Clinical Decision Support and Knowledge Management

Transcription

1 Clinical Decision Support and Knowledge Management Roberto A. Rocha, MD, PhD Sr. Corporate Manager Clinical Knowledge Management and Decision Support, Clinical Research and Development, Partners Healthcare System Lecturer in Medicine Division of General Internal Medicine and Primary Care, Department of Medicine, Brigham and Women s Hospital, Harvard Medical School Biomedical Informa/cs Course Marine Biological Laboratory in Woods Hole, MA - June,

2 Objec:ves Outline the main factors that the need for computerized Clinical Decision Support (CDS) and Clinical Knowledge Management (CKM) Describe the history and benefits of CDS systems Describe the main components of a CDS system Describe the different modali@es of CDS and their associated requirements Provide examples of CDS modali@es integrated with EHRs Describe the CKM processes required to create, deploy, disseminate, and maintain CDS interven@ons Describe the main components of a CKM system Provide examples of CKM tools Outline challenges and opportuni@es related to CDS & CKM

3 Expected Results As a result of par@cipa@ng in this ac@vity, learners will be able to: Explain uses and benefits of Clinical Decision Support (CDS) and Clinical Knowledge Management (CKM) Describe the main components of a CDS system Describe the different modali@es of CDS Describe CKM processes and associated tools Outline important challenges and opportuni@es related to CDS and CKM

4 Outline 1. Background History & Benefits 2. Clinical Decision Support (CDS) CDS (examples) and standards Components of a CDS system 3. Clinical Knowledge Management (CKM) Mo@va@on for CKM CKM Program: processes, people, and infrastructure 4. Challenges and opportuni:es

5 Background History of CDS Demonstrated benefits

6 Informa:on needs needs 47 physicians (self- reported) 269 raised during 409 visits» 2 ques@ons for every 3 pa@ents seen Answers not pursued 70% of Frequent barriers Pursued answers only 55% Doubt that an answer existed lack of usable informa@on Sources: human (informa@on consulta@on) and/or textbook (63%), electronic resource (16%) Unable to find answer in 28% Covell DG, Uman GC, Manning PR. Informa@on needs in office prac@ce: are they being met? Ann Intern Med Oct;103(4): Ely JW, Osheroff JA, Chambliss ML, Ebell MH, Rosenbaum ME. Answering physicians' clinical ques@ons: obstacles and poten@al solu@ons. J Am Med Inform Assoc Mar- Apr;12(2):

7 Informa:on explosion? 800,000 MEDLINE /PubMed Baseline Yearly Cita:on Count Totals 700, , , , ,000 Over 18 million total 50% since , , Sta@s@cal Reports on MEDLINE /PubMed Baseline Data

8 Survival of Systema:c Reviews Shojania KG, Sampson M, Ansari MT, Ji J, Douceoe S, Moher D. How quickly do reviews go out of date? A survival analysis. Ann Intern Med Aug 21;147(4):

9 Cri:cal diges:on of informa:on the expanding of into dimensions greater than can be traversed rapidly and efficiently is raising needs for and of needed Such and synopsis is costly in intellectual effort that is not well rewarded academically or commercially. Huth EJ. The explosion. Bull N Y Acad Med Jul- Aug;65(6):

10 Defini:ons (1) Medical/Clinical Decision Support System a computer program that provides reminders, advice or interpreta@on specific to a given pa@ent at a Wyao JC. Decision Support Systems. J R Soc Med 2000; 93: computer systems that provide the correct amount of relevant knowledge at the and context, ul@mately contribu@ng to improved clinical care and outcomes. Osheroff JA, Teich JM, Middleton B, Steen EB, Wright A, Detmer DE. A roadmap for na@onal ac@on on clinical decision support. J Am Med Inform Assoc Mar- Apr;14(2):141-5.

