stated that they used computers in their practice. Of these, 56.4% reported that they use computers for all the relevant clinical functions and are paperless. Three-quarters of GPs with computers reported that they are able to access the internet (74.5%), and 61.1% of GPs use at work [BRIT2009]. 3.3 Project background and rationale As Australia s lead agency for e-health adoption and the National Release Centre for SNOMED CT, the National E-Health Transition Authority (NEHTA) has a significant interest in the timely and successful implementation of SNOMED CT AU into the Australian healthcare system. At present, SNOMED CT AU has not been implemented in any of the available GP desktop clinical systems commonly used in Australia. It is recognised that in its entirety SNOMED CT is too large and complex to be effectively used in clinical subdomains. As such, the IHTSDO 2 has introduced the concept of reference sets, defined as a mechanism used to identify a subset of content from the source dataset [IHTS2009]. 2 IHTSDO is a registered trademark of the International Health Terminology Standards Development Organisation

309 nehta The general practice environment The development of the GPRS is anticipated to facilitate the uptake of SNOMED CT-AU in Australian general practice, and will encourage the use of SNOMED CT-AU as a standardised terminology across the health sector. In turn, this may enhance the quality of data entered into general practice clinical software, and enable the sharing of clinical information within the wider Australian healthcare system. It will also provide GP clinical software vendors who are interested in using SNOMED CT-AU with a relevant set of SNOMED CT-AU concepts to assist with implementation. The creation of the GPRS is consistent with a recommendation from the General Practice Coding Jury in August 2000 that SNOMED CT should be considered as a standardised coding system for general practice [GPCG2000]. NEHTA contracted the Family Medicine Research Centre (FMRC), University of Sydney to gather and prepare a project plan and requirements specification for the development of the GPRS. Requirements for maps from the GPRS to classifications were also requested. Future elements of this project also include: evaluating data elements used in general practice; designing the GPRS and associated maps; building and testing the GPRS; classifying the GPRS to associated classifications; releasing the GPRS and maps for use in general practice; and trial implementations through an early adopter program. 3.4 Nature and scope of consultation The need to consult widely with known stakeholders was identified by NEHTA as a priority for this project to ensure that the final deliverables are acceptable and reasonable to those who will be impacted by this work. A list of stakeholders can be found in Section 6 of this report. For the consultation, stakeholders were divided into the following groups: GP software vendors General practitioners NEHTA Clinical leads Workshops were the primary method used to formally consult with stakeholders in Phase 1, with one-on-one sessions with stakeholders where necessary. Consultation with stakeholders will continue to be undertaken throughout the life of the project using or through future workshops as deemed necessary. Organisational stakeholders, comprising the GPRS-SG will also provide input to the project development as necessary Organisational stakeholders At the time of this report, the GPRS-SG has been recruited by NEHTA, and this group met for the first time in March This group will provide additional input and advice during the project duration. Members of this group represent the following organisations: Australian General Practice Network (AGPN) Royal Australian College of General Practitioners (RACGP) Australian Medical Association (AMA) Rural Doctors Association of Australia (RDAA)

310 Project Plan & Requirements Specification General Practice Reference Set Project Aged Care IT Council Australian Practice Nurses Association (APNA) Australian Institute of Health and Welfare (AIHW) Medical Software Industry Association (MSIA) Clinical Terminology and Information Reference Group (CTIRG) Vendors An invitation to a workshop designed for GP software vendors was sent to vendors using the Medical Software Industry Association s (MSIA) list, with a reminder sent two weeks later. Vendors who did not attend the workshop, but were identified as having a significant share of the GP software market in Australia, were individually approached requesting involvement in the project, and one-on-one meetings were arranged where possible. Mechanisms to engage software vendors who have not yet been consulted are in process, and consultation with vendors will continue into Phase 2 of the project GPs The GP workshop was comprised of GPs who completed the BEACH study of general practice and met the following criteria: they identified themselves as computer-using GPs, using clinical software for all or most of the available functions; and they were located in New South Wales. BEACH is a national study of general practice activity. Annually 1,000 GPs from around Australia complete information about 100 patient encounters. GPs who participate in BEACH are asked to indicate whether they use computers in their clinical practice, for which functions they use computers in practice, and which clinical software program they use. The aim of the GP consultation was to involve a group of GPs who are representative of GPs using the different brands of clinical software around Australia. Lists of GPs were generated according to the brand of clinical software used. Letters of invitation were sent to GPs requesting their participation in the workshop and non-respondents were followed up by telephone NEHTA Clinical leads NEHTA has recruited a group of clinicians to provide expert input into the adoption of e-health in Australia. An invitation was sent to the NEHTA Clinical leads, and a number of GP Clinical leads attended the workshop

311 nehta The general practice environment 3.5 Breadth of consultation Tables 1 and 2 provide details of the numbers and types of stakeholders consulted. Table 1: Breadth of consultation Vendors Consulted Contacted but yet to consult No response ACS Computing Intrahealth Houston Medical Best Practice Medtech NT Health Communicare Genie Solutions Global Health HCN (Medical Director) isoft Medical Objects MIMS Stat Health Zedmed Note: GP clinical software vendors not consulted during Phase 1 of the project will be contacted early in Phase 2. Table 2: Breadth of consultation Software used by GP attendees Software used by attendees Best Practice Genie Solutions HCN (Medical Director) Medtech isoft (Practix) Software not used by attendees Communicare NT Health Stat Health Intrahealth Houston Medical Promed Zedmed Note: This table includes participants in both the GP and Clinical Lead workshops

312 Project Plan & Requirements Specification General Practice Reference Set Project Table 3 provides a comparison of market share of each GP software program against the consultation conducted. Both GP and vendor consultation is included. Table 3: Comparison of market share against consultation process Company Market share Vendor GP consulted? 3 4 (%) consulted? (Number) (Yes/No) ACS Computing 0.0 Yes 0 Best Practice 6.3 Yes 1 5 Communicare 0.2 Yes 0 Genie Solutions 3.1 Yes 1 Global Health 0.0 Yes 0 HCN (Medical Director) 66.2 Yes 5 Houston Medical 0.1 No 0 Intrahealth 0.9 No 0 isoft (Practix) 6.5 Yes 1 isoft (Monet) 5.1 Yes 0 Medical Objects 0.0 Yes 0 Medtech 5.8 No 1 MIMS Not applicable Yes 0 NT Health 0.2 No 0 Stat Health Not applicable Yes 0 Zedmed 3.1 Yes Based on BEACH study responses to the question What clinical software is used? between April 2008 and March 2009; n =914. Figures will not total to 100.0%. Other programs equals 3.2%. Due to BEACH sampling methods the number of GP Communicare users is an underestimate of the true proportion. Communicare is primarily used in Aboriginal Community Controlled Health Services

313 nehta The general practice environment 3.6 Related projects There are a number of projects either completed or running parallel to this project that may impact the development of the project NEHTA Emergency department reference sets (EDRS) NEHTA has developed reference sets for use in hospital emergency departments (the Emergency department reference set or EDRS). These reference sets are due for release in May Presentations to general practice and emergency departments are often of a similar nature and it is expected that there will be considerable overlap between the reference sets. Once the GPRS has been developed it should be compared with the EDRS to determine the extent of overlapping content between the two IHTSDO General/family practice reference sets The International Health Terminology Standards Development Organisation (IHTSDO) and the World Organisation of Family Doctors (Wonca) signed a collaborative agreement in October 2009 to create an International Family Practice/General Practice Special Interest Group within the IHTSDO. As part of this agreement, the IHTSDO included a project in their workplan for to develop an international reference set for general practice/family practice. This reference set will then be mapped to the International Classification of Primary Care, Version 2 (ICPC-2). Two members of the Family Medicine Research Centre are members of this group and will be conducting some of the work involved to create the reference set and maps. Timelines for this project has been scheduled to minimise overlap between the development of the Australian and international reference sets. The Australian GPRS could be one of the general/family practice termsets used to inform the development of the international reference set National Library of Medicine Clinical Observations Recording and Encoding (CORE) subset The National Library of Medicine in the United States has developed a Clinical Observations Recording and Encoding (CORE) subset of SNOMED CT, released in July 2009 [USNL2009]. This subset contains SNOMED CT concepts suitable for use in problem lists. Although not specifically derived from general practice concepts, this subset could be used as a validation measure of the utility of the GPRS Map from Best Practice terms to SNOMED CT During Phase 1 consultation it was identified that Best Practice has mapped a proportion of their termset to SNOMED CT. Best Practice has indicated that they would be willing to provide the mapped termset as a form of validation of the GPRS Use of SNOMED CT in Houston Medical A pilot study was conducted in South Australia using SNOMED CT in Houston s VIP software, published in This study tested the usability of SNOMED CT in general practice, testing SNOMED CT in its entirety [STAN2008]

