Innovative Fraud Detection at CMS Mike Doane, Technical Director Federal Healthcare April 10, 2013
Agenda
• • • • • •
Fraud Characteristics CMS Requirements MarkLogic POC Functionality Use Cases Future Capabilities Q&A
Slide 7
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Healthcare Fraud is Huge FY 2012 90 80 70 60 50 40 30 20 10 0
$B
80
4.2 Total Fraud (FBI Est.) Amount Recovered
New approaches to detection are needed Slide 8
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
1.6 Spending on Recovery
What does fraud look like?
• • • • • • • • • • •
False storefronts Beneficiary shopping (compromised benes) Networks of fraudsters Ambulance trips – no check-in Home health Double-billing DME proliferation ID theft Prescriptions (forged, resold, etc.) Upcoding Kickbacks
Slide 9
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Today’s Approach
Sharding of databases by states and territories Manual correlation across states Multiple versions of algorithms Looking at only the claims Looking at these claims well after they’ve been submitted Not incorporating other available sources Not sharing results
Slide 10
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
CMS Asked For
Scalability to evaluate the entire country’s data as a single, unified data set The Agility and Flexibility needed to keep up with changing data sources and evolving data analysis strategies Better Performance - a shorter Time To Market for its data Facilitation of Enterprise Data Sharing within CMS and with CMS partners
Slide 11
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Scalability
Over 600M documents, 4+TB in the POC Deliver a 300TB+ cluster and eliminate sharding of data by geography Fifty states plus six territories Unknown additional sources
Slide 12
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Agility and Flexibility
Load data as-is Handle combination of full text (unstructured) and structured data Support in-line data descriptions or enrichment Allow composable queries across all data Full-text search, geospatial constraints, and traditional and fielded constraints all in the same query
Handle Unpredictability New data of any type readily incorporated without schema changes – indictment docs, Dunn & Bradstreet, Facebook, Property records
Support flexible Provider and Beneficiary profiles
Slide 13
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Flexible Profiles Provider
NPI
Associations
Beneficiary
Metadata
Statistics Assoc Hosp RxCount
Assoc -Prov
Providers Visit Bill-Amt
Addresses InclusionAddress reason
ID
Address Address
Benedistance
Slide 14
Annual Billings
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Early- Distance TopschedII refill -Rx
Flexible Profiles Provider
NPI BoardMember Assoc Hosp RxCount
Associations
Beneficiary
Providers
Metadata
Statistics
Visit
Assoc -Prov
Exclusion
Addresses InclusionAddress reason
ID
Bill-Amt
Address Address
Benedistance
MRIcount Slide 15
Annual Billings
DEA
Indictment
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Early- Distance TopschedII refill -Rx
PersonalProperty Auto House $-value
Ambulancecount
$-value
Performance – Time to Market
Operate on valid and contemporary data Handle multiple disparate data sources with varying schemas Enrich in place by reference and related data sources Architect for performance and flexibility Distribute loading across servers in the cluster No model changes for new incoming formats Database and Data Warehouse are the same – no data staging between products
Slide 16
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Enterprise Data Sharing
Support direct data access from external systems Connectivity via Java API, REST API, .NET, ODBC/SQL Facilitate sharing data back to the states or other parties Expose data via SOAP or REST web services Provide efficient transformation of data to meet third party required exchange formats
Slide 17
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
MarkLogic Delivered SAS, Excel
Fraud Review Application
Alerting
APS
Load data “as is”
MSIS Claims
Slide 18
MDEM Claims
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
NPI
Reference Data
Indictments
SAS AFRs
POC Functionality
Provider and Beneficiary Profiles Search Detailed Views Easy Risk Scoring
Granular State Policy Document Search Contextual Enrichment of Alerts Incorporation of External Alerts Co-occurrence Analysis Link Analysis Geospatial Analysis
Slide 19
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Link/Network Analysis
Slide 20
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Geospatial Enrichment & Analysis 1214 Main Austin TX 98143
Slide 21
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
State Policy Documents
Slide 22
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Contextual Enrichment of Alerts
Slide 23
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
POC Findings
Aberrant behavior based on NPI registration Compromised beneficiaries Value of external data sources to support link analysis Aberrant behaviors based on high billings Aberrant behaviors based on high procedure codes Aberrant behaviors based on geospatial data Providers that should have been excluded based on sanctions, exclusion lists Slide 24
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Use Case 1: Provider Profiles Clifford Ubani
