University of Twente Student Theses


Outlier based predictors for health insurance fraud detection within U.S. medicaid

Capelleveen, Guido Cornelis van (2013) Outlier based predictors for health insurance fraud detection within U.S. medicaid.

[img] PDF
Abstract:This paper describes an e�ective method of outlier based predictors for health insurance fraud detection that identi�es suspicious behavior of health care providers. Fraud and abuse on medical claims became a major concern for health insurance companies last decades. Estimates made for the studied U.S. Medicaid health insurance program is that up to 10% of the claims are fraudulent. Unsupervised data mining techniques such as outlier detection are suggested to be an e�ective predictors for fraud detection and should be used to support the initiations of audits. A method, based on comparative research, fraud cases and literature study has been proposed. We evaluated the method, by applying the method in a real life case study, were behavioral metrics were designed and 14 analytic experiments were built using outlier detection. The analysis ran on dental claim data and showed promising results. The proposed methodology enabled successful identi�cation of fraudulent activity in several cases; however linking these identi�ed incidents with irrefutable de jure fraud proved to be a di�cult process. From 17 top suspicions analyzed, we reported eventually 12 of those to o�cials, a precision rate of approximately 71%. In the two interviews conducted with Medicaid Fraud Experts, experiences were gained on requirements for the design of the analytics and an e�ective implementation of the method. We found that outlier based predictors are not likely to succeed as fraud classi�cation technology, though it explored an important role as decision supportive technology for resource allocation of fraud audits.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:85 business administration, organizational science
Programme:Business Administration MSc (60644)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page