314 Project Plan & Requirements Specification General Practice Reference Set Project SNOMED CT UK GP subset Relevant content within this subset, developed in the United Kingdom (UK) for use of SNOMED CT by UK GPs, could be used as a form of validation for the GPRS Other related projects Other related projects (and documents) include the IBM Functionality Specifications, [IBMC1997], the General Practice Data Model and Core Data Set Project, [SIMS2000], and the General Practice Computing Group General Practice Reporting Data Set [BRIT2005b]. Archetype specifications will be utilised during the development of the design documentation in Phase 2. Relationships between the GPRS and the projects listed above, and the potential for leveraging methods used during these projects will be investigated further in Phase 2 during the preparation of the GPRS and map design documentation. 3.7 Results from consultation process Project scope The scope of the project is to create a group of SNOMED CT-AU general practice reference sets for use in Australian GP software systems. Maps from the GPRS to relevant international classifications will then be created. As stated in Section 3.1, definitions of general practice are changing. General practice is no longer synonymous with general practitioner. The GPRS must therefore include all aspects of care managed by GPs and any supporting clinicians (e.g. practice nurses). These healthcare providers must have their terminology needs met both individually and from a practice perspective, if a variety of healthcare professionals are working together within a general practice. During consultation, GP software vendors requested that the scope of the reference sets be expanded to a clinical desktop reference set, including terms used by medical specialists and during aged care encounters. On completion of the GPRS options for including the terminology needs of these users, either as an addition to the GPRS or as a separate set of reference sets will be investigated GP recommendations A clear recommendation resulting from consultation with GPs was the need to record clinical information quickly and easily, and clinical terminologies were seen as a useful tool to facilitate this. GPs stated that the primary reason for coding clinical information was to populate clinical information from patient encounters into other areas of the patient record, specifically for populating problem lists and health summaries and for referrals to other healthcare providers. The ability of a coding system to link to decision support tools within GP software systems was also seen as an important reason to use clinical terminologies

315 nehta The general practice environment Including language used in Australian general practice was seen as imperative for the successful use in general practice of a clinical terminology such as SNOMED CT. Codesets and terminologies currently used in Australian general practice contain multiple synonyms to describe a single medical concept. For example, the terms type 2 diabetes, non-insulin dependent diabetes and adult-onset diabetes are all present in codesets and terminologies currently used by GPs in Australia. The extent to which Australian GPs prefer using SNOMED CT-AU preferred terms or synonyms will be investigated further during Phase 3 of the project to ensure that the breadth of language used in Australian general practice is represented in the GPRS. The language used in Australian general practice changes according to geographic location. Incorporating differences in language based on geographic location, for example state-based differences and differences between metropolitan and rural settings will be important to ensure that the GPRS can be used throughout Australia. GPs stated that the size of the terminology in their GP software system should be constrained to ensure that terms are able to be easily located. GPs stated that the inability to find terms easily was a significant barrier to their use of clinical terminologies. GPs also stated that the search and display mechanisms related to clinical terminologies should be improved compared with current methods. This relates to the previous recommendation about the ease of locating terms within clinical terminologies. Specific proposals relating to search and display mechanisms for the GPRS will be included with the implementation pack prepared during Phase Recommendations for reference sets Identification of source codesets and terminologies Development of the GPRS must incorporate terms currently used in Australian general practice. The terminologies or codesets listed below are currently used in Australian general practice and provide a valuable source of the terms used by Australian GPs in their clinical practices. As well as ensuring that terms contained within these codesets are included in the GPRS, maps from each of these terminologies to the GPRS will be required to ensure access to legacy data stored in general practice software systems: ICPC-2 PLUS This is an Australian general practice interface terminology classified to ICPC-2. This problem list consists of just over 8,000 terms derived from terms used by Australian general practitioners and other primary care practitioners in more than 1.5 million GP patient encounter records. This system is currently being used in nine GP clinical software systems used in general practice in Australia. Mapping to SNOMED CT has been scoped using software developed by the Health Information Technologies Research Laboratory (HITRL), University of Sydney. The University of Sydney has consented to this termset being used in the project. Australian Classification and Terminology of Community Health (CATCH) This terminology was developed for community health clinicians by the Community Health Information Management Enterprise (CHIME) consortium and is in limited use in the public community care sector. Although the structure and content of CATCH mean that large sections of the terminology are out of scope for the GPRS, appropriate sections should be included in the GPRS. Amalgamation into a common term set and mapping of the terms to SNOMED CT-AU would be a prerequisite for inclusion in the GPRS. Permission for using the termset in this project has been obtained from Department of Health and Ageing

316 Project Plan & Requirements Specification General Practice Reference Set Project University of Adelaide GP termset derived from Medical Director terms. While this termset was derived from terms used in only one GP clinical software system, this is the most widely used system in Australia. This termset was rationalised by Dr Don Walker from the University of Adelaide in This termset has not been mapped to SNOMED CT. Permission has been obtained from University of Adelaide and the Royal Australian College of General Practitioners for using the termset during this project. Best Practice termset used in Best Practice. This termset has been partially mapped to SNOMED CT. Consent has been given to provide the termset and map to SNOMED CT for this project. The inclusion of terms from the Medical Director software, including Docle codes, will be discussed with the owners of these codesets during Phase Terminology used in data elements from GP clinical software Implementation of the GPRS requires that content within the reference sets must be linked to various data elements found within GP clinical software systems. Data element labels and definitions vary considerably between GP clinical software systems. Harmonisation of the labels and definitions for these data elements would ensure that the GPRS is implemented in a standardised way across GP software systems in Australia. The scope of this work will be investigated during Phase 2 of this project, for consideration as a future project. A number of data elements were identified as being essential for linkage to GPRS content. These included: Reason for encounter/reason for visit. This group of data elements is used to describe the patient s reason for attending general practice, and may include signs and symptoms, patient diagnoses or other requests for care. Problem/diagnosis/reason for prescription. This group of data elements describes the issue being managed at the consultation as a result of the health provider s assessment of the patient s complaint. This may be described as a symptom, diagnosis, social problem or other ill-defined complaint. Personal past history and family history. Health interventions. Health interventions recorded may relate to diagnostic and therapeutic interventions being undertaken by a health provider at the current encounter, or may refer to procedures performed previously that continue to impact the management of the patient. Counselling, education and administrative procedures may also be included. Allergies. The clear and concise recording of allergies is vital to ensure that patients are appropriately managed. Reason for medication change. Terminology for data elements that are being addressed in other NEHTA projects, for example pathology ordering data elements, will not be included in the GPRS