11 Defini:ons (2) Medical Decision Analysis (decision- making) Assist with the uncertain nature of medical Fundamental concepts include probability, and expected value decision making Make clinicians beoer decision makers need to be directly involved Typical How should I interpret new diagnos@c informa@on? How do I select the appropriate diagnos@c test? How do I choose among several risky treatments? H.C. Sox, M.A. Blao, M.C. Higgins, K.I. Marton. Medical Decision Making. Buoerworth- Heinemann, Stoneham, MA, 1988.

12 Evolu:on: 1960s and 1970s 1960s Early using digital computers, models (Bayes theorem), good accuracy, need for reliable and contextualized data sources; Medical decision making methods, theory 1970s Symbolic and reasoning methods, medical Intelligence (AI), need for knowledge engineering, expert systems Inadequacy of expert judgment, enhance data (large databases), superior accuracy when compared to experts Complexity of decision- models, teaching principles to physicians and health workers Shortliffe EH. Medical Knowledge and Decision Making. Methods in InformaEon in Medicine ,

13 Evolu:on: 1980s and 1990s 1980s of symbolic and models, theory of fuzzy sets, importance of instead of diagnosing (decision support) Personal computers, rapid and faster processing, graphical user interfaces, knowledge authoring tools, knowledge importance of proper 1990s Demise of stand- along model (expert system), with data management systems (clinical systems), Sosware and legal liability, ownership and maintenance of knowledge bases, standard formats for knowledge encoding and exchange, clinical and IT governance ( poli@cal challenges ) Shortliffe EH. Medical Knowledge and Decision Making. Methods in InformaEon in Medicine ,

14 Evolu:on of CDS Enthusiasm for CDS (and AI), research and new ideas Shortliffe EH. Medical Knowledge and Decision Making. Methods in InformaEon in Medicine , Successful CDS showing benefit, but limited Na@onal agendas (call to ac@on), safety and quality (errors, ADEs), roll out of Electronic Health Records (EHRs), Computer- base Provider Order Entry (CPOE), Electronic Prescribing (erx), and Personal Health Records (PHRs) Recognizing knowledge management as enabling CDS at scale Federal Health IT Strategic Plan (ONC): Rector, 1986 ( Defaults, excepeons and ambiguity in a medical knowledge representaeon system Med Inform (Lond). Oct- Dec;11(4): ) Government incen@ves to implement EHR systems with CDS

15 Regenstrief Medical Record System (RMRS) Clement McDonald, Marc Overhage, William Tierney, Paul Biondich, et al. Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 4: Regenstrief Medical

16 RMRS: Timeline (1) 1972: management of diabetes care 1976: results of the first randomized controlled studies rules; effect of reminders (as paper reports); no training effect non- perfectability of man (NEJM) Late 1970s: to a hospital system; rela@onal database; language for rule ( CARE ) 1980: larger study (reminders + literature) 410 protocols; confirming results; no interest in the suppor@ng literature already knew) 1984: results of a 2- year RCT 130 providers, 14,000 pa@ents, +50,000 visits; +140,000 paper reminders generated 1,490 rules; confirming results; greatest effect on preven@ve interven@ons; only 40-50% responded to reminders (did not see, inappropriate reminders due largely to missing data) Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 4: Regenstrief Medical Informa@cs

17 RMRS: Timeline (2) Mid 1980s: Medical Gopher order entry; and CDS Lab test ordering: previous results, test cost, likelihood of result CDS ordering: charge for med, allergy warnings, drug- drug, and drug- diagnosis corollary orders 1990s: Extensions to handle guidelines Problems with guidelines: vague terminology, omit branch points, data is not available, no for concurrent therapy and Physicians ignored most reminders about chronic disease management: intrusive, cost control emphasis, logic is too complex for discrete rules 2000s: Indianapolis Network for Care (INPC) Leveraging commitment to standards (HL7 and LOINC) Ongoing studies at a larger scale, involving mul@ple ins@tu@ons Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 4: Regenstrief Medical Informa@cs

18 Brigham Integrated Compu:ng System (BICS) David Bates, Jonathan Teich, Gilad Kuperman, John Glaser, et al. Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 5: CDS at Brigham and Women s Hospital

