Health IT for empowering citizens and health professionals

Health IT for empowering citizens and health professionals Ilkka Kunnamo, MD, PhD, Chief Editor, EBMeDS Duodecim Medical Publications Ltd Adjunct Professor of General Practice, University of Helsinki Disclosure: I am a salaried employee of Duodecim Medical Publications Ltd., the company that develops and and licenses EBM Guidelines and the EBMeDS decision support service WICC 14.9.2016

Guideline development by Duodecim, the scientific society of Finnish physicians Idea of EBM Guidelines emerged 1987 Pilot electronic version on floppy disks 1989 CD-ROM 1991 Health Internet portal for professionals 2000 Mobile guidelines for smart phones 2001 Health internet portal for citizens 2007 Computerized clinical decision support 2008 rules (EBMeDS) Electronic Health Coaching for citizens 2010 Translations of EBM Guidelines in English, German, Russian, Estonian, Hungarian, Slovenian, Dutch, French, Turkish, Italian

Terveysportti (www.terveysportti.fi) Over 40 million articles opened in 2015 by health care professionals in Duodecim s portal Uusi hakutoiminto

Evidence chain: Guideline Evidence Summary Cochrane review

NICE accreditation Evidence-based medicine methodology

Knowledge In 2012, the annual number of new scientific medical articles exceeded one million Gillam M & al. 2014 (digitalhealthdesign.com)

Medical knowledge is organized into PICOs Patient group (characteristics of the patient + setting) Intervention Comparator (intervention) Outcome

Medical knowledge is organized into PICOs Patient group (characteristics of the patient + setting) Intervention Comparator (intervention) Outcome Computers can process structured evidence: each PICO component must be coded

Evidence summaries and recommendations are authored in MAGICApp

Ann, 35 years Cholesterol 6.8 Treating high cholesterol with statin reduces Ann s relative risk of stroke, myocardial infarction or death by 25 % Tim, 55 years Cholesterol 6.8 Treating high cholesterol with statin reduces Tim s relative risk of stroke, myocardial infarction or death by 25 % Andres Rodriguez Dreamstime Stock Photos Alamar Dreamstime Stock Photos

The absolute risk of stroke or myocardial infarction is estimated on the basis of 9 risk factors

Fewer than 1 out of 140 people like Ann would have any of those events in the next 10 years. At least 600 people like Ann would need to take a statin for 10 years to prevent those events in one of them (NNT = 600)

More than 1 out of 2 people like Tim would have any of those events in the next 10 years. 6 people like Tim would need to take a statin for 10 years to prevent those events in one of them (NNT = 6)

What (CDS is about) Match individual patient data with evidencebased recommendations that would benefit the patient Software providing actionable guidance automatically at the right time at the right point in the workflow

Case history A 40-year-old man visits the doctor because of a hernia. The decision support system automatically shows a reminder that the latest data on blood pressure was from 3 years ago: 185/100. The doctor measures the blood pressure, which is now 220/110. Without the reminder the high blood pressure would have remained undetected, and the patient would not have received treatment

A comprehensive clinical decision support service with accredited evidence-based methodology Based on guidelines and systematic revies (Cochrane reviews) Easy to integrate with electronic health records Available in 10 languages

Demo of the EBMeDS service http://www.ebmeds.org/dev/dssscripts/dsscli ent.htm?clientskin=1&serverskin=1&nat=gb&l an=en Open the last profile from the pull-down menu and click Submit

Drug orders guided by clinical decision support (under development) The patient s diagnosis and need for care (desired outcomes) are first entered. A list of both drugs and non-pharmacologic interventions becomes available for which there is evidence of effectiveness. The drugs are categorized to first or second choice drugs for the given indication (and patient group). The CDS applications tags the drugs on the list automatically with the following characteristics and knowledge: drugs which the patient is already using drugs that are contraindicated or the dose needs to be reduced because of renal insufficiency drugs (of the same class) that have caused adverse effects to the patient in the past and had to be discontinued drugs that have a significant interaction with a drug which the patient is currently using drugs that are unsuitable because of an abnormal laboratory test result drugs that are contraindicated or the dose needs to be reduced because of liver dysfunction drugs that are not recommended because of the patient s age drugs that are unsuitable because the patient is pregnant or lactating drugs that are unsuitable because of pharmacogenetic reasons drugs that significantly increase the cumulative risk of adverse effects of the patient s current medication drugs for which the balance of benefits and harms is questionable or unfavourable

