Big Data In The Medical Industry David Shein, MD dshein@veriskhealth.com Medical Director, Verisk Health Agenda Verisk Health: Who we are What we do Data management in the healthcare environment Example cases Data challenges 2 Verisk Analytics – The Science of Risk Measure, Evaluate and Navigate Risk P & C Insurance Healthcare Mortgage Lending Supply Chain 3 The Markets We Serve Understanding Healthcare Risk Payors Providers Employers Commercial Plans At-risk Physician Groups Self-Insured Third Party Administrators Integrated Delivery Networks Benefit Consultants Disease & Care Management Provider Hospital Organizations Brokers State & Managed Medicaid Accountable Care Organizations Medicare Advantage 4 The Healthcare Crisis: On the Rise ~Healthcare spending is projected to reach nearly $4.6 Trillion in the next decade Employers spent $8,300 on average per employee per year in 2010…costs are projected to rise to more than $13,400 in 2019 Today, GM spends more on healthcare than on steel… Where do healthcare costs rank for you? Today, healthcare spending has reached $2.5 Trillion…and $3 out of every $4 spent is on chronic conditions In 2014, when health coverage is expanded to millions of uninsured Americans, spending is estimated to increase by 9.2% By 2019, healthcare is projected to account for nearly $1 of every $5 spent, or about 19.6% of the national GDP ~CDC ~Health Affairs: http://www.healthaffairs.org/press/2010_09_09.php 5 The Healthcare Crisis: A Smarter Way According to a 2009/2010 Towers Watson Report: • Companies with effective health benefit programs not only improved employee health but also experienced superior human capital and financial outcomes, including: Lower medical trends by 1.2% 1.8 fewer days absent per employee Fewer lost days due to disabilities 11% higher revenue per employee • These organizations believe in identifying the root causes of healthcare cost increases • …and consistently analyze data integrated across programs to identify opportunities, design programs, and measure performance 6 Capability & Capacity: Understanding Risk For… • Budgeting and Cost Containment • Risk adjustment • Cost & utilization driver analysis • High-cost case identification For… • Medical Management • Identification & stratification For… • Provider Network Management • Quality Measurement • Predictive event modeling • Employer reporting • Gaps-in-care measurement • NCQA HEDIS reporting • Trending & reporting • Program Measurement • Fraud, waste & abuse DATA AGGREGATION – ANALYTICS – DECISION SUPPORT - REPORTING – EXPERT INTERPRETATION 7 Data Acquisition and Management is a Core Operational Strength Examples of our Data Capabilities • Data Handling: • Acquisition: We can accept data in virtually any format • We currently accept data from over 300 different vendors with over 1,200 mapping and translation schemas • We process 50 million claims per day and 1.5 billion on a recurring monthly basis. Average turnaround time less than 10 calendar days. 8 Data Captured Enables Analysis of a Complete Picture of Health and Productivity Medical Worker’s Comp. Pharmacy Incentive Disability Wellness HRA Absence Lab Biometrics 9 Data Elements Healthcare trend: Progression from Health Health and Wellness Productivity • • • • • Medical claims Pharmacy claims Eligibility Vision Dental • HRA • Disability (Health Risk Appraisal) • • • Biometric Lab EMR 10 The Data Lifecycle at Verisk Health DATA Source Process Analytics Access • Customer • Payor • Data Warehouse • ETL • Engine • ASP • Service Bureau • Warehouse • Scrub • Cleanse • Map • Quality • Risk • HEDIS 11 Data Analytics Rules Engine CDF (Common Data Format) Enterprise Analytics Medical Intelligence Provider Intelligence DxCG Performance Measurement 14 Demographic Analysis Distribution of spend, health conditions and quality across membership status, age, sex Employee Spouse Dependent Age/sex distribution of the population Expense distribution Population 1% 2-5% 6-15% 16-30% 31-60% 61-100% % of Spend 30+% < 1% 15 Financial Analysis Focus on cost drivers Factors affecting change include: membership, utilization, pricing, intensity • Cost trends • Norm comparison • Modeling Predicted • • Current Future Performance • Actual vs Predicted 16 Clinical Analytics Include disease prevalence, population health status, quality of care • Disease Registry • List of top conditions (chronic and acute) • Prevalence • New diagnoses • Cost metrics • • • Risks: Disease burden in population • Gaps in Care: Care delivery / quality Comparison • Benchmark Commercial norm • Period 1 / 2 Quality and risk measures • Clinical deep dives • • Specific consultative evaluations Comorbidity analysis 17 Predictive Modeling • Relative risk score • Predicts cost and utilization using prior health status and performance • Regression model fits population around normative “1.0” • Benchmarked to: - Book of Business - Norm • Concurrent models Performance • Predictive models Budgeting and forecasting • Likelihood models Medical management Intervention 18 Additional Analytics Understand patterns of utilization and healthcare spending • Utilization • • ER