317 nehta The general practice environment The relationship between codesets and information models used in general practice will impact on the content contained within the GPRS and how they might be implemented. Clinical information stored in these data elements may be used to populate other clinical functions available in GP clinical software, such as problem lists, health summaries and referrals Features of SNOMED CT AU to be included during GPRS development The GPRS must be derived from SNOMED CT-AU foundation reference sets that exclude non-human concepts, and incorporate only currently active terms. Further specification of applicable SNOMED CT-AU foundation reference sets for the GPRS will be provided with the design documentation. Users must have the ability to record either SNOMED CT AU preferred terms or synonyms. As previously stated in the GP recommendations, there is considerable variability in the language used by Australian GPs, and this must be maintained to ensure that users are able to record their preferred term. This will be dependent on the availability of synonyms in SNOMED CT AU. Attributes linked to SNOMED CT AU concepts should be made available to GPRS users to allow the user to include additional detail with the reference set content. Attributes seen as most useful included chronicity (e.g. acute, sub-acute, chronic), laterality and certainty of diagnosis. The relationship between codesets and information models used in general practice will impact on the content contained within the GPRS. For example, the concept of family history can be represented in either the information model as a field or data element to which diagnoses are attached, or in the terminology where family history and the disease are linked in either a precoordinated term or post-coordinated expression. Further examination of the codesets and terminologies currently used in general practice will assist in determining the extent to which contextual information should be included in the GPRS, and included with the GPRS design documentation Mapping recommendations There are two distinct sets of maps (or linkages) required in this project: 1. For GPRS development. This involves maps from the source codesets and terminologies identified in Section to SNOMED CT-AU, to inform the development of the GPRS and allow continued use of legacy data in existing systems to SNOMED CT-AU concepts. 2. Maps to international classifications. This will occur on completion of the GPRS, and involves classifying the reference sets content to the classifications identified. Figure 1 shows the two sets of links required to create the GPRS and links to classifications. The arrows indicate that mapping will only occur in one direction, from the source codesets and terminologies to SNOMED CT, then from the GPRS to international classifications

318 Project Plan & Requirements Specification General Practice Reference Set Project Figure 1: Directions of maps required during the GPRS project Two classifications have been identified as relevant for linkages from the GPRS. As shown in Table 4, nine software vendors currently have the ICPC-2 PLUS interface terminology incorporated into their GP clinical software system. The International Classification of Primary Care, Version 2 (ICPC-2) provides the structure for the ICPC-2 PLUS terminology, and the reporting facility within ICPC-2 PLUS utilises the ICPC-2 classification for data aggregation and reporting. ICPC-2 is the Australian standard for reporting data from general practice and patient self-reported data. Classifying the GPRS to ICPC-2 will aid the ongoing reporting of data from general practice at a variety of levels for the individual GP, practices, general practice divisions and across Australia. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM) is the Australian standard for reporting morbidity data from the acute care sector and for population health. It is used by at least one GP software vendor in Australia, primarily by their medical specialist user base. Classifying the GPRS to ICD-10-AM would also allow comparisons of general practice and acute care data. The International Classification of Functioning (ICF) was also mentioned during consultation as a possible classification to which the GPRS could be linked. Due to its current low clinical usage in Australia this work will not be undertaken during this project

319 nehta The general practice environment Table 4: Terminologies or codesets used currently by GP software vendors GP software vendor ACS Computing Best Practice Communicare Genie Solutions Global Health Terminology/codeset used Not applicable Best Practice termset ICPC-2 PLUS ICPC-2 PLUS/ICD-10-AM Not applicable HCN (Medical Director) 6 Docle termset 7 Houston Medical 8 Intrahealth 9 isoft (Practix) Medical Objects Medtech 10 NT Health Stat Health Zedmed ICPC-2 PLUS ICPC-2 PLUS ICPC-2 PLUS SNOMED CT ICPC-2 PLUS ICPC-2 PLUS ICPC-2 PLUS ICPC-2 PLUS Implementation of the GPRS and associated maps will be covered by NEHTA s compliance and conformance assessment (CCA) scheme. If available, conformance test specifications will be provided on release of the GPRS and associated maps To be consulted during Phase 2. To be consulted during Phase 2. To be consulted during Phase 2. To be consulted during Phase 2. To be consulted during Phase

320 Project Plan & Requirements Specification General Practice Reference Set Project Parallel classification development The International Classification of Primary Care is currently being revised from the second version to the third (ICPC-2 to ICPC-3). The Wonca International Classification Committee, who develop and maintain ICPC, estimates that ICPC-3 will be completed by The International Statistical Classification of Diseases and Related Health Problems (ICD) is undergoing a revision process at present, from Revision 10 of the classification (ICD-10) to Revision 11 (ICD-11). The World Health Organisation has estimated that ICD-11 will be ready for release by It should be noted that the maps from the GPRS to these classifications will need to be updated once the new classifications are released. However, due to the significant amount of time before the availability of the new versions, no benefit would be gained by deferring the development of maps to these classifications until the new versions are released Implementation recommendations There was a clear recommendation stated from both vendors and GPs that additional time and effort should not be required by users to enter data into their GP software systems using the GPRS, as compared against their current practises. Implementation of a system that required additional key-strokes or mouse-clicks was deemed unacceptable by both users and vendors, and vendors stated that they would not be willing to implement a system that was more onerous to their users than the codesets or terminologies they currently have included within their systems. Vendors also stated that the implementation of the GPRS and maps should not need a user to upgrade their computer hardware to manage SNOMED CT-AU. There was a clear need to constrain the size of the GPRS to a similar size as the codesets and terminologies currently used in general practice (which vendors reported as containing between 8,000 and 15,000 terms) to facilitate effective searching. However, there was also acknowledgement that GPs may receive clinical terms from external providers that would not fit within the scope of the GPRS. To ensure that users have the capacity to record such information using SNOMED CT-AU, it was recommended that a secondary search mechanism be implemented with the GPRS, whereby a user who does not find the term they need to use in the GPRS can search further within the relevant foundation reference set in SNOMED CT-AU to find their required term. Specific implementation recommendations relating to search and display mechanisms are still under discussion and require further input prior to being presented in this report. These will be included in the design documentation at the conclusion of Phase 2. Vendors said that the preference was to receive the GPRS and maps in either SNOMED CT format (Release Format 2), or as comma separated value (CSV) files. In the first instance, the GPRS and maps should be prepared and released using both formats. Vendors stated that they need a comprehensive set of implementation guidelines to assist them during their implementation of the reference sets. Users should also receive education during the implementation process. This could be provided through a general practice SNOMED CT-AU education program made available to each GP who uses the GPRS in their GP software system

321 nehta The general practice environment Prior to implementation the GPRS should be tested to ensure that it does contain the terms used in Australian general practice. As well as internal testing by the Family Medicine Research Centre, GP software vendors and potential users of the GPRS should be recruited to test the validity of the GPRS and maps. Tentative agreement from both vendors and users was received during consultation. Vendors stated that there may be a significant cost factor involved with testing, and may require funding to do this Maintenance requirements To ensure its ongoing usefulness and representativeness, the SNOMED CT-AU GPRS must be updated regularly. Maintaining the GPRS has three facets: 1. Adding concepts to the GPRS that are already present within SNOMED CT-AU. 2. Adding concepts or synonyms to SNOMED CT AU, then including these in the GPRS. 3. Maintaining the GPRS in line with changes to SNOMED CT (International Release) Codesets and terminologies currently used in general practice are maintained at varying time intervals. The codesets contained within both Medical Director and Best Practice are maintained on an ad hoc basis, with terms added on user request. ICPC-2 PLUS is updated on a three month cycle, with users requesting terms throughout the year. Updates are released in January, April, July and October. A release may be postponed if there are no user requests. The GPRS must be maintained at least every six months in line with the SNOMED CT AU release cycle. However, there was concern from vendors that six months between updates might be too long and additional updates may be required to guarantee that the GPRS contains all necessary content. Ways in which this could be handled include: Incorporating updates into the monthly AMT release cycle. Developing a system for allowing temporary codes to be used within a GP system until the following six monthly update. Creating an alternate release cycle for the reference set for the first few releases until new term requests become less frequent and can fit into the six monthly cycle. The frequency of the update cycle should be influenced by NEHTA s processes for updating SNOMED CT AU. For example, how NEHTA is planning to manage interim updates or patches for new conditions that arise between releases (e.g. the 2009 swine flu pandemic) will need to be taken into consideration. If NEHTA has processes in place that allow for interim updates to occur, any new terms added to the SNOMED CT AU core will need to be considered for inclusion in the GPRS. Specific responsibilities for maintenance will also need to be considered. A defined maintenance schedule should be devised during Phase 5 to ensure the ongoing representativeness of the reference set for use in Australian general practice. Maps from concepts in the GPRS to classifications must also be updated with each update to the GPRS