19 BICS: Timeline 1984: decision to develop clinical system; emphasis on decision support; complete hospital system (clinical, financial, and derived from systems developed at Beth Israel Hospital (Slack & Bleich MIIS system); implemented in MUMPS; independent since : electronic medical record system (Miniamb); free text notes 1993: computerized physician order entry (CPOE Glaser & Teich); embedded with decision support; front- end 1997: longitudinal medical record (LMR); now Web- based Teich JM, Glaser JP et al. The Brigham integrated system (BICS): advanced clinical systems in an academic hospital environment. Int J Med Inform Jun;54(3): Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 5: CDS at Brigham and Women s Hospital

20 BICS: Studies 1995: Adverse drug events (Leape et al. & Bates et al.) common (6.5/100 admissions); osen preventable (28%); osen serious (43%) 1997: Costs of ADEs (Bates et al.) $2,600 (2.2 days) per ADE or $5.6M/year at BWH; $4,700 (4.6 days) per preventable ADE or $2.8M/year at BWH ADEs major for CPOE Wrong dose, wrong choice, known allergy, wrong frequency, drug- drug etc. CPOE reduced serious error rate by 55% (Bates et al. 1998) Overall error rate fell 83% with CPOE (Bates et al. 1999) Display of lab test charges, redundant lab tests, corollary orders, radiology ordering, follow- up of abnormal results, among others Many remain: what best to deliver and how to deliver it Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 5: CDS at Brigham and Women s Hospital

21 Health Evalua:on through Logical Processing (HELP) Homer Warner, T. Allan Pryor, Reed Gardner, R. Scoo Evans, Peter Haug, et al. Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 6: CDS at LDS Hospital

22 HELP: Timeline (1) : design and of the HELP system; designed from the outset as a clinical system clinical data stored in a common database; most data stored in a coded format knowledge base organized as medical logic modules (MLMs); ranging simple rules to complex logic using data from mul@ple sources ability to data- drive the knowledge base; all data is inspected by the decision engine 1976: pharmacy applica@on; adverse drug events: drug- drug, allergy, laboratory, disease, dose, diet, and interval; alerts displayed to pharmacists (not a CPOE) MDs changed therapy for 77% of the alerts (Hulse et al.) 1975: interpreta@on of blood gas results 1980s: bedside char@ng by nurses; installa@on of bedside computers Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 6: CDS at LDS Hospital

23 HELP: Timeline (2) 1985: disease monitoring; hospital- acquired reportable diseases, resistant pathogens, in sterile body sites; hospital- wide surveillance 1990: monitor; appropriate based on culture and results 1989: 2- hours prior to incision; surgery schedule improvement from 40% to 96%; decreased wound (Classen et al) 1989: of high- risk for hospital- acquired (diagnosis) 1990: computerized laboratory system; life- threatening flashing yellow lights; paging nurses unaware of result (Tate el al.) 1990: blood ordering Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 6: CDS at LDS Hospital

24 HELP: Timeline (3) : protocols; treatment of acute respiratory distress syndrome (ARDS); distributed to other hospitals survival with computer protocol was 67%, compared to 33% (East et al.) 1991: adverse drug event monitor; events (lab results, serum drug levels, treatment of ADEs, physiologic signs) increased annual number of ADEs from 10 to over 500 (Classen & Evans et al.) 1992: of therapy; longer that 48 hours; significant cost savings 1994: agent assistance; regression models using accumulated data; suggest proper prior to culture results reduced inappropriate use, excessive dosages, number of ADEs caused by reduced cost (Evans et al.) Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapter 6: CDS at LDS Hospital

25 CDS: addi:onal evidence review of 70 studies (RCTs), up to 2003 the ability of CDS to improve clinical Focus on 15 CDS features (derived from literature) CDS improved in 68% of trials Key features (independent predictors) CDS as part of clinician workflow rather than just assessments CDS at and of decision making CDS triggered by computerized data analysis Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical using clinical decision support systems: a systema@c review of trials to iden@fy features cri@cal to success. BMJ. 2005;330(7494):765