NPI registration (data enrichment) Address coordinates Dunn & Brad Street incorporation (external data sources) Ability to link additional participants (unknown players)
Slide 25
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Profile of Clifford Ubani – Multiple NPI Clifford Ubani
Family Health Services 8313 Soutwest Fry, Suite 109 Houston, TX Home Health
Directcare Link 8313 Southwest Fry, Suite 113 Houston, TX DME
Family Health Services (Family DME) 8313 Southwest Fry, Suite 113 Houston, TX DME
Slide 26
St. Michaels EMS 9896 Bissonet St. Suite 340 Houston, TX Ambulance
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Special Care 8313 Southwest Fry, Suite 215 Houston, TX DME
Ubani – Common Address
Family Health Services 8313 Southwest Fry, Suite 109 Houston, TX Home Health
Slide 27
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
CKCC 8313 Southwest Fry, Suite 109 Houston, TX Home Health
Ubani – Dunn & Bradstreet Dunn & Bradstreet links Clifford Ubani to: Family Healthcare Person of Interest – Princewill Njoku (CCK DME) 8313 Southwest Fry, suite 107
Future EMS Persons of Interest Monica Stahl, Franklin Godfrey, Wissam Moustafa Zanhar
St. Michaels EMS Person of Interest Stephen Onyechi
Ballet Healthcare Person of Interest Ededem Edem (Edison Medical Corporation)
Special Care Person of Interest Caroline Njoku
UBA Holdings Person of Interest Ezinne Ubani Slide 28
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Shared Beneficiaries Family Health Services and CKCC Home Health
Family Health Services DME and CKCC Home Health
Robert Hill
Robert Hill
Deborah Fenroy
Deborah Fenroy
Mitzie Wilson
Elbert McNeil
Elbert McNeil
Iva Parkhill
Iva Parkhill
Ester Dickey
Jeff Ferguson
Clarence Green
Ester B. Dickey
Esther Hunter
Clarence Green
Lee Parkhill
Esther Hunter
Ruth Johnson
Lee Parkhill
Gerald Daniels
Carolyn Schexnayder
Ruby Haywood
Ruth Johnson
Lulla Brooks
Gerald Daniels
Norma Lezine
Ruby Haywood
Morris Paul
Lulla Brooks
Zelda Paul
Norma Lezine
Dorothy Anderson
Morris Paul Zelda Paul Copyright © 2013 MarkLogic® Corporation. All rights reserved. Slide 29 Dorothy Anderson
Ubani - Medicare Indictment Medicare Indictment Document - June 20, 2010 Clifford Ubani, Ezinne Ubani, and Princewill Njoku (registered nurse) owners of Family Health Services Mary Ellen (vocational nurse) (recruiter) Rolondae Mitchel Straughter (recruiter) Caroline Njoku (recruiter) Michelle Turner (recruiter) Cynthia Garza Williams (recruiter)
Slide 30
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Use Case 2: Link Analysis Devon Michael Spicer Multiple NPI Brainex Consulting 1962702597
7447 Harwin Drive, Suite 217 Houston, TX 77036
Mindex Consulting 17506811466
POC Value – Linking Devon Michael Spicer multiple NPI Slide 31
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Indictment Link – Devon Michael Spicer Mobility Store
Devon Michael Spicer
Indictment
Adding external data sources – Indictments – elevates Devon Michael Spicer as Person of Interest and need to further examine his associations and ties to other Medicaid Providers
Slide 32
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Persons of Interest – Verona Spicer
Slide 33
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Linking Known Entities to Examine Aberrant Behaviors and Relationships Verona Spicer
Invision Pros Devon Michael Spicer
Elite Care Medical Services 7447 Harwin Drive, Suite 109 Houston, TX 77036 1740298868
Brainex Consulting 1962702597 Mindex Consulting 17506811466
POC Value – Linking Claims and Non Claims Data (Dunn & Bradstreet) to find Suspect relationships and billing behaviors Slide 34
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Use Case 3: Using Density Map to find “unknown” patterns and risk behaviors 8303 Southwest Freeway, Houston, Texas 7 Home Health Agencies – Do they share any beneficiaries – Any beneficiaries billed with aliases – High billable procedure codes Coastal Medical Services – 1144305004 509249318 DANIELLE HAMILTON Coastal Medical Services 1144305004 510322382 DANNY HAMILTON Coastal Medical Services 1144305004 513606514 DE A HAMILTON Coastal Medical Services Houston Optimum Care Associates – 1144305004 512127894 ROBERTHA COX Coastal Medical Services 1740300540 512127894 ROBERTHA COX Houston Optimum Care Associates – Suspicious Doctor – Doctor Emanuel Nwora – Vestibular Diagnostic Codes – one patient billed 800 times Slide 35
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
What’s Next More States, more claims, more fields provided Enrichment of provider’s notes
A better format for data delivery Additional Data Sources Access DMV and home ownership records Hadoop integration Enhance algorithms with profiles – run as UDFs Leverage Semantic Tech Store Provider to Provider and Beneficiary to Provider as triples Provider and Beneficiary profiles as hierarchical docs in XML Traverse triples graph for relationships and focus using queries on profiles
Slide 36
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
And There’s Too Much Waste Estimated Waste in US Healthcare Spending
Unnecessary Care Fraud & Abuse Administrative Inefficiency Medical Mistakes Preventable Conditions
Slide 37
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
Any Questions?
Slide 38
Copyright © 2013 MarkLogic® Corporation. All rights reserved.
For More Information Mike Doane, Technical Director Federal Healthcare
[email protected]
Slide 39
Copyright © 2013 MarkLogic® Corporation. All rights reserved.