The Blue Matrix: How Big Data provides insight into the health of the population and their use of health care in British Columbia Martha Burd, Director, Health System Analytics Health System Planning and Innovation Division Ministry of Health, British Columbia, Canada 1
Multiple Encounters, Separate Data Patient has multiple encounters with Health System in year Information on encounters collected in separate databases No one database can tell the whole story! Each encounter with health system collects point-in-time info on patient s health Data collected different data; different definitions; ICD10, ICD9, neither May miss the pattern or trajectory Hard to see who are the major users of health care Tests Need a BIG Data Approach Tests 2
2009: Inspiration from Bridges to Health A single journal article changed the way we analyze data Using Population Segmentation to Provide Better Health Care for All: The Bridges to Health Model, by Joanne Lynn et al. Milbank Quarterly, Volume 85, Issue 2, pages 185-208, June 2007. We were given the challenge to create population segments Heath Status Groups based on available ministry data. Pathway of our discussions and thinking. What populations do we want to separate out? What data could we use? We ll need to bring all the data into a single database What service lines should we separate out? How granular do we want the data? 3
Putting the Picture Together: BIG DATA Big Data: Use information about the patient from across the health care system Doesn t need all data in one huge database Needs Key information from each data set BC s Big Data approach takes data from administrative databases: Physician billings, Hospitalizations, Prescription drugs, Inter-Rai assessments, Home & Integrated Care programs, Emergency Department visits, Eligibility for programs Synthesizes data into value-added information Chronic conditions Population segmentation Result: Better understanding of health of BC population and their use of Health Care system Achievement: Better planning, Improved Care 4
Population Segmentation Guidance from Bridges to Health The people in each population segment must have sufficiently similar health care needs, rhythms of needs, and priorities to make the segment useful for planning, but each segment must be different enough to justify separate consideration. The definitions of population segments should be easy to understand by clinicians, planners The number of population segments should be relatively small Ministry s considerations in developing our population segmentation Build upon Ministry s work and definitions Should reflect the person s health status over the entire year For multiple conditions: priority assignment to highest need in year Population segmentation will be an additional tool for analysis, not a replacement for a population risk adjusted grouper 5
Identifying people with Chronic Conditions Data sets may only collect single diagnosis per encounter Chronic condition recorded when initially diagnosed. If stable, may not appear on later encounters Analysis based on diagnoses recorded that visit or that year may miss chronic conditions BC s Virtual Chronic Condition Registries Specific algorithms identify earliest diagnosis of 23 chronic conditions Uses diagnoses from physician / hospital / prescription drugs databases If criteria met, add person to registry Achievement: Registry concept has made a difference Analysis: to estimate prevalence rates, analysis of CC populations To improve care: identify target populations for specific care programs, aid physician to identify their CC patients, showed need for care pathways, resulted in physician incentive programs for good CC care 6
Chronic Conditions Population Segments The Chronic Condition Registries identify more than 2 million British Columbians who have one or more of these 18 chronic conditions, or one of these 5 health events / interventions. Challenge: How to analyse people with multiple comorbidities? Matrix divides these Chronic Conditions into High, Medium and Low Complexity groups, based on complexity & acuity of individual conditions or selected co-morbidities 7
Key Concept: Population Segments BC s Blue Matrix used Big Data to develop 13 health status segments End of Life (palliative care) Frail in Care (residential care) Cancer High Complex Chronic Conditions Frail in the Community Maternity and healthy newborns Mental health & Substance Use Medium Complex Chronic Conditions Low Complex Chronic Conditions Child and Youth with Major conditions Adults with Major conditions Healthy / Minor acute illness Non-users Highest health care needs Lowest health care needs 8
Development of Health Status Groups How should population be segmented? Consultation: Which populations were important to separate out? Which ones could be inferred from utilization data or other Ministry data? End of life: People in end of life programs Frail in Care: Living in long term residential care Blue Matrix database summarized use of service in entire year. People who end the year in these groups will have used services differently during year. Cancer: Ministry does not have data from BC Cancer Agency, after patient is referred by family physician. Therefore, assume that patient is receiving non-reported cancer treatment in year following cancer diagnosis on family physician fee for service billing. Cancer diagnosis will put person in Cancer health status group for 2 years. Frail in Community: Living at home with assistance in activities of daily living Eligibility for most of these publicly provided supports is based on assessed need Maternity and healthy newborns: Wanted to separate healthy newborns from newborns with health issues (who are assigned to Major Conditions) 9