322 Project Plan & Requirements Specification General Practice Reference Set Project Other recommendations All vendors wishing to incorporate the GPRS into their GP clinical software must establish licensing arrangements with NEHTA. This involves obtaining two licences from the NCTIS: 1. SNOMED CT Affiliate Licence Agreement; and 2. Australian National Terminology Release Licence Agreement. The affiliate licence provides worldwide rights to use the SNOMED CT International Release. The national terminology release licence provides the right to use the SNOMED CT-AU, which includes the International Release along with all Australian developed terminology and documentation for use in Australian clinical software applications. No fee applies when these licences are exercised in Australia. Users who wish to utilise the maps from the GPRS to classifications (e.g. ICPC-2, ICD-10-AM) must be in possession of a licence for these classifications. At present, negotiations for an ICPC-2 national licence are underway. Completion of this licence prior to the GPRS and map release will allow all users to access the map between the GPRS and ICPC-2. Implementation of the GPRS and links to classifications should conform to NEHTA s compliance and conformance assessment (CCA) processes if available

323 nehta Preliminary project plan 4 Preliminary project plan 4.1 Overview The proposed approach for this project involves six phases. Each phase will build on the outcomes of the previous phases. At the end of each phase there will be a checkpoint to ensure that planned outcomes have been delivered and that the basis for proceeding is sound. In summary the six phases are: Phase 1: Requirements and analysis phase (current phase) Gathering and collating requirements from clinicians, vendors and other stakeholders, incorporating workshops with these groups. Preparation of a preliminary project plan and requirements specification (this document). Phase 2: Design GPRS Continuation of requirements gathering as needed. Preparation of design documents for the general practice reference set, and for the linkages needed to and from the GPRS. Preparation of the final project plan and requirements specification. Phase 3: Build GPRS Create and populate the GPRS. Create linkages between the GPRS and international classifications as defined in Phase 2. Development of an implementation pack for GP software vendors and educational material for end users. Phase 4: Validate and test GPRS Appropriate testing of GPRS content internally. Evaluation of GPRS content by GPs in clinical practice. Validating of links from the GPRS to classifications. Refinement and revision of the GPRS and links after testing. Phase 5: Finalise GPRS pack Define roles and responsibilities for ongoing maintenance of the GPRS and linkages to classifications. Design a maintenance process and schedule. Provision of the final project report. Phase 6: GPRS Early Adopter Program Deployment of the GPRS into selected GP practice and vendor packs. Support the implementation of the GPRS in selected GP software vendors and associated general practices. Provide education to selected vendors and GP practices. Post-implementation support for early adopters. Refinement of the GPRS and training from lessons learnt. Refinement of the implementation plan

324 Project Plan & Requirements Specification General Practice Reference Set Project 4.2 Indicative project delivery time frame Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Phase 1: Requirements and Analysis Phase 2: Design GPRS Phase 3: Build GPRS Phase 4: Validate and Test Phase 5: Finalise Package Phase 6: Early Adopter Program Figure 2: Indicative project delivery time frame 4.3 Steps for completion of the GPRS Phase 2: Design phase The following list specifies the design phase: 1. Identify coded data elements identified during the consultation process in Phase 1 to be populated in the GPRS. 2. Continue requirements gathering, as a result of gaps identified during consultation in Phase 1. Groups who may be consulted during Phase 2 include: Organisational stakeholders Vendors not consulted during Phase 1 Practice managers Practice nurses GPs/health workers in Aboriginal health settings The Royal Flying Doctor Service 3. Leverage evaluation and recommendation of mapping software undertaken by IHTSDO for linking the codesets and terminologies used in general practice to SNOMED CT-AU. Obtain appropriate licences for the mapping software. 4. Prepare design documents for the GPRS. These documents will include final decisions about the scope, number and size of the reference set. The structure of the reference set will also be finalised. Validation and testing mechanisms for the reference set and linkages will be developed in consultation with NEHTA

325 nehta Preliminary project plan Phase 3: Build phase The following list specifies the build phase: 1. Create and populate the initial version of the GPRS. General practice termsets currently in clinical use in Australia (identified and procured during Phases 1 and 2 as previously described) will be amalgamated, overlaps identified and rationalised, and inconsistencies resolved to form a single Australian termset of general practice terms. This amalgamated termset will then be used to create a database of terms and concepts linked by best fit to matching SNOMED CT-AU concepts and Fully Specified Names to create the SNOMED CT-AU GPRS. The methodology will incorporate techniques for computerised linking of each of the terms in the amalgamated preliminary set to SNOMED CT-AU concepts and descriptions, to identify the most closely related concept in SNOMED CT-AU (the best fit ) to match the term in the amalgamated termset. Where terms or concepts in the amalgamated termset do not have a matching SNOMED CT-AU synonym or concept, recommendations will be provided to NEHTA for inclusion of the synonym and/or concepts in SNOMED CT-AU through the NEHTA submission request process. 2. Create links between the GPRS and international classifications identified during Phases 1 and 2. The ICPC-2 PLUS terms (one of the source termsets used in the reference set development) are already classified to ICPC-2 and ICD-10-AM, so where appropriate this classification can be used to link that part of the GPRS to ICPC-2 and ICD-10-AM. The remainder of the GPRS may be classified to ICPC-2 and ICD-10-AM involving a combination of automated and manual methods of linking, similar to those specified in the methodology applied during the development of the reference set above. Due to the coarser granularity of the classification labels in ICPC-2 and ICD-10-AM, criteria for inclusion and exclusion within ICPC-2 rubrics will be utilised during the classification process. 3. Develop an implementation pack for GP software vendors and educational material for end users Phase 4: Validate and test phase Methods used to validate and test the reference set will be developed during Phases 2 and 3. The GPRS and linkages from the GPRS to classifications will be refined and revised after testing as needed Phase 5: Finalise GPRS pack The following list outlines the steps for the finalisation of the GPRS pack: 1. Define roles and responsibilities for ongoing maintenance of GPRS and links. 2. Design a maintenance process and schedule. 3. Provision of the final project report

326 Project Plan & Requirements Specification General Practice Reference Set Project Phase 6: GPRS Early Adopter Program The following list outlines the steps involved in implementing the GPRS Early Adopter Program: 1. Select suitable vendors for inclusion in the early adopter program and recruit GP practices willing to participate in the program. 2. Assist the selected vendors in the planning and implementation of the GPRS. 3. Provide education to the selected vendors and GP practices. 4. Support the implementation of the GPRS in selected vendor applications and practices. 5. Provide post-implementation support for early adopters. 6. Identify lessons learnt from implementation planning and deployment. 7. Refine the implementation plans and training program based on lessons learnt

327 nehta Preliminary project plan 4.4 Key deliverables The key deliverables from the GPRS project are tabulated below. Phase Deliverables Expected Delivery 11 Phase 1: Requirements and analysis (current phase) Phase 2: Design GPRS Phase 3: Build GPRS Phase 4: Validate and test GPRS Phase 5: Finalise GPRS pack Phase 6: GPRS Early Adopter Program Preliminary project plan and requirements specification (i.e. this document). Design specification. Final project plan and requirements specification. Initial version of GPRS content and populated reference set structures. Linkages between the reference set and international classifications. Test plans, scenarios and test results for testing of GPRS content. Validated GPRS. Validated links from the GPRS to classifications. Implementation pack for GP software vendors. Educational material and scripts for GPRS. Ongoing maintenance plan for GPRS and links. Early adopter vendor packs utilise GPRS. Early adopter GP practices utilise GPRS. Lessons learnt from implementations. Refined GPRS and training materials. Refined implementation plan Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q3 11 Indicative timeframe - refer to Section 4 Indicative project delivery time frame