Admissions • • • • Specialty services • • • Acute Subacute Med specialties DME Non-PBM drugs* Office Network utilization • • • Geography Pricing / discount differentials • Tests • • • Lab Imaging Vascular 19 Pharmacy Analytics Utilizing PBM and claims data Examples of analytic capabilities: • Pricing • • • Brand Generic Fill location • Mail order • Prescribing patterns • • • Brand Generic Opportunities for generic conversion • Medication possession ratio (MPR) • Measure of Rx compliance Ratio of: # Days filled # Days Rx • Non-PBM drug • Costs by location • e.g. office, outpatient hospital 20 Case #1: Employer • Large multi-state employer High rates of chronic disease Benefit strategy includes • • • High deductible health insurance and HMO options • Disease management through outside vendors • Challenges: • • • Manage health care costs Provide insight into vendor performance Budgeting and forecasting 21 Case #1: Cost Drivers • Answers questions including: • • • • What is the overall performance across all carriers? How do specific subgroups perform? What is driving changes in medical spend? Are there areas to focus for controlling cost? 22 Case #1: Cost Drivers • Aggregate carrier information to common format for overall analysis • Analyze cost trends and determine drivers • Report on demographic and membership patterns across all carriers • Analyze patterns of health spend by cuts across membership • Influences include health policy, general economy • Impacts on membership changes (e.g. hiring or layoffs) • Monitor changes 23 What Are Key Cost Drivers? Change in Member Months (Thousands) -5% 352 Total Expenses $ Millions 335 +1% 128 285 300 P1 P2 129 P1 P1 Change in Medical PMPM +5% $ P2 P2 Change in Total PMPM +7% $ 364 390 Change in Pharmacy PMPM $ +9% 89 P1 97 P2 Sep 08 – Aug 10: Calculations are based on Total expenses and membership P1 P2 24 Medical Cost Drivers Change in Unit Pricing $ / Event +2% 552 564 P1 P2 Change in Medical PMPM $ +5% 285 300 Change in Utilization Events / Member Months +4% P1 P2 0.56 P1 Full Cycle - Demo 0.59 P2 25 Case #1: Health Analytics • Answers questions including: • What are the key conditions driving health care spend in the population? • What is the quality of the care delivered to the population? • Overall • Subgroup analysis • What approaches can be taken to address the health issues? • How to track interventions? 26 Case #1: Health Analytics • Disease Registry • Identifies key conditions • Prevalence and change in prevalence • Cost by member (PMPM/PMPY) • Comparison to norm • Future capabilities will include industry-specific norm (NAICS) • View by subgroup • Business unit • Individual carriers • Insight into significant population health conditions and patterns of change • Provides a focus for disease management and wellness 27 Disease registry key findings: PMPY cost Condition Prevalence Per 1000 VH Norm • PMPY for the top 4 prevalent conditions is higher than the VH Norm. $11,799 Hypertension 202 107 Hyperlipidemia $10,401 175 81 Diabetes $15,111 85 56 Coronary Artery Disease 51 24 Co #1 VH Norm Analysis Period. Based on current members $10,781 $21,602 $17,393 • Members with Coronary Artery Disease and Diabetes contribute to highest costs. • Opportunities for wellness and disease management programs 28 Case #1: Health Analytics Cont’d • Quality and Risk Measures • Detail on health status of overall population • Compared to BOB and norm • Quality measures – gaps in care • Baseline evaluation and comparison • Evaluate range of quality across multiple conditions • Enable monitoring over time for changes • Evaluation can be done for overall population or subgroup analysis • Carrier • Business unit (location, division) • Demographic cut (membership, geographic) 29 Quality of care comparison: Coronary Artery Disease 29.0% 26.8% Patients diagnosed with CAD and without a lipid profile test in the last 12 months 38.9% 1.6% Patients diagnosed with CAD and without an office visit in the last 12 months 3% • No apparent issues with access to care 2.2% 36.8% Patients diagnosed with CAD and without antihyperlipidemic drugs in the analysis period Patients diagnosed with CAD and HTN without antihypertensive drugs in the analysis period • Plan performance is similar 42.5% 29.2% 7.7% 8.1% 13.6% • Disease management to improve lipid treatment may lower future CAD costs VH Norm Plan A Plan B 30 Case #1: Risk Modeling • Answers questions including • How does the health risk for population of interest compare to other populations? • How do specific areas compare? • How to budget for next year’s health costs? 