Development of Health Status Groups Groups identified through insight from health care records over time Chronic conditions Based on Chronic Condition Registries + Cystic Fibrosis PharmaCare plan 2009: had single Chronic Condition group for all people in chronic condition registries 2010: separated into Low, Medium, High Complex Chronic Condition groups Severe Mental Illness and Substance Use: Mental Health Program list of diagnoses to identify severe mental health conditions and substance use, but no measure of severity Recommended a definition of severe condition: If person was hospitalized for mental illness anytime in a 5 year period, or in PharmaCare Plan G (Mental Health) How to divide the remaining 51% of BC Residents? Non-Users Medical Service Plan registrants (mandatory coverage) who had no services in year (as reported to Ministry databases) Healthy & Minor acute illness Used less than $1,500 of physician services and/or less than $1,000 of PharmaNet (includes out of pocket prescription drugs), and no other health care service. Arbitrary decision! Major Condition (Under 18 years, 18 and older) All remaining BC residents who were not assigned to another health status group. 10
Multiple Conditions? Unique Assignment by Priority People can qualify for more than one health status group in that year, based on their multiple conditions and co-morbidities. Most analysis needs unique assignment to 1 group, to avoid double counting. People are assigned to the group that represents their highest need in the year. Priority used to assign Health Status Groups was determined by consensus Source: 2009/10 Blue Matrix 4.0 11
Key Concept: Service Lines Primary care Obstetrics & Gynaecology Mental health & addiction Ambulatory medical specialist Oncology Emergency departments Acute medical care Elective surgical Transplant surgery Trauma & emergency surgery Palliative care Diagnostic pathology & imaging Clinical support therapies Pharmaceuticals Anaesthesia & pain management Physical medicine & Rehab Community supports for daily living Residential Care Missing from matrix Healthy living Environmental health Health emergency management Community mental health Emergency transport Case management Service lines group together services from different types of providers: example: Fee-for-Service physician billings for services in hospital are in hospital service lines 12
Health Status Groups and use of Health Care, 2012/13 Summarizes how BC residents use approximately $10 billion in publicly funded health care Based on information in Ministry s administrative databases All services converted to dollars based on estimated unit costs. Known gaps in data include Community Mental Health services; Population Health and Wellness; Cancer Agency; physicians on salary; hospital outpatient services; Source: 2012/13 Blue Matrix 5.0 13
Insight from the Blue Matrix: Examples 14
Health Care per Capita Matrix project shows how BC population used over $10 billion in publicly funded health care in 2012/13 Average per person: $2,210 Further insight: by age But is age really the driver? Source: 2012/13 Blue Matrix 5.0 15
Health Care per Capita: Insight using Health Status Groups Population segmentation shows different utilization by pop segments Source: 2012/13 Blue Matrix 5.0 16
Health Care per Capita: Insight using Health Status Groups Health conditions drive use of health care Source: 2012/13 Blue Matrix 5.0 17
Insight: Population segments use health care differently This view divides people by the number of times they used ED in year and shows the distribution across pop segments. High % of people who visit ED 6+ times in the year: High Chronic Conditions, Low Chronic Conditions, Mental Health Source: 2009/10 Blue Matrix 4.0, All Interior HA residents who used Fee for Service ED P:hysician services 18
Insight into health trajectories: Progression of Chronic Conditions People with High Complex Chronic Conditions in 2008/09: What was their major health condition in previous years? 2008/09 2007/08 2006/07 2005/06 2004/05 2003/04 2002/03 Non User 0% 0% 1% 1% 1% 2% Healthy 1% 1% 2% 3% 4% 5% 18+ yrs Major Non-Chronic 0% 1% 2% 2% 3% 3% <18 yrs Major Non-Chronic 0% 0% 0% 0% 0% 0% Low Complex Chronic 4% 8% 11% 15% 18% 22% Medium Complex Chronic 7% 11% 15% 18% 21% 22% Mental Health & Addict. 1% 1% 1% 2% 2% 1% Maternity 0% 0% 0% 0% 0% 0% Frail in Community 1% 1% 1% 1% 1% 1% Disablity 0% 0% 0% 0% 0% 0% High Complex Chronic 100% 84% 72% 62% 53% 45% 38% Cancer 1% 3% 3% 3% 3% 3% Frail in Residential Care 0% 1% 1% 1% 1% 1% End of Life 0% 0% 0% 0% 0% 0% Not in BC 0% 1% 1% 2% 2% 3% Total 100% 100% 100% 100% 100% 100% Analysing progression contributed to Ministry s strategic plans for prevention, and early intervention / better care to slow progression to higher complexity levels Source: 2009/10 Blue Matrix 4.0 19
Insight: use of health care is higher in transition year People use more services in transition year To diagnose condition Failing health Comparison of the services used per capita of new entrants compared to people who were in group in previous fiscal year Frail in Residential Care High Complex Chronic Conditions Per Capita Cost $70,000 $60,000 $50,000 $40,000 $30,000 $20,000 $10,000 $- Existing Patients New Patients Per Capita Cost $18,000 $16,000 $14,000 $12,000 $10,000 $8,000 $6,000 $4,000 $2,000 $- Existing Patients New Patients Blue Matrix 4.0, 2009/10 20
Blue Matrix.Insight into the BC Health System works on 3 Levels 1. As a Broad Concept: Think about the health needs of different populations Think about the health conditions that drive service use Think about all the services that people use, across the system 2. As a Data Base: Blue Matrix database tables in Health Ideas Summarizes all services that each BC resident used each fiscal year that are reported to the Ministry databases, 2002/03 to 2012/13. Useful for wide range of analyses, can be linked to other MOH databases 3. Blue Matrix summary tool in Excel Summary of service line utilization by Health Status groups, where they live, age groups % using service Average use per user Estimated dollars of services used Uses blue shading. Therefore, the BLUE MATRIX! 21
Thank you Martha.burd@gov.bc.ca 250 952-2842 22