Carolina s Journey: Turning Big Data Into Better Care Michael Dulin, MD, PhD
Current State: Massive investments in EMR systems Rapidly Increase Amount of Data (Velocity, Volume, Veracity) The Data has enormous potential to solve major healthcare issues Minimal ability to use data to improve healthcare delivery at population or individual level Future State: Health data becomes integrated Data governance allows patients to control their data Data & Analytics: Identify Best Practices Drive Population Health Personalize Healthcare Achieve Triple Aim
Licensed Beds 7,800 Employees 62k+ Care Centers 900+ Patient Encounters 11M Hospitals 42 EMR transactions 20M
CHS/DA 2 : Differentiation & Recognition Huge primary care base Provider owned with instant access to tremendous clinical expertise Our clinical data volume and breadth is a significant asset Dickson Advanced Analytics (DA 2 ) 6
Level 7: Personalized medicine Level 6: Waste elimination Level 5: Cultural data literacy Level 4: Evidenced-based population management Level 3: Automated external reporting Level 2: Automated internal reporting Level 1: Vocabulary, metadata, & data governance Level 0: Core data integration
Population Health Advanced Analytics Predictive / Forecasting Cost and Operational Analytics Quality Informatics (Acute and Ambulatory) Outcomes Research / Qualitative Evaluation Marketing Analytics / Strategic Support Geospatial and Community-Based Analytics Project Management / Internal Operations Support
Building Analytic Competencies: Structure Business Operations & Strategy Population Health, Research & Evaluation Central & Shared Services Dickson Advanced Analytics (DA 2 ) 9
Building Analytic Competencies: Operating Model Dickson Advanced Analytics (DA 2 )
Work Effort: Tracking Dickson Advanced Analytics (DA 2 ) 11
Analytics in Healthcare: Impact Differentiator Advanced Analytics & BI Improve Patient Outcomes Predict Healthcare Needs Create Transformative Solutions Based on CHS Data Promote Community Health Dickson Advanced Analytics (DA 2 ) 12
Becoming a Data Driven Healthcare System
CASE STUDY #1 Segmentation and Care Management
Big Data: Management & Analysis CHS Billing Systems CHS EMR System IDX STAR Cerner EDW Panorama Statistical Algorithm Segments Behavior & Consumer Data Experian Address- Based Geospatial Data ESRI I/S Data Services Analytic Services Dickson Advanced Analytics (DA 2 ) 16
Our Results: Compare with Consultancies High Risk 5% High Risk 5% Advanced Cancer 0.6% Complex Chronic 6.6% 7.2% High Risk Aging, Rising Risk 16.7% 16.8% Rising Risk Rising-Risk 20% Mental Health 0.1% Pregnancy & Delivery 2.6% Newborns & Toddlers 2.5% Prescriptive, Occasional, Chronic 95% Sparse Information, Acute & Well 70.8% 76.0% Low Risk Low-Risk 75% Sg2 The Advisory Board Company Dickson Advanced Analytics (DA 2 ) 17
Pareto Rule: 24% of Patients With 76% of Billed Charges High Risk Rising Risk Advanced Cancer 0.6% Complex Chronic 6.6% Aging, Rising Risk 16.7% Mental Health 0.1% Pregnancies & Deliveries 2.6% Newborns 2.5% % of Population Jan 2011 Dec 2013 Total Billed Charges Jan 2011 Dec 2013 11.3% 40.5% Low Risk Sparse Information, Acute & Well 70.8% 24.1% Note: Mental Health Focus segment is 0.1% of the patient population and about the same amount of the total billed charges too 0.1% 4.5% 2.2% 17.4% Dickson Advanced Analytics (DA 2 ) 18
Fine Tuning: Clinical Insights Patients Complex Chronic Segment 142,247 and Living 131,392 and Avoidable Utilization 1 in Last 12 Months 27,782 and Active Primary or Specialty Care 2 22,569 and Residence in Core Market 21,814 and > 4 Body System Conditions/Diseases 18,697 and 8 or more Therapeutic Classes 15,714 and no Mental Health Conditions 3 7,031 1. Avoidable Utilization consists of Avoidable ED visits (NYU Algorithm) and Avoidable Inpatient Hospitalization (PQI / PDI as defined by AHRQ). 2. Active Primary or Specialty Care patients having a PCP / Specialty (as defined by PCP attribution logic) in the last 18 months. 3. 200+ mental health diagnosis codes applied 19
Care Management: Patient Density Across Our Region Care management patients are spread across the CHS footprint Patient density across the geography supports broad implementation of the regional care team approach Dickson Advanced Analytics (DA 2 ) 20
Outcomes: Surveillance and Evaluation Dickson Advanced Analytics (DA 2 ) 21
Panorama: The single source of truth and analytic results Other Clinical Data (EMR) Other Canopy IDX DATA SOURCES STAR Esri Billing Data Address-Based Geospatial Data SYSTEM SINGLE SOURCE OF TRUTH Enterprise Data Warehouse (EDW) Experian Behavior & Consumer Data PANORAMA Full 360 view of patients and potential customers. Including segmentation, communications preferences, appointment information, gaps in care and more. Care Management Customer Relationship Management Care Delivery
