Leveraging Big Data to Reduce Workers’ Compensation Fraud
Identified $150+ million in fraud; delivered evidence in minutes instead of days
Workers’ compensation fraud costs the industry $5 billion annually. This case study illustrates how an established state-funded workers’ compensation agency successfully reduced insurance fraud using big data technology. The respected agency has over 100,000 policy holders, and collects and pays out billions in premiums and claims annually. The agency has an in-house fraud investigation unit staffed with hundreds of lawyers who investigate and prosecute fraud.
The agency encountered these challenges when proactively attempting to reduce fraud with its 1990's legacy data warehouse/business intelligence platform:
- A laborious system - new data requests were sent to business analysts, who would build and run SQL queries and compile reports
- Queries executed against a traditional data warehouse took hours to retrieve results
- Manual process to fetch the original bill scans including writing down record numbers, then going to the appropriate server and retrieving them individually
Because of these issues, fraudsters could manipulate billing faster than the investigators could audit.
The agency selected Search Technologies to architect a modern analytics platform built on an open source big data stack that combined:
- Cloudera Hadoop for streamlined content processing and indexing
- Cloudera Search for faster and easier access to data
- Query Processing Language (QPL) for aggregating query results for real-time charts and graphs
- Custom user-friendly, intuitive graphical interface
This anti-fraud investigator's desktop helps reduce costs and improve scalability. Investigators no longer have to rely on business analysts to formulate queries (SQL) and data is available in real-time. Faceted search allows investigators to drill down and retrieve individual claim scans to build a legal case.
By leveraging big data to flatten complex databases, the agency realized immediate ROI within a few months.
- Ability to analyze 10+ million of claims, 100+ million of bill line details, and related records to compute key fraud indicators and create "red flag" data sets for suspicious activities rather than just conducting a statistical sampling
- Ability to export billing scans and compile evidence to prosecute $100+ million in fraud
- Kudos from judges as evidence is delivered in minutes, compared to the industry's average of three days
For additional details on this case study, also see our entry in the Cloudera Vision Blog tilted Empowering an Organization to Discover Fraudulent Medical Claims Effectively.