Using a computer-generated diagnosis-specific summary reduced the time needed to retrieve all relevant data from the electronic health record from 5.5 to 1.7 minutes saved 57 mouse clicks Richelle J. Koopman et al. Annals of Family Medicine 2011;9 (no 5)

How are health care information systems built? EHR 1 - Data EHR 2 - Data EHR = electronic health record PHR 1 - Data PHR 1 - Data PHR = personal health record Device 1 Device 2

Wearable devices

How are health care information systems built? EHR 1 - Data PHR 1 - Data EHR 2 - Data PHR 1 - Data One patient one record Data - Recorded by professionals - Recorded by patients/citizens - Collected by devices Device 1 Device 2

Summary medical record Problems/diagnoses Medication Test results Measurements (e.g. blood pressure) Risk factors (e.g. smoking) Procedures Treatment plan Functioning in standardized, coded format

Patients recording their own data Symptom history and monitoring Family history Blood pressure, weight, height Peak expiratory flow Blood glucose ECG Pain intensity Functioning and checking their current medication list diagnosis list

Who are the users of the health record?

Who are the users of the health record? 1. Patients and their family members 2. Health professionals other than doctors 3. Doctors

How are health care information systems built? Finland: national ehealth archive Applications Data - Recorded by professionals - Recorded by patients/citizens - Collected by devices EHR 1 PHR 1 EHR 2 App 1 PHR 2 App 2

What is the optimum level of granularity in the coding of data Low for viewing High for processing by computer and assisting in decisions

What is the optimum level of granularity in the coding of data Low for viewing High for processing by computer and assisting in decisions A hierarchic coding system can meet both needs Different users focus on different levels Labels (names) of concepts may be different for different users

Big data how many data elements in health care? Future: millions (billions?) Now: about 100 000 Nigam Shah (Stanford) 2013

GenBank contents: 200 billion = 200 000 000 000 base pairs in August 2015 http://www.ncbi.nlm.nih.gov/genbank/ Human genome sequenced The explosion of genome data started in the 2000 s 2000 2003 2006 2008 The doubling rate of genome data was 18 months in 2008

Health data of the individual EHR PHR User interface Use of genomic data Genome of the individual Query interface Validated associations Directory or data repository Biobank 1 Decision support International database of genomic associations Biobank 2

ehealth and esocial strategy 2020 - Information to support well-being and service renewal 1. Citizens as service users - doing it yourself 2. Professionals - smart systems for capable users 3. Service system - effective utilisation of limited resources 4. Refinement of information and knowledge management - knowledge-based management 5. Steering and co-operation - from soloist to harmony 6. Infostructure - solid foundation

Types and volume of services (National ehealth strategy in Finland) Web services Web services with login Contact to service system Health care center Hospital University hospital Informing Selfcare eservices Face to face primary services Face to face secondary services Demanding special services Number of clients Portion of digital services Cost/client Web pages (Personal Health and Wellfare Record) Mental health house Regional and national systems (EHR data repository) Health library Self care path Virtual Hospital Providers directory and comparison Digital primary care (ODA) Next-gen EHRs: Apotti / UNA

edeliveries from pharmacies compared to reimbursed prescriptions years 2009 2015 Million prescriptions 5,0 Pharmacies 31.03.2012 Public health care 31.03.2013 Private health care 31.12.2014 (>5000 res/v) 4,5 4,0 3,5 3,0 eprescriptions reimbursed prescriptions 2,5 2,0 1,5 1,0 0,5 8/2015 eprescriptions for 4,6 M people 0,0 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 2009 2010 2011 2012 2013 2014 2015 (84 % of population)