328 Project Plan & Requirements Specification General Practice Reference Set Project 4.5 Project governance The governance for this project is the responsibility of NEHTA's National Clinical Terminology and Information Service (NCTIS) Management team with consultation and advice from the GPRS Support Group (GPRS-SG). The purpose of the GPRS-SG is to provide guidance in the establishment and maintenance of SNOMED CT terminology components that would be applicable in the primary care sector, specifically in general practice. Key project deliverables will be made available for review to the GPRS-SG. 4.6 Resources In addition to NEHTA and other external resources the GPRS project will also require some time from GPRS-SG members for review of key project deliverables and participation in periodic GPRS-SG meetings. Tooling resources required to deliver this project are: A tool to enable matching and linking terms from general practice termsets currently used in general practice to SNOMED CT-AU. There are two potential tools that could be used in this project: 1) the CSIRO Snapper tool; and 2) the tool developed by the Health Information Technologies Research Laboratory at the University of Sydney. Access to the IHTSDO workbench to prepare the GPRS for release. 4.7 Assumptions and dependencies Most assumptions and dependencies have been identified and documented within each of the stated requirements. Further assumptions and dependencies are: Timely establishment of the tooling requirements prior to the GPRS build phase. The final GPRS product could be constrained by the input and outcomes of the clinical engagement

329 nehta Preliminary project plan 4.8 Risks Insufficient buy-in from GP software vendors. This would lead to the GPRS and links not being implemented in GP software systems. Strategies to overcome this risk include: involvement of software vendors from the beginning of the project; ongoing consultation as the project progresses; development of a clear and understandable implementation pack; acknowledgement during development that the GPRS should not adversely affect end-users' use of their software; and availability of links from existing terminologies or codesets to the GPRS. Insufficient buy-in from GPs. Strategies to overcome this risk include: consultation with GPs throughout the life of the project; promotion of SNOMED CT-AU and the GPRS to GPs; ensuring co-operation from software vendors to implement the GPRS; and ensuring seamless transfer from existing systems to SNOMED CT-AU. Insufficient buy-in from stakeholder groups. Strategies to overcome this risk include: ongoing consultation with stakeholders throughout the project; and requesting co-operation from stakeholder groups to promote the project to end users. Parallel development and distribution of other general practice reference sets. The distribution of multiple GP reference sets of SNOMED CT will lead to a lack of standardisation of the SNOMED CT-AU concepts used by Australian GPs. Lack of clear conformance, compliance and accreditation (CCA) programs within NEHTA for the GPRS and linkages created during this project. The lack of clear conformance, compliance and accreditation could lead to lack of standardisation of the manner in which the GPRS and SNOMED CT-AU concepts are implemented in vendor packs

330 Project Plan & Requirements Specification General Practice Reference Set Project 5 Use cases 5.1 Explanation of use cases The use cases supplied below are indicative use cases only. The use cases illustrate the potential uses of the SNOMED CT-AU GPRS and maps from the reference sets to appropriate international classifications. They are indicative use cases only. Please note that these uses cases may not be representative of future implementation. The use cases will be presented to the GPRS-SG for feedback and indicative endorsement Patient recall A GP practice sets up a practice recall query, focussing on women who have not had a pap smear for more than two years. The GP clinical software system used in the practice uses the SNOMED CT-AU GPRS to record all pap smears undertaken. In the clinical software reporting facility, the SNOMED CT concept identifier for 'pap smear' (and all associated synonyms of 'pap smear') can be concatenated with the date of encounter at which the last pap smear was undertaken, allowing identification of patients who have not had a pap smear for a certain period of time. Patient details can be extracted using this report, and the practice can send patients recall letters as appropriate Coded data for the Australian Primary Care Collaboratives Program A GP practice that is involved in the Australian Primary Care Collaboratives Program needs to ensure that all the patients in their practice with chronic obstructive pulmonary disease (COPD) are able to be identified for inclusion in the program. All patient diagnoses are recorded using the GPRS in their GP clinical software system. The use of the GPRS allows GPs in the practice to describe synonyms of COPD using their preferred terms (alternative terms include 'chronic obstructive airways disease COAD', 'chronic airways limitation CAL', or 'emphysema'), without the practice dictating or recommending to their GPs which term they should use to describe the condition. SNOMED CT-AU includes most of these synonyms within the same concept, however 'emphysema' is not included within the COPD concept in the SNOMED CT International Release. The use of the map from the separate concepts in the GPRS to either ICPC-2 or ICD-10-AM will allow the practice to automatically group COPD (and its included synonyms) and the concept of emphysema for reporting purposes. In addition, if further research were undertaken into emphysema rates, these data could be directly sourced from records coded with the GPRS Research purposes A GP in a practice that uses the GPRS is taking part in a research study that requests de-identified data from medical records to be extracted from the GP s clinical software system. The research study requests that the GP sends all data from patients problem lists to determine the prevalence of diseases experienced by patients in general practice. The GP exports the requested information using SNOMED CT-AU codes, processes the information as required and sends it through to the research study. Receiving SNOMED CT-AU codes facilitates the analysis of data for the researchers, allowing them to easily aggregate data from multiple patients and practices into ICPC-2 for comparison with other GP-based studies

331 nehta Use cases Electronic communication using HL7 - referrals A GP with a clinical software system containing the GPRS and links to classifications is referring a patient to a specialist for consultation and further care. The GP clinical software system automatically constructs an HL7 structured referral document containing the SNOMED CT-AU coded concepts in the patient s health summary (e.g. diagnoses, past and family history, and previous procedures and treatments). This information is sent electronically to the specialist, whose clinical software is populated with this information, reducing the data input required by the specialist to populate their electronic record. The SNOMED CT-AU concepts are classified to both ICPC-2 and ICD-10-AM to assist data coding should the patient be admitted to hospital Electronic communication using HL7 discharge summaries A GP receives an HL7 structured electronic hospital discharge summary where diagnosis and procedure details are expressed as coded SNOMED CT-AU concepts. The presence of the GPRS and map to ICPC-2 enables these details to be automatically imported into the GP clinical software system and expressed within the GP s software in SNOMED CT concepts and associated synonym descriptions. If the SNOMED CT-AU concepts included in the discharge summary are not available within the GPRS, the GP can access a secondary search mechanism located within their GP clinical software. This secondary search is a broader search that allows the GP to search all relevant foundation reference sets within SNOMED CT-AU. If the most appropriate clinical concept cannot be found within SNOMED CT-AU after the secondary search the text description for the clinical concept should be displayed within the message. This results in discharge information being available to the GP as soon as the discharge summary is received, enabling the GP to commence post-discharge care planning immediately. The SNOMED CT-AU to ICPC-2 map provides linkage for audit of subsequent care and to best practice evidence and guidelines and for subsequent statistical use of data Decision support A patient visits a GP practice. The patient s regular GP is on holidays, so another GP at the practice sees the patient. The patient complains of worsening pain in the knee due to osteoarthritis. The patient has been taking paracetamol to manage the pain, but this is no longer providing enough relief, and the patient requests another form of medication to manage the pain. The GP decides to prescribe a non-steroidal anti-inflammatory drug (NSAID). On entering this medication into the record (using the Australian Medicines Terminology) a pop-up warning appears on the screen. The use of SNOMED CT-AU and Australian Medicines Terminology codes by the decision support provider within the drug-disease interaction system provides the link to determine the presence of such a warning. In this case, the presence of the SNOMED CT-AU coded concept of duodenal ulcer in the patient s record triggers the drug-disease interaction module stored in the GP clinical software system to warn the GP that the use of NSAIDs is contraindicated for patients with duodenal ulcers. Using the SNOMED CT-AU concept for duodenal ulcer, the GP searches the clinical decision support system within the software for duodenal ulcers to investigate appropriate treatments for the patient

