Workers Compensation outlier identification
3AI September 10, 2020
Leading US Insurer
Problem Statement
- The $ impact of medical provider fraud on US Insurers is enormous
- For the majority of insurers, efforts on medical claim fraud identification tend to be concentrated on (and limited to) opposite ends of the spectrum: (a) rule-based and transaction-level reviews
(b) SIU investigations - The sheer volume of invoices and the detail they contain, as well as the challenge of medical data can confound the process of manual claim review and allow fraudulent claims to filter through
Analytics Led Approach
- An analytical solution to fraud identification that focuses on providers rather than transactions, and looks for patterns in provider behavior
- The solution methodology is end-to-end: it ingests medical invoice data available with insurers, creates its own data assets and classifies outlier behavior using unsupervised machine learning techniques
- Resourcing with functional, technical and clinical experts; high-touch execution; risks proactively identified & mitigated
Business Impact
- Led to a larger (current) project encompassing a much larger span of data with a monthly refresh
- Exploring bolt-on analyses (provider collusion, provider preference/ referral, flagging for clinical review
- Delivered on-spec, on-time & with highly positive feedback
- Strong buy-in from Client Analytics and SIU organizations
Critical Success Factors
- Identified ~600 outliers among Company’s regular trading partners meeting conservative floor criteria
- Identified ~$8+ MM in anomalous claims (from abovementioned ~600 outliers alone, closed claims alone)