26 CDS: value & impact Most profound impact of ambulatory CPOE arises with CDS Advanced CPOE systems cost as much as basic CPOE, but were projected to generate greater financial return Model projected reduc@on of more than 2 million adverse drug events (ADEs) annually with na@onwide implementa@on of ambulatory CPOE Annual savings of approximately $44 billion from reduced medica@on, radiology, laboratory, and ADE- related expenses Johnston D, Pan E, Middleton B, Walker J, Bates D. The Value of Computerized Provider Order Entry in Ambulatory SeSngs. Center for Informa@on Technology Leadership (CITL), 2003.

27 Clinical Decision Support (CDS) CDS Examples: content + process

28 CDS: modali:es 1. Reference knowledge selec:on and retrieval e.g., infobuoons, crawlers 2. Informa:on aggrega:on and presenta:on e.g., summaries, reports, dashboards 3. Data entry assistance e.g., forcing evidence- based templates for ordering and 4. Event monitors e.g., alerts, reminders, alarms 5. Care workflow assistance e.g., protocols, care pathways, guidelines 6. Descrip:ve or predic:ve modeling e.g., diagnosis, prognosis, treatment planning, treatment outcomes

29 Knowledge Lifecycle Generate Maintain Acquire Deploy Represent

30 CDS Modali:es: example 1 Health Maintenance Reminders Acquire

31 Reminders: overview Extensive literature about reminders studies results Specific example: Method that facilitates the authoring, discussion, review, and approval of reminders by clinicians (expert panels) Promotes and Virtual sessions are recorded (retrievable)

32 Reminders

33 Structure of rules IF THEN Risk Group Trigger in risk group} AND State has triggering condi:on} AND {NOT previously suppressed by user} AND {NOT suppressed by another reminder} AND {Ac:ve at specific {Reminder message and/or Display Regier R, Gurjar R, Rocha RA. A clinical rule editor in an electronic medical record setting: development, design, and implementation. AMIA Annu Symp Proc Nov 14;2009:

34 Narra:ve specifica:on Risk Group Triggers Display

35 New & reminders

36 Focused with hyperlinks to evidence

37 Opinions and for discussion

38 Logic of each Reminder is made available through the KM Portal

39 Example 1: Reminders Emphasis on acquisi:on & review Virtual spaces Overcome knowledge and boolenecks Stakeholder involvement during different lifecycle phases and use Direct involvement of domain experts (panels) Knowledge content accessible and maintained with a detailed audit trail Access to knowledge assets using open portal

40 CDS Modali:es: example 2 Problem List Infobudons Represent

41 Infobudons: overview Extensive literature about infobuoons Successfully balloted HL7 dras standard Specific example: Method that facilitates the retrieval and of common guidelines by physicians at the point of care Infobuoons linked to problems in an EHR Each infobuoon displays a list of common ques@ons that can be answered by the guideline Poon SK, Rocha RA, De Fiol G. Rapid Answer Retrieval from Clinical Practice Guidelines at the Point of Care. 19th IEEE International Symposium on Computer-Based Medical Systems, pages

42 Problem List Infobuoons

43 Poon SK, Rocha RA, De Fiol G. Rapid Answer Retrieval from Clinical Practice Guidelines at the Point of Care. 19th IEEE International Symposium on Computer-Based Medical Systems, pages driven Infobuoons

44 Infobudons: knowledge model Poon SK, Rocha RA, De Fiol G. Rapid Answer Retrieval from Clinical Practice Guidelines at the Point of Care. 19th IEEE International Symposium on Computer-Based Medical Systems, pages

45 Infobudons: requirements EHR with coded Problem List Problem List concepts: terminology or Problem List records: structured Models (previous slide) ranking, classes, answer sources Guideline metadata Indexed with same codes used by EHR Problem List Guideline tagging Tagged with same codes used by