332 Project Plan & Requirements Specification General Practice Reference Set Project Management of legacy data A GP uses the GPRS to record patient reasons for encounter and diagnoses within a GP clinical software system. Periodically the GP profiles her practice relative to the reasons that patients are presenting for care. The GP has a legacy profile of patient reasons for encounter and diagnoses that were previously recorded using ICPC-2 codes. The GP wishes to compare the current diagnosis profile of her patients against those documented in the past. The GP software vendor has incorporated the map from SNOMED CT-AU to ICPC-2 to support the translation from SNOMED CT-AU to ICPC-2 classification of clinical data allowing the GP to make the necessary comparison

333 nehta Stakeholders 6 Stakeholders This section identifies the key stakeholders and stakeholder groups for the GPRS project. A stakeholder matrix will be added to this document prior to the conclusion of Phase 2. The key stakeholders and stakeholder groups are: NEHTA (including NEHTA Clinical leads) and IHTSDO The Family Medicine Research Centre Organisational stakeholders: Aged Care IT Council Australian College of Rural and Remote Medicine Australian Association of Practice Managers Australian General Practice Network Australian Institute of Health and Welfare Australian Medical Association Australian Practice Nurses Association General Practice Registrars Association Medical Software Industry Association Australian Information Industry Association Royal Australian College of General Practitioners Rural Doctors of Australia Association The Improvement Foundation (responsible for the Australian Primary Care Collaboratives Program). GP software vendors: Best Practice Software Pty Ltd Communicare Systems Pty Ltd Genie Solutions Pty Ltd Health Communication Network Limited (HCN) Houston Medical Australia isoft Australia Intrahealth Limited Medtech Global Ltd Stat Health Systems Pty Ltd Northern Territory Department of Health and Families Zedmed Pty Ltd Royal Flying Doctor Service. General practice staff: General practitioners Practice Nurses and Practice Managers. Related vendors: Pen Computer Systems Pty Ltd Ocean Informatics Medical Objects

334 Project Plan & Requirements Specification General Practice Reference Set Project 7 Requirements overview 7.1 Purpose of providing requirements 7.2 Scope The purpose of gathering requirements is to provide the GPRS project team with an overview of the requirements for the development and release of the GPRS and maps and for the GPRS-SG to oversee the project and its outcomes. The requirements are intended to: facilitate the planning of tasks required for the design, development and implementation; and articulate the scope and objectives of the release to allow validation of the project deliverables. The document outlines the requirements for the GPRS product/project. Where relevant, each requirement contains: a title that identifies the topic it covers; a description; 7.3 Context an imperative that indicates the optionality of the requirement; the priority of the requirement; a fit criterion which serves as a baseline for assessing the requirement s implementation; considerations that provide contextual information that may influence the implementation of the requirement; and related and conflicting requirements. The GPRS will be the principal source of clinical coded data that Australian clinical IT systems will use to allow GPs and other supporting clinicians working in general practice to record, retrieve and process information in an electronic health record at the point of care. 7.4 About the requirements Source of requirements The requirements specified in this document were elicited from consultations with GPs, GP software vendors and NEHTA Clinical leads

335 nehta Requirements overview Identification The requirements in this document are uniquely identified using the following format: Figure 3: Requirement Identification Format Example The requirement types presented in this document are described in the table below. Table 5: Requirement Types Requirement Type BR FR NF Requirement Type Description Business Requirement: A business requirement defines the business needs expressed in terms of broad outcomes the business requires rather than specific functions the system may perform. Functional Requirement: A functional requirement defines a function of a software system or its component. A function is described as a set of inputs, the behaviour, and outputs. Non-functional Requirement: A non-functional requirement is a requirement that specifies criteria that can be used to judge the operation of a system, rather than specific behaviours

336 Project Plan & Requirements Specification General Practice Reference Set Project Wording The requirement wording in this document uses the keywords MUST, SHOULD, MAY, MUST NOT and SHOULD NOT and are to be interpreted as described in [IETF1997]. An excerpt of these definitions is shown in the following table. Table 6: Requirement Wording Keyword MUST SHOULD MAY MUST NOT SHOULD NOT Interpretation This word, or the terms REQUIRED or SHALL, means that the definition is an absolute requirement of the specification. This word, or the adjective RECOMMENDED, means that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course. This word, or the adjective OPTIONAL, means that an item is truly optional. One implementer may choose to include the item because a particular implementation requires it, or because the implementer determines that it enhances the implementation while another implementer may omit the same item. An implementation which does not include a particular option must be prepared to interoperate with another implementation which does include the option, perhaps with reduced functionality. In the same vein, an implementation which does include a particular option must be prepared to interoperate with another implementation which does not include the option (except of course, for the feature the option provides). This phrase, or the phrase SHALL NOT means that the definition is an absolute prohibition of the specification. This phrase, or the phrase NOT RECOMMENDED means that there may exist valid reasons in particular circumstances when the particular behaviour is acceptable or even useful, but the full implications should be understood and the case carefully weighed before implementing any behaviour described with this label

337 nehta Requirements overview Priority Each requirement is prioritised based upon its contribution to the development of the GPRS. A requirement is ranked using one of the following categories depending on its perceived impact on the efficiency and effectiveness of the GPRS. Highly Desirable Business Critical Low High Desirable Highly Desirable Figure 4: Requirement priorities Effectiveness Table 7: Requirements priorities Requirement Priority High Medium Low Requirement Priority Definition Business Critical: Requirements in this category are considered mandatory and are necessary for the effective and efficient operation of the GPRS. Highly Desirable: Requirements in this category are considered optional but would provide considerable efficiency or effectiveness benefit to users of the GPRS. Desirable: Requirements in this category are considered optional and non essential but would be improve the effectiveness of the GPRS

338 Project Plan & Requirements Specification General Practice Reference Set Project 8 Detailed requirements 8.1 Content requirements BR.GPRS.0001 Scope of content - [HIGH] Description The GPRS MUST be sufficiently comprehensive to cover clinical terms used in general practice Fit criteria Codesets and terminologies identified for inclusion in this process are: CATCH Australian Government Department of Health and Ageing. University of Adelaide GP termset (derived from Medical Director terms) University of Adelaide and Royal Australian College of General Practitioners. ICPC-2 PLUS University of Sydney. Best Practice termset Dr Frank Pyefinch from Best Practice has indicated that he is willing to provide the termset. Data held within the existing codesets or terminologies that cannot be mapped to SNOMED CT-AU will be considered on a case-by-case basis for inclusion in SNOMED CT-AU and subsequent mapping. The GPRS must be sufficiently comprehensive to allow the recording of attributes and contextual information, including: Finding site Laterality Chronicity Certainty of diagnosis Severity Considerations The GPRS will be maintained and refined over time to ensure that it contains all relevant general practice content. The GPRS should be appropriate for use within general practices regardless of practice geographic location or patient socio-economic status. The GPRS should incorporate all relevant terms used in Australian general practice. Some terms within existing codesets and terminologies may be deemed as inappropriate for inclusion in the national SNOMED CT-AU GPRS. When vendors are considering implementing the GPRS it would be advisable to consider the safe retirement of duplicate or erroneous terms from their systems. The GPRS should contain both SNOMED CT-AU concepts and synonyms to reflect that existing systems use differing semantics

