Multi-stage methodology to detect health insurance claim fraud

Marina Evrim Johnson, Nagen Nagarur

Research output: Contribution to journalArticlepeer-review

16 Scopus citations

Abstract

Healthcare costs in the US, as well as in other countries, increase rapidly due to demographic, economic, social, and legal changes. This increase in healthcare costs impacts both government and private health insurance systems. Fraudulent behaviors of healthcare providers and patients have become a serious burden to insurance systems by bringing unnecessary costs. Insurance companies thus develop methods to identify fraud. This paper proposes a new multistage methodology for insurance companies to detect fraud committed by providers and patients. The first three stages aim at detecting abnormalities among providers, services, and claim amounts. Stage four then integrates the information obtained in the previous three stages into an overall risk measure. Subsequently, a decision tree based method in stage five computes risk threshold values. The final decision stating whether the claim is fraudulent is made by comparing the risk value obtained in stage four with the risk threshold value from stage five. The research methodology performs well on real-world insurance data.

Original languageEnglish
Pages (from-to)249-260
Number of pages12
JournalHealth Care Management Science
Volume19
Issue number3
DOIs
StatePublished - 1 Sep 2016

Keywords

  • Classification
  • Decision trees
  • Fraud
  • Insurance
  • Provider profiling
  • Risk

Fingerprint

Dive into the research topics of 'Multi-stage methodology to detect health insurance claim fraud'. Together they form a unique fingerprint.

Cite this