46 Example 2: Infobudons Emphasis on Modeling & Enhanced retrieval based on a ques@on- answer paradigm Avoid knowledge modeling and data availability roadblocks related to ac@onable/executable CDS Leverage widespread availability of authorita@ve reference content (guidelines) Maximize the usefulness of these documents Provides a passive alterna@ve to CDS, without interfering with clinician workflow Rela@vely simple to implement (infobuvon manager) Cumbersome tagging of reference documents

47 CDS Modali:es: example 3 CPOE Inpa@ent Order sets Maintain

48 Order Sets: overview Extensive literature about order sets for CPOE success, but cumbersome to create and maintain Specific example: Method that enables refinement of order sets using tracking data of prescribers with order sets are recorded data is analyzed and presented back to order set authors Hulse NC, Del Fiol G, Bradshaw RL, Roemer LK, Rocha RA. Towards an on-demand peer feedback system for a clinical knowledge base: a case study with order sets. J Biomed Inform Feb;41(1):

49 changes to order sets Select values Add new orders Check or Uncheck Enter values Hulse NC, Del Fiol G, Bradshaw RL, Roemer LK, Rocha RA. Towards an on-demand peer feedback system for a clinical knowledge base: a case study with order sets. J Biomed Inform Feb;41(1):

50 Mutually exclusive orders (pre- Items from a pull- down menu (sequence) Report u@liza@on using authoring tool and/or CPOE UI Hulse NC, Del Fiol G, Bradshaw RL, Roemer LK, Rocha RA. Towards an on-demand peer feedback system for a clinical knowledge base: a case study with order sets. J Biomed Inform Feb;41(1):

51 Comprehensive to improve order set Hulse NC, Del Fiol G, Bradshaw RL, Roemer LK, Rocha RA. Towards an on-demand peer feedback system for a clinical knowledge base: a case study with order sets. J Biomed Inform Feb;41(1):

52 Example 3: Order sets Emphasis on maintenance (CQI) Detailed tracking providing aggregated end- user feedback (constantly updated) Refinements based on how knowledge is used Reduce the cost and increase efficiency of knowledge maintenance Enable (passive) end- user involvement Avoid liability due to lack of coverage or update for Monitor quality, safety, and business drivers requires oversight provided by expert panels

53 CDS Modali:es: example 4 Disease Management: Care Pathway Deploy

54 Care Pathway: overview Extensive literature about guidelines/ pathways for disease management, but very difficult to computerize (research prototypes) Specific example: Method that implements EHR Smart Forms to integrate of CDS Data and Ordering guidance and tracking Schnipper JL, Linder JA, Palchuk MB, Einbinder JS, Li Q, Postilnik A, Middleton B. "Smart Forms" in an Electronic Medical Record: documentation-based clinical decision support to improve disease management. J Am Med Inform Assoc Jul-Aug;15(4):

55 Smart Forms (1) View: Data Display Documenta:on Assessment, Orders, and Plan Assessment and generated from rules engine Lipids platelet therapy Blood pressure Glucose control Microalbuminuria Smoking Weight Eye and foot

56 Smart Forms (2) Medica:on Orders Lab Orders Referrals Handouts/Educa:on

57 Example 4: CDS- enabled workflow Detailed structured and coded data authoring and maintenance of protocols and workflows (complex knowledge) Able to merge overlapping protocols and workflows diseases, various user roles, of care, Able to rollback triggered and revise context, including and protocol states Proper handling of errors and uncertainty data and workflow Detect unexpected and fail gracefully Performance

58 CDS: implementa:on strategies CDS modality Implementa:on Strategies 1. Reference knowledge and retrieval Reference 2. and 3. Data entry assistance 4. Event monitors 5. Care workflow assistance 6. or modeling Complexity Cost Ac:onable Executable Availability Maintainability

59 Clinical Decision Support (CDS) Components of a CDS system CDS standards

60 Clinical Data! Data are the diamonds of medical Computer systems come and go Data is forever Or at least it should be The data you have constrains what you can do with decision support Be very conscious of these limits Do not assume you can capture data that you don t have Slide from Clem McDonald, MD