339 nehta Detailed requirements BR.GPRS.0002 Navigational groupings - [HIGH] Description The values in the GPRS MUST be capable of being restricted based on the context of their usage Fit criterion Users SHALL be able to select clinical terms from the GPRS in support of: Reason for visit/reason for encounter (including symptoms and signs) Problem/diagnosis/reason for prescription Past and Family history Health interventions Allergies Reason for medication change Considerations Definitions for the navigational groupings listed above will be developed during Phase 2, and the names of groups may change after these have been finalised. The GPRS must be sufficiently complete to be used to populate health summaries, problem lists and referrals and linked to decision support systems BR.GPRS.0003 Exception handling - [HIGH] Description Users MUST be able to record clinical terms that do not exist within the GPRS Fit criterion There is a defined mechanism and/or supporting process for handling terms that do not exist within the GPRS Considerations Existing systems have their own exception processes, for example the inclusion of temporary codes. Some vendors have ad hoc release cycles with updates provided to users as needed. If a clinical term is not contained within the GPRS it should be possible to search for the term within SNOMED CT-AU. If the term is not available within SNOMED CT-AU, there should be a mechanism to request its inclusion. Nongeneral practice users of the GPRS (e.g. specialists) may also utilise an additional search to record clinical terms with greater granularity than present in the GPRS. Data received at the practice from other healthcare providers (e.g. hospital discharge summaries) may legitimately contain SNOMED CT concepts that have not been, nor should be, included in GPRS. If the vendor s system is populated with a wider SNOMED CT-AU dataset, the SNOMED CT concept received should be able to be uploaded and interpreted. If the vendor s system contains only the GPRS, the GP will need to add the text from the concept description as a free-text entry in their system

340 Project Plan & Requirements Specification General Practice Reference Set Project BR.GPRS.0004 Legacy data - [HIGH] Description The implementation of GPRS MUST enable the retention of legacy information Fit criteria Maps from existing codesets and terminologies to SNOMED CT-AU MUST be developed Considerations As stated in Requirement BR.GPRS.0001 (see 8.1.1), some coded content in current GP codesets and terminologies will not be mappable to SNOMED CT (anecdotally reported as up to 20% in some systems). This is low value data so it should not be added to SNOMED CT-AU. An alternative mechanism for managing this data needs to be developed BR.GPRS.0005 Links to classifications - [HIGH] Description The GPRS MUST be linked to relevant international classifications Fit criteria The GPRS MUST be classified to the International Classification of Primary Care, Version 2 (ICPC-2). The GPRS SHOULD be classified to the International Statistical Classification of Health Problems and Diseases, 10th edition, Australian modification (ICD-10-AM) Considerations None BR.GPRS.0006 Maintenance - [MEDIUM] Description Values contained within the GPRS SHOULD be kept up-to-date Fit criterion Users would always be able to select the latest values in their systems. Where an appropriate value is not available an exception process would be in place. Data updates subsequent to initial software implementation should be rolled out without the user having to re-install their software Considerations The maintenance schedule needs to be sufficiently regular to ensure that it is adequate for most users. In some cases it will not be sufficient, so there must be a mechanism for the GPs to enter terms that are not currently available. It is expected that if the vendors have not updated the GPRS within their software, complaints and issues about GPRS content will be handled by the NCTIS Helpdesk, and a process for handling these must be developed

341 nehta Detailed requirements Appropriate new content should be added to SNOMED CT-AU and then included in the GPRS. For systems that have SNOMED CT-AU as their base this would have to be updated prior to the updating of the GPRS. SNOMED Release Format 2 (RF2) allows for structural changes to reference sets. This means that in some cases data updates would require vendor software changes also. However, normally data updates should have no impact on operating software. 8.2 Implementation requirements BR.GPRS.0007 Release format - [HIGH] Description The GPRS MUST release all files in SNOMED CT AU Release format 2 (RF2) and fully comply with the SNOMED CT Release Format 2 Specifications Fit criterion All files are released in Release Format 2. The current release format for reference sets distributed by NEHTA via its SNOMED CT-AU product is described in [IHTS2009], where member reference sets are in the Attribute value reference set pattern. The GPRS as a member reference set needs to be SNOMED CT compliant to the IHTSDO s technical specifications for reference sets as described in [IHTS2009] Supporting materials Material SNOMED CT Release Format 2.0 Reference Set Specifications URL or appendix BR.GPRS.0008 Terminology viewers - [HIGH] The terminology viewers included in the SNOMED CT-AU product MUST allow viewing of the GPRS in addition to its current specification Fit criterion The terminology viewers for Windows operating systems and MacOS include the current version of all reference sets Considerations None

342 Project Plan & Requirements Specification General Practice Reference Set Project BR.GPRS.0009 Impact of implementation - [MEDIUM] Description Implementation of the GPRS SHOULD have a minimal impact on the users operation of a vendor s system Fit criteria Implementation of the GPRS SHOULD NOT necessitate process change for data entry. Implementation of the GPRS SHOULD NOT require hardware changes to the user s existing system. Implementation of the GPRS SHOULD NOT negatively impact a user s ability to enter clinical information into their system Considerations Some users may be using older versions of software and will require a more significant upgrade to their systems. If vendors choose to significantly alter/improve their business processes simultaneously with implementation of the GPRS, more significant system upgrades may be required with a greater potential impact on users. It is anticipated that the GPRS will contain between 8,000 and 15,000 terms BR.GPRS.0010 Usability - [MEDIUM] Description Terms within the GPRS SHOULD be fast and easy to find, and minimise the amount of typing at the point of entry Fit criterion The system SHOULD use a minimum number of mouse-clicks and keystrokes to search, select and save terms from the GPRS Considerations It is advisable to create a set of implementation guidelines for vendors giving advice on how best and most effectively to search, display and use terms within their software environment. Possible enhancements to vendor systems search and display mechanisms recommended include the use of wildcards for searching, highlighting words included in searches, keyword searches, use of logical operators or utilising synonyms and acronyms for searching BR.GPRS.0011 End user education - [HIGH] Description End users of the GPRS MUST receive education in its use Fit criterion Users of the vendor systems can competently find, select, record and interpret GPRS terms as appropriate for their role

343 nehta Detailed requirements Considerations All education must be provided in conjunction with the GP software vendor. Education may be provided in the form of: User guides Face-to-face training Online tutorials Context-sensitive help systems BR.GPRS.0012 Technical education - [HIGH] Description Implementers of the GPRS are REQUIRED to undergo appropriate training as to the usage and implementation of the GPRS Fit criterion All implementers of the GPRS are suitably informed as to the usage of the GPRS Considerations Technical education may be provided in the form of: Implementation guide Release notes Data structures specification Reference set specification. 8.3 Governance BR.GPRS GPRS licence - [HIGH] Description GPRS will be released under two licences: 1) the SNOMED-CT Affiliate Licence Agreement and 2) the Australian National Terminology Release Licence Agreement Fit criterion GPRS will be available to licence holders only Considerations Vendors must be licence holders. These licences are granted and maintained by NEHTA

344 Project Plan & Requirements Specification General Practice Reference Set Project BR.GPRS GPRS governance - [HIGH] Description There MUST be a governance structure in place that is responsible for the ongoing management of the GPRS Fit criterion An existing or new broad based governance structure that represents the interests of all of the stakeholders and controls the ongoing evolution and maintenance of the GPRS Considerations It will be important to define the governance structure for the GPRS at a national level to provide a forum for GPs and the other stakeholders that are identified above to direct the evolution of the GPRS. It is important that this structure is balanced and represents the interests of all users thereof. The governance structure will be responsible for clarifying any rules under which local decisions about the termset (for example, addition of locally used abbreviations) may be made BR.GPRD GPRS maintenance process - [HIGH] Description There MUST be a process for maintaining the ongoing development and maintenance of the GPRS that has been endorsed by the GPRS governing body Fit criterion The process provides representatives from all stakeholders across all the appropriate stakeholder groups with the means to request adding, changing or retiring terms Considerations Without formalising an approach for the GPRS s maintenance and ongoing development, the initially implemented and agreed reference set will be extended and amended in an ad hoc manner that, though serving the needs of local users does not serve the national health information agenda BR.GPRS Change management - [MEDIUM] Description The implementation of the GPRS SHOULD coincide with a change management initiative to ensure its successful adoption Fit criterion Vendors, clinical staff and IT staff in all the GPRS implementation areas will know what the GPRS is, what its implications on system performance and usability will be and what the benefits of its introduction will be

