SRS Project the use of Big Data in the Swedish sick leave process EUMASS Scientific program 2016-03-04 Stöd för rätt sjukskrivning
Aim of presentation To present how Big data analysis can be used to support decisions in the Swedish sick leave process, and the possible benefits of such a decision support system To invite to collaboration
Content of presentation About decision support systems Knowledge versus statistical based Predictive models from Big Data Project idea Results so far Work this year EUMASS survey
About decision support systems Stöd för rätt sjukskrivning
Decision support systems (DSS) Assist practitioner to make a better analysis Combine practitioner knowledge and experience with support from the DSS Improved and knowledge based practitioner performance Normative effect, reduce disparities
Two types of DSS Knowledge based ( expert systems ) Based on explicitly described human knowledge Ex drug interaction: IF drug X is prescribed AND drug Y is prescribed THEN alert doctor In Swedish sick leave process: Guidelines (FMB), since 2008. Statistically ( Big Data ) Based on patterns in observed Big Data Ex: trading systems, weather forecast In Swedish sick leave process: SRS decision support system
Pros and Cons Knowledge based Statistically based Pros +Precise knowledge, clear hypotesis +Mostly understandable +High precision +Easy to explain +Easy to build +Feed-back learning included +Dynamic +High coverage +Easy to include new knowledge Cons - Expensive to develop - Expensive to maintain - Static - Do not cover bad data or data in between - Requires a lot of data - Difficult to explain - If low quality or noisy data => bad result
Predictive models from Big data Data sources Electronic sick leave certificate Collect and structure data and text Structured data Analyzing, finding patterns Predictive models Registers Medical Health records Statistic analysis
Project idea Stöd för rätt sjukskrivning
Project idea a common support system for several parties A statistically based decision support system to Give predictions for length of sick leave Give early identification of individuals with needs of specific and coordinated interventions Propose interventions to increase RTW The system shall also Increase knowledge of effective interventions Serve as pedagogical support in dialogue with patient Increase cooperation between parties/actors
Parties involved The Government of Sweden Agreement - Sick leave and rehabilitation The National Board of Health and Welfare The Swedish Social Insurance Agency Swedish Association of Local Authorities and Regions (SALAR) 20 County Councils and Regions Utveckling Stöd för i rätt nära sjukskrivning 290 Municipalities samarbete
Big data Collection and use in SRS Supports the 2 dialogue with Interventions, statistically most the patient 3 likely to increase positive outcomes 1 Evaluation of predictive factors 5 Outcome data looped 4 Outcome data collected in database
Vision: A common decision support system
Expected benefits Increased RTW by a correct intervention at the right time Increased knowledge of effect of risk factors and interventions on RTW, common to all parties Increased conditions for equal care despite region and experience, rookies are brought to a higher level
Challenges Engagement, committment and cooperation needed from several stakeholders Legal aspects Ethical aspects Operational (from the end user s point of view) Technical and architectural Big Data analysis and predictive model accuracy
Results so far Stöd för rätt sjukskrivning
Pre-study completed Pre-study investigated if project idea was possible to realize Legal Technical and architectural Predictive models User needs Report to government in October 2015. Project idea is possible to realize, in theory. Decision: Proceed, more practical investigation needed.
No of days with sick leave pay within 30 days Patterns in Big data sick leave for osteoarthritis Growth trajectory models 30 25 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 Months, after day 21 Högt, sedan sjunkande (8 %) Sjunkande, sedan stigande (12 %) Sjunkande (noll vid 10 mån) (15 %) Sjunkande (noll vid 6 mån) (29 %) Sjunkande (noll vid 4 mån) (36 %)
Work this year Stöd för rätt sjukskrivning
Work this year Proposal on organization and responsibilities for development, maintenance and operation of SRS DSS Cost-benefit analysis Operational analysis User needs Ethical aspects Testing prototypes List of interventions including codes
EUMASS survey Stöd för rätt sjukskrivning
EUMASS survey Question: Does your country have, or have plans for, decision support systems or other tools to support decisions of sick leave, rehabilitation or other similar questions?
Results so far No system (3 countries) Semi-automatic data analyses, for fraud detection etc. (3) Knowledge based decision support systems (6) Statistically based decision support systems (3) No answer (10)
Thank you! Anne Snis Project manager SRS-Project Anne.Snis@kvadrat.se Stöd för rätt sjukskrivning