61 Haug PJ, Rocha BH, Evans RS. Decision support in medicine: lessons from the HELP system. Int J Med Inform Mar;69(2-3):

62 Basic model User Interface Pa:ent database Inference Engine Knowledge Base Editor

63 CDS Rules Manager ( Event Monitor ) Rocha R.A,Bradshaw R.L., Hulse N.C., and Rocha B.H.S.C. The clinical knowledge management infrastructure of Intermountain Healthcare. In: Clinical Decision Support: The road ahead, RA. Greenes (ed.). Academic Press, Boston, 2007, pp

64 CDS: standard representa:on formats CDS modality 1. Reference knowledge and retrieval 2. and 3. Data entry assistance 4. Event monitors 5. Care workflow assistance 6. or modeling Standard Formats Guideline Elements Model (GEM): ASTM Context- aware Info Retrieval (Infobuoon): HL7 (dras) Clinical Document Architecture (CDA): HL7 Quality Measures (emeasure): HL7 (dras) Order sets: HL7 (in progress) Arden Syntax for Medical Logic Systems: HL7 GELLO - A Common Expression Language: HL7 Decision Support Services: HL7 (dras) Virtual Medical Record: HL7 (in progress)

65 Clinical Knowledge Management (CKM) for CKM CKM Program

66 Implica:ons of CDS strategy Knowledge Management Development of CDS content Standard CDS content formats CDS content available for EHR use Implement EHRs with CDS Acquisi:on Representa:on Dissemina:on Deployment

67 Desirable CDS Features CKM Knowledge is based on the best evidence available Knowledge covers problem in detail allow problem solving, advice, Knowledge can be readily updated by a clinician without unexpected effects Knowledge base provides links to related local and Internet material lifelong learning Most pa@ent data drawn from exis@ng sources ease of use System (knowledge) performance is validated against suitable gold standard Demonstrated prac@ce or outcomes improvements in rigorous study Clinician always in control Receive advice, browse the knowledge base, get help and explana@ons, try out what- if scenarios, and obtain a cri@que of the pa@ent management plan System is easy to access for example via the Web Modified from Wyatt JC. Decision Support Systems. J R Soc Med 2000;93:

68 CKM: mo:va:on of knowledge (explosion) towards clinical Complex decision making process demanding computerized support Distributed care delivery processes (fragmented) Extensive knowledge is needed beyond boundaries Learning leading to care and stewardship Global trends towards knowledge Consumers constantly seeking knowledge (empowerment) Shared responsibility only possible with proper understanding Knowledge content maintainability (long- term) Content diversity and makes

69 Engineering vs. Deployment CDS developers Knowledge Engineering CDS consumers Crea:on/Revision Review/Approval Configura:on/Tes:ng Deployment/Valida:on Evalua:on/Monitoring Knowledge Deployment

70 CKM: deployment models Knowledge Content only Knowledge Services + Content Import Integrate Update Configure EHR with CDS EHR with CDS Configure Both require content localiza:on (configura:on)

71 CKM: concurrent lifecycles Import Update Configure Dic:onaries Generate Maintain Acquire Dic:onaries Deploy Represent Integrate Rules EHR with CDS Configure Maintain Generate Rules Acquire EHR with CDS Deploy Represent Generate Generate Maintain Acquire Order Sets Deploy Represent Import Update Configure Monographs EHR with CDS Maintain Acquire Templates Deploy Represent Import Update Configure Templates EHR with CDS Maintain Generate Reports Acquire Generate Maintain Acquire Workflows Integrate Guidelines EHR with CDS Configure Maintain Generate Protocols Acquire Deploy Represent Deploy Represent Deploy Represent

