Using Database Mining Techniques to Detect Fraud Within Health Care Insurance Databases

The GAO has reported to Congress that the loss due to fraudulent health care activity amounted to an estimated $84 billion in 1992. The ability to detect fraud and abuse is critical to the reduction of health care costs. The claims processing system will soon undergo rapid change with the introduction of rapid payment transaction processing systems. The emphasis on rapid turn around combined with increases in data volume will exacerbate the fraud and abuse detection problem. We propose to prove the feasibility of an automated high speed, pre-pay system for processing claims data to detect instances and patters of fraud and abuse. MRJ developed this technology for classified specail-access projects within the Department of Defense, the Intelligence Community, the Department of Justice, as well as within Department of the Treasury. MRJ has demonstrated the potential of using prallel processing for very high speed analysis of patterns of abuse to the law enforcement community. We will develop the system using knowledge mined from historical claims databases and test its heuristics on actual claims data. Blue Cross/Blue Shield of South Caroline (BCBSSC) has expressed interest (Appendix I) and will provide access to its experts and data. Anticipated Benefits: The goal of the project will be to demonstrate the technical feasibility of an automated high speed near real time, pre-pay system for processing helath care claims data to detect instances and patterns of potential fraud and abuse. The fraud detection technology will also lend itself to other types of fraud -- auto insurance, life insurace, phone fraud, mail fraud, etc.