Logins to My Kanta Pages and Number of Visitors by Month 05/2010 07/2016 1 000 000 2010 2011 2012 2013 2014 2015 2016 800 000 600 000 400 000 200 000 0 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 1 4 7 10 614 518 All logins 320 292 Number of visitors 1.6 Mill. persons have used service by 30 Apr 2016 Use by citizens (health records and prescriptions) 14.9.2016 44

456.2 Mill. (31.07.2016) 436.7 Mill. (30.06.2016) 370.2 Mill. (31.03.2016) 301.2 Mill. (31.12.2015) Documents Registered / Archived in the Patient Data Repository Concerning 5.3Mill. (31.07.2016) 14.9.2016 5.2 Mill. (30.06.2016) 4.9 Mill. (31.03.2016) 4.7 Mill. (31.12.2015) 45

Information Management Service in Patient Data Repository: Number of Informings, Consents and Denials 31.03.2016 3 840 812 informings 3 283 758 Kanta informings total 837 580 Informings Registered in My Kanta Pages 1 767 809 consents 758 450 Consents Registered in My Kanta Pages 30 254 denials 14.9.2016 46

In Finland, health related data will be gathered in one place for the generation of new knowledge and for use as the basis of new services Antti Kivelä, Sitra Isaacus was the bishop of Turku who ordered that births, deaths, and causes of death must be recorded

Antti Kivelä, Sitra

Coded patient data Health problems Analysis of data by CDS based on best evidence Potential for health benefit Updating data Patient Health care

Implementing evidence-based care on populations In a virtual health check all decision support rules are executed in a population of patients, and resulting reminders are listed.

Two purposes for the virtual health check (VHC); executing decision support rules in a population of patients Clinical: find people who need interventions and contact them Persons identified Analytic: create statistics on quality of care Anonymous

Care gaps identified in a Virtual Health Check for a population of 16 000 by a set of 100 CDS rules Blood-pressure lowering drug not used in moderately high BP and high cardiovascular risk 396 All beneficial drugs not in use in heart failure 143 No visits for a patient with diabetes during last 13 months 58 Check drug dosing (renal dysfunction) 1164 Stop/change drug (contraindicated in renal dysfunction) 28

Table 1. Example of the calculation of potential to benefit (PTB) for one patient. In this format the table can be used for prioritization of interventions in the situation where no intervention has yet been selected to be implemented. If the same outcome (in this example death) can be influenced by two or more interventions, the ARR obtained from the second intervention is not any more as large after the first intervention has been implemented, because the first intervention will reduce the baseline risk (BR). The total potential to benefit (the sum of PTBs of different interventions, which is 3.16 in the table) would be smaller for a patient who has already stopped smoking, because the baseline risk would be reduced. Diagnosis Intervention Outcome Impor-tance (IO) RRR (and time unit if applicable) BR and time unit ARR (RRR x BR) NNT (1/ARR) PTB (IO x ARR) Coronary disease Smoking cessation counseling Death 9 0.1 0.3/10 y 0.03 33 0.27 Coronary disease Statin Death 9 0.18 0.3/10 y 0.054 19 0.49 Knee osteoarthritis Arthroplasty Pain VAS < 4 5 0.6 0.8/1 y 0.48 2 2.4 Total 3.16

Coded patient data Health problems Analysis of data by CDS based on best evidence Potential for health benefit Updating data Patient Health care Quantitation and synthesis of benefits and harms The citizen/patient chooses - which outcomes he/she wants to pursue - which targets he/she wants to meet - which interventions he/she wants Care plan

Prioritization for health benefit The population is listed and sorted by care gap and potential health benefit For each patient, the most important interventions are put on top The patient s values and preferences influence the selection of interventions to the care plan

