A Novel Approach to Uncover Health Care Frauds through Spectral Analysis

Fraudulent activities exist in many areas of businesses and our daily lives, and have been studied in many research articles. Such activities are most prevalent in credit card transactions, telecommunications, network intrusions, finance and insurance, and scientific applications. However, not much emphasis has been put on healthcare fraud detections and hence the research in this area has been very limited. The lack of research is not because the loss in health care fraud is insignificant, rather, the loss is significantly huge. According to NHCAA estimates in 2011, the financial losses due to health care frauds are in the tens of billions of dollars each year. Because of privacy concerns, the health care data are seldom released to the research communities. In this article, we will use a de-identified health claims dataset to propose and test a novel fraud detection technique based on a community detection algorithm through spectral analysis. Our result shows good performance and promising results in terms of identifying potentially fraudulent patterns in potental physician collusions. We evaluated our findings by going to detailed discussions of potential fraudulent scenarios. This community detection algorithm and a list of other similar algorithms could be expanded to other areas of fraud detection problems.