31 Case #1: Risk Modeling • Normative benchmark for overall population risk assessment • Normalized to BOB for division analysis • Carrier • Business unit (location, division) • Demographic cut (membership, geographic) • Concurrent models • Predictive models 32 Quality comparison: Risk-adjusted spend and utilization by carrier Well managed sick pop. Carrier Lives RRS PMPM (unadjusted) PMPM Admissions (RRS-adjusted) Index ER Index Imaging Index 1 3,180 1.45 285.12 196.63 1.02 0.98 0.92 2 8,327 0.71 85.97 121.08 0.72 0.80 0.95 3 16,784 1.00 200.41 200.41 1.01 0.95 1.05 4 1,903 2.12 532.86 251.35 1.32 1.41 1.21 5 1,201 0.86 189.22 220.02 1.1 1.22 0.85 Poorly managed “healthy” population pop. Analysis for demonstration purposes 33 Case #2: Provider Organization Large single-state provider organization Accepts risk from insurance contracts Looking to perform at the forefront of healthcare with innovative reimbursement approaches • • • Alternative Quality Contract • Experience with local clinical information (EMR) Own data warehouse • • • Challenge: Rising costs and medical expense management • • • Tools for financial analysis, modeling and forecasting Monitor network utilization and cost factors Practice pattern variation: Provider dashboard • • • Quality Efficiency Utilization patterns 34 Case #2: Clinic and Physician Manager and Dashboard • How does performance compare across physicians and clinic sites? • “My patients are sicker than yours” • Who are the top performing providers and clinics? • Efficiency • Quality of care • Where to focus for performance improvement? • How to evaluate utilization patterns… • • • • Drug ER Specialists DME … and take action? 35 36 Case #2: Clinic and Physician Manager and Dashboard • Metrics for comparison based on quality, volume, risk and efficiency • Quality (gaps in care) • Relative risk models • Generic utilization rates • Drill down functionality across drug class to dose • Readmission rates • Efficiency scores: • • • • actual utilization concurrent predicted Total admissions Potentially avoidable admissions ER visits Imaging • Drill down capability to view individual physician or clinic • Comparison to BOB or norm 37 Case #2: Outside Utilization • How to evaluate referral patterns: • Where are outside referrals going? • What diagnoses are being treated outside? • What procedures are being done outside? 39 Case #2: Outside Utilization • Define “events” • Typical views by claim or unit vs episode group • Includes ancillary claims • Capture true event cost • Evaluate utilization patterns • Cost and quality • View events by: • • • • • Provider: Professional Location: Facility Specialty Group (similar events) Code level 40 Case #2: Outside Utilization • Evaluation of referral patterns: 41 Case #3 Large benefit management company • Needs to provide information on wellness and augment disease management for clients • • Chronic disease Clients are capturing HRA (health risk appraisal) • • Challenge • • Create HRA report Understand implications of self-reported health outcomes 42 Augmenting Member Level Data for Analysis: HRA Health Risk Appraisal Claims data • Self-reported data • History of disease • Lab results (glucose) • Biometrics (height, weight) Example conditions with claims and HRA data Diabetes Tobacco use Obesity 43 Diabetes* by Claims and HRA (30%) Claims (66.6%) HRA (33.3%) Individuals with claims for Diabetes have a relatively low rate of diagnosis reporting on HRA Adding HRA will increase the population of diabetic individuals by about 5% *Diabetes claims: Medical Intelligence disease registry criteria *Diabetes HRA: Blood glucose value, blood glucose range, or diabetes history 44 Compliance Rates in Diabetes: Identification by Claims vs HRA Highest compliance Medium compliance Lowest compliance HbA1c in last 12 months Renal disease Eye exam in screening in Lipid profile ACE/ARBs in last 12 last 12 in last 12 last 12 Statins in last months months months months 12 months Overlap (Claims and HRA) Claims HRA only • Individuals identified by HRA only have a lower compliance rate • Individuals identified by claims and report having diabetes in HRA (overlap) have the highest compliance rate 45 Data Hosting Platform - ASP Primary Site • Physically hosted at Class A Datacenter, SAS 70 Type II certified • Industry-leading managed services and enterprise infrastructure provider • Located in downtown Boston with alternate sites throughout the US • Direct connection to Tier 1 Sprint backbone, MCI, Verizon backup • Raised floor, redundant grid power, climate control, smoke detection, fire suppression systems Secondary Site • Duplicate attributes of Primary Site • Located 30 miles northwest of Boston • Direct connections to multiple Tier 1 backbones (Level 3, Global Crossing) • End User Data Warehouse migrated to secondary site during Final QC to Production Post phase • Transactional updates migrated nightly in batch 47 Security • • • • • • • FTP HTTPS Password change 90 days Idle time out Lock out after 3 unsuccessful login attempts Client admin for all users User-level rights access • PHI • Data level • Clinical vs financial 48 Challenges • Processing • Normative data set • 10 mil lives • Clinical quality measures • 500+ • Cloud • Private vs public • Maintaining accuracy of imported data • Security • Managing PHI 49 Challenges • Meeting the needs of a changing industry • PPACA • ACOs • Changing paradigms for care, reimbursement and reporting • Clinical progress • New drugs and diagnoses • Changing clinical guidelines • Coding evolution (ICD-10) • New reimbursement policies • Readmissions • Episodes of care • New and evolving reporting needs • ACO 50 Questions 51