345 nehta Detailed requirements Considerations The implementation of the GPRS should be preceded by a change management process whereby its adoption is socialised across all of the stakeholders that are involved in the collection and use of the data. Vendors are largely satisfied with the status quo and the introduction of the GPRS will take place in an environment where the change is not driven by internally initiated business improvements and may result in lower business performance BR.GPRS.0017 Release cycle - [HIGH] Description The GPRS MUST be released at least every six months Fit criterion A release of GPRS is available on a regular and scheduled basis Considerations A regular and systematic refresh of the GPRS data set will assist vendors in coordinating the release of data to their customers. Incremental updates would ideally be made available to cover data deficiencies; however, these would not be practical if they relied on an updated SNOMED CT AU release

346 Project Plan & Requirements Specification General Practice Reference Set Project 9 References [AGDH2009] [BIND2007] [BRIT2005a] [BRIT2005b] [BRIT2009] [GPCG2000] [IBMC1997] [IETF1997] [IHTS2009] [IHTS2010] Australian Government Department of Health and Ageing 2008, GP super clinics, accessed 15 December 2009, < tent/pacd-gpsuperclinics>. Bindman AB, Forrest CB, Britt H, Crampton P & Majeed A 2007, 'Diagnostic scope of and exposure to primary care physicians in Australia, New Zealand, and the United States: cross sectional analysis of results from three national surveys, British Medical Journal, 334 (7606) Britt H, Miller GC, Knox S, Charles J, Pan Y & Henderson J 2005, General practice activity in Australia , General Practice Series No. 18. Cat. no. GEP 18, Australian Institute of Health and Welfare, Canberra. Miller G, Britt H & O'Halloran J 2005, General Practice EHR and data query minimum data set development, Phase 3 Development Grants, General Practice Computing Group, Melbourne. Britt H, Miller GC, Charles J, Henderson J, Bayram C & Pan Y 2009, General practice activity in Australia General Practice Series No. 25. Cat. no. GEP 25, Australian Institute of Health and Welfare, Canberra. General Practice Computing Group 2000, Final report of the General Practice Coding Jury, accessed 26 February 2010, < ocs/gp_coding_jury.pdf>. IBM Consulting Group 1997, Clinical and administrative general practice systems consultancy report. Produced for the Commonwealth Department of Health and Family Services, accessed 5 March 2010, < ocs/ibmfinal.pdf>. Internet Engineering Task Force 1997, Key words for use in RFCs to indicate requirement levels, accessed 2 March 2010, < International Health Terminology Standards Development Organization 2009, SNOMED CT RF2 reference set specifications, v0.6, IHTSDO, Copenhagen. International Health Terminology Standards Development Organization 2009, About IHTSDO, accessed 5 March 2010, < [KNOX2008] Knox SA, Harrison CM, Britt HC & Henderson JV 2008, 'Estimating prevalence of common chronic morbidities in Australia', Medical Journal of Australia, 189 (2) [MSIA2008] [SIMS2000] [STAN2005] Medical Software Industry Association, Discover MSIA, accessed 5 March 2010, < 25>. Simsion Bowles and Associates 2000, General practice data model and core data set project, Simsion Bowles and Associates, Sydney. Standards Australia 2005, AS The language of health concept representation, Standards Australia, Sydney

347 nehta References [STAN2008] [USNL2009] [WONC2002] Stanek J & Symon B 2008, Coding diagnosis in general practice using SNOMED CT, accessed 11 December 2009, < CT%20coding%20diagnosis%20in%20General%20Practice% 20FINAL.pdf>. United States National Library of Medicine 2009, The CORE problem list subset of SNOMED CT, accessed 14 December 2009, < et.html>. Wonca Classification Committee 2002, Wonca International dictionary for general/family practice, Manedsskrift for Praktisk Laegegerning, Copenhagen

348 Project Plan & Requirements Specification General Practice Reference Set Project Appendix A: Definitions, Acronyms and Abbreviations Term BEACH CATCH CHIME Classification Classifying Codeset Concept CTIRG Docle Definition 'Bettering the Evaluation and Care of Health' is a research study that collects information about general practice clinical activities in Australia. It is conducted by the Australian General Practice Statistics and Classification Centre, a collaborating unit of the Australian Institute of Health and Welfare and the University of Sydney. The Australian Classification and Terminology of Community Health is a coding/classification system which enables the capture, storage and retrieval of information about client/provider contact within a community-based health services environment; the issues or problems clients present with, the services provided to clients and resources used in providing those services. Custodianship of CATCH is jointly managed by the NCCH and the AIHW. The Community Health Information Management Enterprise is a joint venture between the NSW Health Department, ACT Health, Queensland Department of Health and the South Australian Department of Human Services, which commenced in The categorisation of relevant natural language for the purposes of systematic analysis. A classification, a type of aggregating terminology, is a logical system for the arrangement of knowledge [STAN2005]. To aggregate concepts from a domain into groups/classes for a purpose [WONC2002]. A list of defined concepts within a domain, with a unique representation (code) allocated to represent each concept [STAN2005]. A concept is a clinical data object within the SNOMED CT Concepts table and is identified by a unique and fixed numeric identifier. Clinical Terminology and Information Reference Group. Docle (Doctor Command Language) is a coding and classification system. It is incorporated into the Medical Director clinical software system

349 nehta Definitions, Acronyms and Abbreviations Term FMRC Foundation reference set GP GPRS ICD-10-AM ICPC-2 ICPC-2 PLUS IHTSDO Linking/links/linkages Mapping Definition The Family Medicine Research Centre is a centre of the University of Sydney located at Westmead Hospital. The FMRC has responsibility for the Australian primary care terminology ICPC-2 PLUS (classified to the International Classification of Primary Care, ICPC-2), and research and reporting on primary care data use through the Australian General Practice Statistics and Classification Centre. Reference sets that form the basis from which all NCTIS clinical and administrative content reference sets will be developed. General Practitioner. General Practice Reference Set. A group of reference sets for general practice based on SNOMED CT-AU concepts. The International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Australian Modification. This classification is maintained and distributed by the National Centre for Classification in Health. The International Classification of Primary Care, Version 2 is an official classification of the World Organisation of Family Doctors (Wonca) compiled by its International Classification Committee [WONC2002]. An interface terminology classified to ICPC-2, developed and distributed by the Family Medicine Research Centre. The International Health Terminology Standards Development Organisation owns and administers the rights to SNOMED CT and other health terminologies and related standards. The purpose of IHTSDO is to develop, maintain, promote and enable the uptake and correct use of its terminology products in health systems, services and products around the world [IHTS2010]. The establishment of a relation between a concept in one system and the most similar concept in another system (modified from [WONC2002]). A relationship between the code or term used to represent a health concept in one system, and the code or term that would be used to represent the same concept in another coding or terminology system [STAN2005]

350 Project Plan & Requirements Specification General Practice Reference Set Project Term Medical Software Industry Association NCTIS Reference set SNOMED CT SNOMED CT-AU Terminology (clinical) Termset Definition The MSIA represents the interests of the Australian commercial software industry which develops, supplies and services information management products and services for healthcare practitioners, healthcare service providers and healthcare organisations [MSIA2008]. The National Clinical Terminology and Information Service, NEHTA. A subset of SNOMED CT concepts that can be used in particular scenarios. Typically each reference set is used to represent a set of components for a specific purpose within a defined scope. Systematized Nomenclature of Medicine - Clinical Terms. A systematically organised computer processable collection of medical terminology covering most areas of clinical information such as diseases, findings, procedures, microorganisms, pharmaceuticals etc. It allows for a consistent way to index, store, retrieve and aggregate clinical data across specialties and sites of care. It also helps in organising the content of medical records, reducing variability in the way data is captured, encoded and used for clinical care of patients and research. Abbreviation for the SNOMED CT Australian Release. The component of health language used at the point of care for the purpose of clinical management of subject(s) of care [STAN2005]. Coded and non-coded lists of terms

351 Appendix 2: BEACH encounter form,

352 (a) 330

353 Appendix 3: ICPC-2 pager 331

354 332

355 333

356 Appendix 4: GP characteristics questionnaire,

357 Appendix 5: Letter of invitation to GPs attending GPRS workshop 335

358 336