(1 in 4) Fraud (43%) (12.6 M) 2013 Identity Theft Report Javelin - - PowerPoint PPT Presentation

THE THE NEW NORM NEW NORM – IDENTIT IDENTITY Y THE THEFT FT AND AND GO GOVERNMENT VERNMENT FRA FRAUD UD

Data Breach (1 in 4) Identity Fraud (12.6 M) Government Fraud (43%)

2013 Identity Theft Report – Javelin Strategy & Research FTC Consumer Sentinel Network Report - CY 2012

TODAY THE FUTURE DATA SIZE

AN ANAL ALYT YTICS ICS IN IN THE THE BIG BIG DATA A ER ERA

THE “HOLY GRAIL” – PREPAYMENT DETECTION Persistent Myth? Real But Distant Future?

WHA WHAT T DOES DOES PRE PREPAYMENT YMENT LOOK OOK LIKE? LIKE?

Provider Enrollment Recipient Eligibility MMIS - Provider MMIS- Pharmacy MCO

ADJUSTING ADJUSTING THE THE TAR ARGET GET

• Complete data
• Timely data
• Detect outliers without

patterns

• Predictive models on

single claims

• Relies on persistent

analysis

• Utilize key decision points
• Uses parallel analytical

models for early detection

• Quick intervention
• Light touch interventions

vs.

PRE PREPAYMENT YMENT COST COST AVOID OIDANCE ANCE

Alert Generation Process

Network Analysis Network Rules Network Analytics Alert Administration Rules and Analytics Anomaly Detection Predictive Models Fraud Data Staging Intelligent Fraud Repository

Exploratory Analysis & Data Transformation

Operational Data Sources

Case Management Alert Management & Reporting

Learn and Improve Cycle

Recipients Providers Claims External Data

1 2 3 4 5

BEST PRACTICE – OPERATIONALIZE ANALYTICS

Alert Rules

BES BEST T PRA PRACTICE CTICE - ENT ENTITY ITY RES RESOL OLUTION UTION FR FROM OM MUL MULTIP TIPLE LE DATA S A SET ETS

Claim data sets

• Aggregate claims for entities to be used in outlier

detection at entity level

• Benefit: Analysis by provider ID might not be

detected as outlier, but analysis by entity ID could show an obvious outlier Fraud data set

• Some providers matched to be the same entity as

a known bad provider

• Benefit: Discovery of bad providers disguised by

masking identities Linkage Analysis

• Link entities together based on attributes
• Used to create both hard and fuzzy relationships

SOLUTION

• New Data Warehouse with 1 Billion+ Rows
• Pilot Followed by Successful Rollout
• Over 40 Custom Analytical Scenarios

HEALTH CANADA – DEVELOPING IN-HOUSE SURVEILLANCE OBJECTIVES

• Move Detection In-House
• Develop Surveillance Approach to

Pharmacies, Prescribers and Clients

RESULTS

• Multi-Million $ Savings
• Continuous Surveillance

and Rapid Intervention

• Medicine Cabinet

Improving Safety

Programmatic Eligibility Data MMIS Medicaid Data

Investigative Unit

MCO Encounter Data MCO Encounter Data MCO Encounter Data MCO Encounter Data

Third Party Liability Contractor Medicaid Recovery Audit Contractor Medicaid Integrity Contractor Program Integrity Contractor

Other Data SNAP TANF Child Support

Program Integrity Data Access Data Access Data Access Data Access Data Access Program Integrity Program Integrity Program Integrity Program Integrity

BEST PRACTICE 3 – ENTERPRISE APPROACH

Consolidated View of Data for Cross-Program Fraud Detection

Common Technology Framework

Prepare Data Model Optimization Monitor & Report Alert Generation Decision Flow

BEST PRACTICE 3 – ENTERPRISE APPROACH

Best Practice – An Enterprise Approach to Program Integrity

SOLUTION

• Data From 15 Programs, 5 Agencies
• Phased Approach with Seven Years of Data.
• Single, Integrated Scoring and Ranking

RESULTS

• 80% Drop in Triage Time
• Decreased False Positives
• 50% Increase in $/Case and 30:1 ROI
• Light Touch Interaction and Policy Effects

WASHINGTON STATE WORKERS’ COMPENSATION

OBJECTIVES

• Maximize Impact of Limited Staff
• Reduce False Positives
• Improve Cross-Program Detection
• Single Detection System

BEST PRACTICE – HYBRID FRAUD DETECTION

Hybrid Hybrid Det Detection ection Engine Engine

Predictive Models Anomaly Detection Business Rules Data Matches Text Mining Link Analysis

BEST BEST PRA PRACTIC CTICE E – TAR ARGET GET NETW NETWOR ORKS KS

• High risk and ROI
• Connect various

participants

• Multiple data sources

and link types

• Network risk multiples
• Utilize geospatial
• verlay to investigate

BES BEST T PRA PRACTI CTICE E – VIS VISUALIZA ALIZATION ION AND AND AD AD-HOC HOC AN ANAL ALYSIS SIS

• Visualiz

isualize e ca case se tr tren ends ds

• Explor

Explore to e to iden identify tify

• u
• utli

tlier ers

• “Hotspots”

ar are e e eviden vident

• Dete

Determine mine co correla elation tions

• Per

erson sonal and al and un unit me it metrics trics

SOLUTION

• Integrated Analytics Platform and Multi-Year

Analysis

• Parallel Analytics for Continuous Monitoring
• Pre-Refund Analytics

OBJECTIVES

• Combat Identity Theft and “Ghost Returns”
• Analyze Big Data on a Single Platform
• Prioritize Audits and Investigations

RESULTS

• Pre-Refund Detection
• Pervasive Real-Time Analytics – 14M

Returns on Peak Day

• New Treatment Streams

Billion Dollar Savings in Partial Year

LEARNING FROM OTHER GOVERNMENT PROGRAMS

IRS RETURN REVIEW PROGRAM – RETHINKING ANALYTICS AND INTERVENTION

Final Final Thou houghts ghts