72 Strategic Partners Enable all knowledge content to be accessible, updatable, and maintained with an audit trail Reduce the cost and increase efficiency of both design and implementa@on maintenance Enable stakeholder involvement in the design process to support effec@ve adop@on and use Ensure alignment with quality, safety, and opera@ng business drivers (HPM, Joint Commission, etc.) Avoid poten@al liability of making incorrect or incomplete recommenda@ons due to lack of coverage or update

73 CKM: program components Personnel Domain Experts Framework Knowledge Engineers Governance Processes Assets Informa:on Modelers Terminology Engineers Lifecycle Processes Sonware Plaoorm Knowledge Repositories Logical Data Templates Concept Dic:onaries

74 Clinical Content Commidee and Sponsors Stewardship of Content Safety Quality Disease Management Trend Management Primary Care Disease Areas Pharmacotherapy Imaging Studies SME Groups Adult, Geriatrics, Pediatrics, Women s Health CAD/CHF, Diabetes, Heme- Onc, Asthma, ID/HIV, Nephrology, Psych Medica:on Knowledge Commidee PCHI P&T MGH ROE BWH Precipio Knowledge Engineers Tools Knowledge Repositories

75 CKM: staffing challenges is a lengthy process Technology professionals with clinical training (exposure) Partnerships with local academic programs Training is quite intense and (6-18 months) Processes, domains, mul@tude of systems, KM Framework Informa@cs Principles: external courses and internal mentoring Reten@on can be problema@c Uncommon skills (differen@a@on) Compe@@on with similar organiza@ons (and EHR vendors) Crea@on of specific job families (interrelated) Knowledge and Terminology Engineers Clinical Subject Maoer Experts Informa@cians

76 CKM: sonware plaoorm requirements Enable consistent lifecycle and Proper handling of dependencies of reference sources (with control) Enable authoring and Promote modularity and reuse Extensible and intelligent Maintain assets with meta- knowledge Leveraging technologies Built- in monitoring Built- in (CQI/Discovery)

77 CKM: sonware infrastructure View Edit Publish Analyze Lifecycle Management Collabora/on Management Knowledge Repositories Logical Data Templates Concept Dic:onaries Store Import Map Archive

78 Challenges & Opportuni:es CDS and CKM

79 CDS: important lessons compared RMRS BICS HELP User is right; data gaps - override alerts and reminders Workflow CDS can become overwhelming Con@nuous user input, collabora@on, and feedback Prospec@ve evalua:on studies o Importance of standards: interoperability o Speed is everything An@cipate user needs Workflow o Details are important o Stopping vs. changing direc@on Simple interven:ons Data is expensive Prospec@ve evalua:on studies Knowledge must be managed and maintained Greenes RA, editor. Clinical decision support: the road ahead. Academic Press, Chapters 4 (RMRS), 5 (BICS), and 6 (HELP) Data is cri@cal Knowledge is a team effort Workflow o Proper tes:ng before Simple interven:ons Evidence- based and matching local prac@ces Knowledge has to be periodically reviewed Ease of use Evalua:on is difficult Cost- effec:ve to implement and maintain

80 Healthcare IT: Meaningful Use HIT-Enabled Health Reform Meaningful Use Criteria HITECH Policies 2011 Meaningful Use Criteria (Capture/share data) 2013 Meaningful Use Criteria (Advanced care processes with decision support) 2015 Meaningful Use Criteria (Improved Outcomes) Diagram from: Tang & Mostashari (chairs) et al., Meaningful Use Workgroup Presentation. HIT Policy Committee, June 16, 2009.

81 Quality improvement with EHR use Study looked at the quality of care delivered in ambulatory and of EHR use: Survey of physicians of EHR (Massachuseos) 137 physicians using an EHR; average of 4.8 years Claims data quality of care as indicated by widely used quality measures Healthcare Data & Set (HEDIS): Breast cancer screening, HbA1c LDL screening, Well- child visits, No between of using an EHR and performance with respect to quality of care Intensifying the use of key EHR features, such as clinical decision support, may be needed to realize quality improvement from EHRs Zhou et al. The relationship between Electronic Health Record Use and Quality of Care. J Am Med Inform Assoc. 2009;16:

82 CPOE with advanced CDS 62 hospitals volunteered to assess CDS applied to ordering simulated test orders likely to cause serious harm entered in local CPOE (8 vendors) Metzger J, Welebob E, Bates DW, Lipsitz S, Classen DC. Mixed results in the safety performance of computerized physician order entry. Health Aff (Millwood) Apr;29(4):

83 CKM: implementa:on challenges (1) Clinical governance and stewardship is poorly defined Clinical strategy does not drive KM nor is informed by KM principles for strategic not or planned Projects and resources defined in with other IT efforts not aligned with clinical quality and safety Inadequate and of KM strategic Cost of not having knowledge is not frequently considered Deployment of knowledge assets is and inconsistent Domain experts (clinicians) frequently unavailable; limited commitment Processes for configuring and ve ng knowledge are not explicitly defined Lack of a systemic view promotes overlapping efforts (varia@on) Modified from Tonya Hongsermeier, MD, MBA Partners/CIRD

84 CKM: implementa:on challenges (2) Sosware tools to create and maintain knowledge are inadequate Knowledge once deployed for use is not easily accessible ( locked ) Tools frequently ignore content dependencies and lifecycle requirements (subsequent updates) Maintenance of knowledge assets is an aserthought Long- term commitment to content maintenance is underes@mated Liability resul@ng from outdated or incorrect recommenda@ons not recognized Analy@c data regarding impact on clinical processes and outcomes is generally not available Modified from Tonya Hongsermeier, MD, MBA Partners/CIRD

85 Personalized Medicine Greatly expanded space 1920s: 1 leukemia & 1 lymphoma 1940s: 3 leukemia & 2 lymphoma Today: 38 leukemia & 51 lymphoma (outdated?) Greatly reduced (targeted) therapeu@c space Blockbusters (e.g., atorvasta@n, sildenafil) Niche busters (e.g., ima@nib magic bullet to cure cancer) Orphans (e.g., imiglucerase Gaucher s disease) Modified from Michael G. Kahn MD, PhD University of Colorado

86 Test for gene:c differences

87 Knowledge Sharing Clinical Decision Support Consor:um Goal of the CDS is to assess, define, demonstrate, and evaluate best for knowledge management and clinical decision support in healthcare technology at scale across ambulatory care se ngs and EHR technology pla}orms. hop://

88 CDSC Portal hdp://cdsportal.partners.org

89 Acknowledgements Blackford Middleton Tonya Hongsermeier Saverio Maviglia Beatriz Rocha CIRD/KM Team at Partners Stanley Huff David Burton KB Team at Intermountain

90 Addi:onal readings Book Clinical decision support: the road ahead Greenes RA, editor. Academic Press, Paper Ten commandments for clinical decision support: making the of evidence- based medicine a reality. Bates DW, Kuperman GJ, Wang S, Gandhi T, Kioler A, Volk L, Spurr C, Khorasani R, Tanasijevic M, Middleton B. J Am Med Inform Assoc Nov- Dec;10(6): (PMID: ) Paper A roadmap for na@onal ac@on on clinical decision support. Osheroff JA, Teich JM, Middleton B, Steen EB, Wright A, Detmer DE. J Am Med Inform Assoc Jul- Aug;13(4): (PMID: ) Paper Just- delivery comes to knowledge management. Davenport TH, Glaser J. Harv Bus Rev Jul;80(7):107-11, 126. (PMID: ) Paper Using commercial knowledge bases for clinical decision support: opportuni@es, hurdles, and recommenda@ons. Kuperman GJ, Reichley RM, Bailey TC. J Am Med Inform Assoc Mar- Apr;14(2): (PMID: ) Paper Predic@ve data mining in clinical medicine: current issues and guidelines. Bellazzi R, Zupan B. Int J Med Inform Feb;77(2):81-97 (PMID: ) Web site Open Clinical (UK): hop://

91 Thank you! Roberto A. Rocha, MD, PhD!