Medicaid & Predictive Analytics Thomas J. Kessler, Esq. Acting Director, Division of Fraud Research and Detection, Data Analytics and Control Group, Center for Program Integrity, Centers for Medicare and Medicaid Services September 11, 2013
Discussion Items Data Analytics and Control Group Medicaid and Predictive Analytics Medicare Experience with Predictive Analytics 2
CMS Center for Program Integrity Center for Program Integrity (CPI) Dr. Peter Budetti, Deputy Administrator and Director Ted Doolittle, Deputy Center Director for Policy Elisabeth Handley, Deputy Center Director for Operations MPIG MIG DPSG DACG PIEG PEOG 3
Center for Program Integrity (CPI) Data Analy:cs and Control Group (DACG) Kelly Gent Director Raymond Wedgeworth Deputy Director Analytics Lab Division (ALD) Linda Smith, Acting Marin Gemmill Toyama, Deputy Director Systems Management Division (SMD) Craig Mooney, Director Kathy Wolf, Deputy Director Command Center Division (CCD) Brenda Emanuel, Director Division of Fraud Research & Detection (DFRD) Thomas Kessler, Acting Director 4
DACG Func:ons Use sophisticated analytics and technologies to support identification of improper payments and ineligible providers in Medicare and Medicaid. Support and manage information technology investments critical for program integrity activities. Manage the Command Center in support of the Center for Program Integrity s mission. 5
DACG Organiza:on Analytics Lab Division (ALD) Systems Management Division (SMD) Command Center Division (CCD) Provides statistical and data analysis for program integrity Identifies emerging fraud trends through data mining and other advanced analytical techniques Leads model development for the Fraud Prevention System Manages system development and enhancements in support of information gathering and analysis to detect fraud and abuse. Implements data management technologies and strategies to support sophisticated analysis to identify fraud and abuse. Provides a collaborative environment for a multi-disciplinary team, including ZPICs and law enforcement, to develop consistent approaches for investigation and action. Division of Fraud Research & Detection (DFRD) Provides statistical and data analysis for program integrity activities related to the National Medicaid Audit Program. Supports evaluation of predictive analytics across programs. 6
Discussion Items Data Analytics and Control Group Medicaid and Predictive Analytics Medicare Experience with Predictive Analytics 7
Medicaid Predic:ve Analy:cs Many State Medicaid Programs are in various stages of applying sophisticated predictive analytics technologies in their program integrity efforts. In addition, the Small Business Jobs Act of 2010 (the Act), section 4241(e)(3) requires CMS undertake an analysis to determine the feasibility and cost effectiveness of applying predictive analytics in the Medicaid and CHIP programs. The results of the analysis will be included in the Report to Congress due March 31, 2015 8
Challenges and Opportuni:es Challenges: State Resources and funding Competing priorities State staff necessary to support information technology Identifying the right technology Applying predictive analytics prepayment Measuring outcomes Federal role limited to assistance in light of claims data access Opportunity: Advanced technologies have the capability to identify fraud earlier and prevent improper payments
Ac:vi:es CMS is providing technical assistance to States: General TA package: for States considering predictive analytics technologies Targeted TA: for States moving forward with tools Algorithm exchange: considering opportunities to share successful algorithms among States and CMS Medicaid Integrity Institute training sessions Command Center missions focused on algorithm development, investigation approaches, and outcome measurement
Ac:vi:es CMS may approve enhanced Federal Financial Participation for certain allowable activities and resources for Predictive Analytics related to the MMIS, including: Planning, Requirements, IT Hardware/Equipment, Software, Staffing and Contractor Support, Required Reporting (CMS and State) Advanced Planning Document template is available thru the CMS Regional Office
Ac:vi:es CMS will evaluate the feasibility of applying predictive analytics in Medicaid: Focus groups with State Medicaid Agencies that are applying predictive analytics Evaluate outcomes of incorporating available post payment Medicaid data into the Fraud Prevention System to support Medi Medi activities Conduct environmental scans with States
Discussion Items Overview of the Data Analytics and Control Group Medicaid and Predictive Analytics Medicare Experience with Predictive Analytics 13
Fraud Preven:on System The Small Business Jobs Act of 2010 mandates that CMS implement predictive modeling and other advanced analytic technologies to prevent potential fraud, waste, and abuse. CPI implemented this requirement through the launch of the FPS on June 30, 2011. The FPS applies effective predictive models and other advanced algorithms to identify providers exhibiting a pattern of behavior that is indicative of potential fraud, waste, and abuse. 14
Fraud Preven:on System The FPS system currently screens all national Medicare Part A, Part B, and DME claims during the adjudication process and consolidates alerts by provider. Monitors 4.5 million claims each day using a variety of analytic models. 15
Fraud Preven:on System The FPS presents the alert results in a prioritized list, provides detailed information (including claims lines, beneficiaries, associated providers and claims). Results are provided to the Zone Program Integrity Contractor analysts and investigators with views by regions. Results are available to CPI and law enforcement partners in a prioritized national view. 16
National Fraud Prevention Program Claim For Payment IDR CPI Analytics Lab Claims Processing Integrated Data Repository Rules Anomaly Detection Predictive Models Social Network Analysis Fraud Prevention System NGD STARS One PI PECOS FID APS FPS INFORMATION NOT RELEASABLE TO THE PUBLIC UNLESS AUTHORIZED BY LAW This information has not been publicly disclosed and may be privileged and confidential. It is for internal government use only and must not be disseminated, distributed, or copied to persons not authorized to receive the information. Unauthorized disclosure may result in prosecution to the fullest extent of the law. Zone Program Integrity Contractors 17
Examples of Models in Credit Card Fraud Rule Charge for TV in FL Cardholder lives in CA (Unlikely charge) Anomaly Charges for 3 TVs in one day (99% of people buy less than 3 in a single day) Predictive Model Charges for multiple TVs out of state, after a $1.00 charge, on Wednesdays after midnight (Based on experience, these charges have a very high probability of being bad) Social Network Charge for a TV at an address known to have bad charges using a card with a phone number used by a known bad actor (relationship suggests a problem) 18
Models Run Simultaneously Rule Anomaly Predictive Model Social Network Risk Score Health Care Claims Trigger FPS Investigations Complaints Stolen IDs Information from Enrollment Risk Score by a Provider s Book of Business, Not Individual Claim 19
Fraud Preven:on System (FPS) Background The Fraud Prevention System Report to Congress for the first implementation year was published in December 2012. Prevented or identified $115.4 million in payments Generated leads for 536 new investigations and augmented information for 511 pre-existing investigations 20