Coded patient data Health problems Analysis of data by CDS based on best evidence Updating data Patient Potential for health benefit Quantitation and synthesis of benefits and harms Tools Self management Action Action Care plan

On-line health check report for the citizen

On-line health coaching

Coded patient data Health problems Analysis of data by CDS based on best evidence Potential for health benefit Updating data Patient Health care Care plan Quantitation and synthesis of benefits and harms Self management Action Tools Action Action Action Workflow Resource planning software (EPR)

1 17 7 3 Patient data 2 Health problems Guidelines: Analysis by CDS Potential for health benefit 4 Care plan 5 Quantitation and synthesis of benefits and harms 6 16 Updating data Patient Self care Action 9 History data (big data) 15 14 16 13 Tools Health care Action 10 Action 8 Action Workflow 11 ERP 12

1. All data about the patient (from the EHR, PHR, wearable devices, Kanta earchive, biobanks) is the starting point in making a care plan. 2. Clinical decision support based on trustworthy guidelines analyzes the data by using evidence-based rules, risk calculators and databases (including big data and genomic databases). A PICO ontology links evidence to the health problems and charcteristics of the individual patient. 3. Clinical decision support identifies care gaps and interventions that could improve health outcomes of the patient. 4. s are constructed to fill the care gap. If the patient has many health problems, individual recommendations from many clinical practice guidelines and care pathways will be listed. 5. Clinical decision support tools that utilize risk calculators, prognostic models and interactive summary of findings tables of research evidence are used to quantify benefits and harms individually for the patient, so that the interventions that would benefit the patient most are on top. Interactions of interventions (such as drug-drug interactions), and concordant and discordant recommendations are taken into account at this stage. 6. The recommendations are shown to the patient, using decision aids that make the benefits, harms and burdens of interventions easier to understand. The patient chooses which interventions he or she is willing to use. The patient defines his or her individual targets (together with the professional) according to the principles of the chronic care model. 7. The interventions that have been chosen to be performed are recorded in the structured care plan. Care protocol templates can be used for recording bundles of interventions. 8. The actions recorded in the care plan have codes that can be analyzed to guide the process of care and the provision of care for the whole population. 9. The patients are offered self-care interventions and tools and on-line health coaching. 10. Actions needed from health care professionals serve as input to resource planning tools that link the actions with the competencies, equipment, rooms, and other resources needed for their completion. Bookings can be automated and can also be made by the patient.

11. The resource planning tools place the actions on the task list and schedule of professionals. Tools are provided that make the work easier and faster. The right thing is made the easy thing to do. 12. The resource planning tools have access to all care plans of all people in the population. In this way the volume of care needed, and the availability of resources is known when the care plans are made for individual patients. If overuse of resources threatens, the care plan can be modified. When prioritizing actions for individual patients in the population, the conclusions from steps 5 and 6 are used as guidance. 13. The patient and the professional meet face-to-face or virtually. 14. The professionals record observations and interventions in the stuctured EHR from where they are forwarded to the national earchive and big data repository. 15. The patient records his or her health data, symptoms, and functional ability, as well as measurements from home monitoring into the PHR from where they are available for analysis by CDS. 16. The data recorded by the professionals, patients, and devices are anonymized and stored in a big data repository where they are used for the creation of new knowledge and for developing prediction models. The big data repository can also received data from the patient s environment, and position data can be linked with patients. 17. CDS uses both individual patient data and big data for determining the patient s baseline risk for events, and making recommendations ( search from history earlier patients that are similar to the index patient and see what happened to them ). In a learning health care system every single data item (such as a single blood pressure measurement) contributes to knowledge. Similarly, every path of the patient can be analyzed for finding shortcuts in the care of future patients.

