Use of Predictive Modeling to Detect Overpayments/ Abuse The National Medicare RAC Summit December 5, 2012 Amy Caro Vice President, Health Information Technology Programs
Strategic Direction Section 4241 of the Small Business Jobs Act of 2010 (SBJA) mandates that CMS implement a predictive analytics system to analyze Medicare claims to detect patterns that present a high risk of fraudulent activity, and enables CMS to employ real-time, pre-payment claims analysis to identify emerging trends of potentially fraudulent activity. Established Approach New Approach Pay and Chase One Size Fits All Prevention and Detection Risk Based Approach Legacy Process Stand Alone PI Programs Innovation Coordinate and Integrate PI Programs 2
3 National Fraud Prevention Program Two-pronged Approach
4 FPS Organization Chart
5 FPS Operation Flow
6 Advantages of Fraud Prevention Program
7 Limitations and Challenges for Predictive Modeling
Informing Operations, Programs, and Policy Evaluate Program Results Evaluate Constraints Communicate with Decision Makers Modify as Necessary Establish Baselines Identify Priorities Set goals Monitor and Measure Document Results Analyze Trends and Processes Identify Areas for Policy Changes Implement Tools Implement Methods 8
The Need for Data-driven Distributed Analytics Transformation s Three Part Aim: Improve care Improve population health Reduce per-capita costs The Challenge: Different members of the health community have access to different sets of data Few have been able to look across the data sets to get a real and timely sense of the health ecosystem Transformation could be accelerated through integrated health analytics information to inform strategy, guidance, operations, evaluation The Need: A flexible and scalable analytics platform that can rapidly provide integrated insights to a broad range of health decision makers 9
integrated Health Analytics Platform (ihap) Analytics Fan Layered Framework Service cost comparisons, outliers Proactive fraud detection Predictive Modeling Program Integrity Estimations of future costs Statistical Geospatial Care seeking behavior of populations Schema Matching Data Model Semantic Integration Encryption/Decryption Ontology Data Cleansing Data Governance Need Analysis Web Service Data Virtualization Data Warehouse Clinical Informatics Health Analytics Data Standards Data Security Data Sources Public Health Surveillance Systems Analyst 10
11 ihap Conceptual Framework
Analytics Maturity Model Foundational / Tactical Strategic Enablers Highly Strategic Descriptive Analytics provides limited overview Limited data governance Limited quality assurance Analyses are typically ad hoc and reactive Inconsistent use of BI tools More detailed reports require laborious data gathering and aggregation Predictive Analytics allows forecasting and planning Formal data management exists for critical projects Enterprise reporting with standard BI tools is established for relevant centrally controlled data sources Decision makers still depend on data mining specialists for more detailed information Advanced analytics and predictive models periodically available to provide decision support Prescriptive Analytics suggests possible interventions Holistic systems approach to data governance Automatically available analyses of key performance indicators Power users can run additional ad hoc queries and reports Data mining system allows users to apply analytical tools without deep expertise Continuous real-time monitoring and alerts with drill-down capabilities Rich visualization tools BI integrated with business process management in a closed-loop to improve results 12 Northrop Grumman capabilities can provide an improved path to evidence-based decision-making
Summary of ihap Capabilities Provides flexibility to work with pre-existing architecture as well as new architectures Reduces costs and time for integration among different data sources Offers robust analytics, visualizations and reporting customized to customer needs managing big data Cuts operational costs (e.g., eliminates need for a data warehouse) Generates resources and support for evidence-based decisionmaking within big data 13 Partnering opportunities provide a win-win situation for Northrop Grumman and its partners.
Fraud Analytics Workstation: Anomaly Detection Key Point High cost outliers for specific types of service codes are identified among diabetic claims, User interface for FAW used to demonstrate different fraud scenarios Outliers Detection tab connects to SAS product for identifying anomalies Using CMS PUF of over 9.7 million rows of claims data sample from 2008. Subset of claims by ICD-9 coding for diabetics. Identifies the high cost outliers for different type of service codes Several kinds of charts can be output for user.
Medicaid Eligibility Projections* Linear Fit BEN_BOE_AGE_65_AND_OLDER = 1831633.3-873.03636*YEAR Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) Analysis of Variance Source Model Error C. Total DF 1 8 9 Sum of Squares 62880881 17584916 80465797 Parameter Estimates 0.781461 0.754144 1482.604 82504.9 10 Mean Square 62880881 2198114.5 F Ratio 28.6067 Prob > F 0.0007* Term Intercept YEAR Estimate 1831633.3-873.0364 Std Error 327030.3 163.2293 t Ratio 5.60-5.35 Prob> t 0.0005* 0.0007* 15 *Excludes expansion population
Dynamic Cost Projections from Existing Data Use Case: Enabling dynamic what if scenarios to project future Medicaid costs Context: LA Medicaid Director adjusts various population parameters to project annual cost with the new population Results: Ability to estimate future costs based on historical data and growing understanding of future population Rapidly gain insights to main factors contributing to Medicaid cost expenditures Explore correlations among variables to gain insights to key cost drivers 16 Estimated Enrollment Dynamic Cost Projection
FAW Prototype: Portal Showing SNAP Cases & Medicaid Eligibles in New Mexico Counties This user interface tab shows a flash file of a bubble chart that displays the percent of Medicaid eligibles and percent of population on SNAP (food stamps) over time Bubbles float to show changes: population, percentage of SNAP recipients as well as percentage of Medicaid eligibles over time for the counties shown 17 12/3/2012
Analytic Needs determined via Joint Planning Macro Level Research Identification of Avoidable Expenses Population based Analysis Geographic based Analysis Cost and Performance Trends Procedural Effectiveness Preventative Campaign Effectiveness Member & Patient Analytics Gaps in care High ED Utilization Unfilled Prescriptions High Risk Members High Prescription Utilization Measures and Benchmark Reports Identification of Avoidable Expenses Cost Measures Quality Measures Meaningful Use Reporting Key Performance Indicators Operations Reporting Hospital Average Length of Stay Hospital Readmission Rate Hospital Infection Rate Procedure Effectiveness Cost per Incidence of Care Evidence Based G/L Compliance Patient Centered Medical Home Analytics Quality Analytics Single Patient Visit Report Prioritized Patient Panel Report Complete Patient Panel Report Non engaging Patient Report Population Performance Report HEDIS Measures Affordable Care Act Measures AHRQ Measures Bayou Health Measures 18 = analytic needs of interest to Hood River County Public Health Department, Oregon
Enrollees in Payment Programs (Age Distribution) Payer 1 Payer 2 Payer 3 Payer 4 Payer 5 Payer 6 Payer 7 Payer 8 19 Number of Enrollees (Diabetics) Northrop Grumman Private / Proprietary Level 1
20 Average payment over time by age group
21 Questions?