CASE STUDY #2 Predictive Analytics at the Point of Care Dickson Advanced Analytics (DA 2 ) 23
Readmissions Model: Patient Risk Assessment Prior State Current State Done After EMR and Patient Review Done Prior to Seeing Patient Limited Capability Risk Assessed from Predictive Model Case Manager Variation Automation Decreases Variation Done at Admission Real Time Dickson Advanced Analytics (DA 2 )
KEY PREDICTORS Demographics READMISSION RISK MODEL 8 Psychosocial 2 Labs & Medications Co-morbidities Utilization 11 8 7 25 Dickson Advanced Analytics (DA 2 )
HAI Rick Model: Genetics, Utilization, Clinical, Lab
Implementation Patient Infection Risk Model Transmission Pattern Model Infection & culture data Patient clinical data Patient billing data Low risk for CRE High risk for CRE CRE genomics Patient temporal & spatial CRE genomic transmission probabilities Epi transmission data G A B D E C H H F Improve surveillance efficiency Alert clinicians of infection risk Identify how CRE spreads Use genomic models to change care delivery 27
Initial Hospitalization Risk Model 7% 29% 7% US population Hospitalized in 2011 29% of all healthcare expenses 2010 National Hospital Discharge Survey: Number of discharges: 35.1 million Discharges per 1,000 population: 114 Average length of stay in days: 4.8
Time-to-Event Analysis Eligible Patients X Event occur O Censored Interest is time to hospitalization Extended Cox Model Predict risk of hospitalization over time X O O O O X X O 01/01/2013 12/31/2013 06/30/2014 Recruitment Period Follow-up Period
Hospitalization Survival Analysis Primary Outcome: Factors Influencing Hospitalization Top 10 higher risk factors: Hospitalized (1-year prior to end point) Outpatient service (1-year prior) English as primary language heart diseases (2-year prior) fluid and electrolyte disorder (2-year prior) ER visits (1-year prior) Had major procedures (2-year prior) CCI score greater than 4 Had advanced imaging (2-year prior) bacterial infections (2-year prior) Lower risk factors: Black or Hispanic Currently married Had 7 acute conditions (2-year prior) Having primary care physician
Probability of hospitalization Survival Estimates 1 0.9 0.8 0.7 Patient1: all top 10 higher risk 0.6 0.5 0.4 0.3 0.2 0.1 Patient3: half 'top 10 higher risk factors' & half lower risk factors' are 1 & all other variables 0. Patient2: lower risk factors' are valued 1 & all other 0 0 100 200 300 400 500 variables 0. Days
Other Analytics Applications
DA 2 Clinical Quality Informatics (CQI) Team: Registries CMS & TJC Inpatient and Outpatient Quality Reporting (i.e., Core Measures) American College of Cardiology NCDR CathPCI Registry ACTION Registry -GWTG ICD Registry The Society of Thoracic Surgeons Adult Cardiac Surgery Database American Heart Association Get With The Guidelines Stroke Registry Code STEMI Database Code COOL & CARES Database Global Trigger Tools (Adult Inpatient and Perinatal) American College of Surgeons - National Surgical Quality Improvement Program Dickson Advanced Analytics (DA 2 ) 33
DA 2 Cost Analytics: Clinical Optimization Dickson Advanced Analytics (DA 2 ) 34
DA 2 Population Health Analytics: Clinic Reporting Our clients find them extremely useful. (City of Gastonia) is providing the 2014 YTD onsite clinic utilization report information to the city manager/board Dickson Advanced Analytics (DA 2 ) 35
DA 2 Quality Analytics: Ambulatory Quality Reporting (AQR) Huddle Report Population Management Performance Report Quality Dashboard Dickson Advanced Analytics (DA 2 ) 36
% Colorectal Cancer Screening 70.0% 65.0% 64.0% 62.9% 62.4% 62.7%63.0% 62.4% 62.5% 63.4% 62.0% 61.0% 60.0% 58.3% 55.0% 52.3% 55.4% Cervical Cancer Screening 50.4% 80.0% 50.0% 46.3%46.3% 45.0% 45.0% 43.4% 47.9% 78.0% 76.0% 74.0% 72.5% 74.5% 74.7%75.0% 74.7% 74.3% 74.4% 74.7% 73.8% 40.0% Oct-13 Nov-13 Dec-13 Jan-14 Feb-14 Mar-14 Apr-14 May-14 Jun-14 Jul-14 Aug-14 Sep-14 Oct-14 Nov-14 Dec-14 Jan-15 Feb-15 Mar-15 Apr-15 72.0% 70.0% 69.2% 70.6% 71.7% 2015 Target: CHSMG HEDIS Payor Mix Adjusted 75th Percentile (67.4%) 2014 Target: CHSMG HEDIS Payor Mix Adjusted 50th Percentile (57.5%) 68.0% 67.8%67.9% 66.3% 66.0% 64.0% 62.0% 60.0% 64.6% 65.2%65.4% 63.9% Oct-13 Nov-13 Dec-13 Jan-14 Feb-14 Mar-14 Apr-14 May-14 Jun-14 Jul-14 Aug-14 Sep-14 Oct-14 Nov-14 Dec-14 Jan-15 Feb-15 Mar-15 Apr-15 37
DA 2 Quality Analytics: Quality Care & Comfort (QCC) Dickson Advanced Analytics (DA 2 ) 38
Our Teammates
Dickson Advanced Analytics (DA 2 ) 40
Michael Dulin Carolinas Healthcare System