Table 2. An example of health benefit from filling the care gap in a population for three interventions (for simplicity, the PTB averages in the population has been chosen to be the same as the PTBs for the patient in Table 1. In the real world, interventions that have acceptable cost-effectiveness should be grouped as bundles, and the PTB calculated in a sequential manner (see explanation of Table 1). Intervention PTB average in the population Cost of the intervention for one patient Cost/PTB Number of patients in need of intervention (N) Health benefit in the population (PTB x N) Cost of treating all patients (Cost x N) Smoking cessation counseling 0.27 500 1851 3000 810 1 500 000 Statin 0.49 1000 2040 1800 882 1 800 000 Arthroplasty 2.4 20 000 8333 200 480 4 000 000

The larger the triangle, the larger the health benefit in the population The higher (sharper) the triangle, the more cost-effective intervention Additional benefit from reduction of health inequality

Traditional medical knowledge Big data P I C O Patient data (problems, drugs, tests, measurements, function, procedures, genome) Interventions - Self care - Tasks/services Tests (order sets) Procedures Therapies Support Desired outcome - Change in health status - Improved function - Coping - Avoiding adverse effects Components of the interventions - Recording of data - Preparation - Guidance - Details of procedures - Follow-up Values and preferences

Traditional medical knowledge Big data P I C O Patient data (problems, drugs, tests, measurements, function, procedures, genome) Interventions - Self care - Tasks/services Tests (order sets) Procedures Therapies Support Desired outcome - Change in health status - Improved function - Coping - Avoiding adverse effects Decision support Components of the interventions - Recording of data - Preparation - Guidance - Details of procedures - Follow-up Values and preferences

Traditional medical knowledge Big data P I C O Patient data (problems, drugs, tests, measurements, function, procedures, genome) Interventions - Self care - Tasks/services Tests (order sets) Procedures Therapies Support Desired outcome - Change in health status - Improved function - Coping - Avoiding adverse effects Classification of interventions Components of the interventions - Recording of data - Preparation - Guidance - Details of procedures - Follow-up Values and preferences

Traditional medical knowledge Big data P I C O Patient data (problems, drugs, tests, measurements, function, procedures, genome) Interventions - Self care - Tasks/services Tests (order sets) Procedures Therapies Support Desired outcome - Change in health status - Improved function - Coping - Avoiding adverse effects Freedom of choice Components of the interventions - Recording of data - Preparation - Guidance - Details of procedures - Follow-up Values and preferences

Future: a learning health care system Every data item in the electronic health record contributes to the body of medical knowledge and becomes a part of a prediction tool. Every series of actions of the EHR user and every path of the patient in the IT system helps to understand workflows and find shortcuts.

Additional material WONCA Policy statement on ehealth: http://www.globalfamilydoctor.com/getfile.aspx?oid=1ce61f79- A9AB-4243-87EC-EBDC73B4C040 This video describes a comprehensive medication review tool https://dl.dropboxusercontent.com/u/14785933/videos/comprehensive%20medication%20review%20d emo.wmv that is used as the intervention in the ongoing PRIMA-eDS study (www.prima-eds.eu). The protocol of the study (which has recruited 4000 patients) has been published: http://trialsjournal.biomedcentral.com/articles/10.1186/s13063-016-1177-8 How to build an ideal healthcare information system? Essay in WONCA Europe World Book of Family Medicine 2015 http://bit.ly/1ovol5q Presentations on the Finnish ehealth strategy: https://dl.dropboxusercontent.com/u/14785933/presentations%20in%20english/20151015_finland_esoc ial%20and%20ehealth%20strategy%202020.pdf https://dl.dropboxusercontent.com/u/14785933/presentations%20in%20english/201608_thl- OPER_Situation%20Review%2007-2016_VJormanainen_%20EN%20%282%29.pdf https://dl.dropboxusercontent.com/u/14785933/presentations%20in%20english/nhs2016_finland_sillan aukee_compressed.pdf European Health Observatory Evaluation of Finland s Health and Social Services Reform: http://alueuudistus.fi/en/european-observatory-finland-workshop-6.-7.9.2016

Thank you! i ilkka.kunnamo@duodecim.fi www.ebmeds.org Twitter: @ilkkakunnamo