Document Type

Thesis

Publication Date

4-2015

Advisor

Yu Zhang, Computer Science

Abstract

On October 1st, 2015, the tenth revision of the International Classification of Diseases (ICD-10) will be mandatorily implemented in the United States. Although this medical classification system will allow healthcare professionals to code with greater accuracy, specificity, and detail, these codes will have a significant impact on the flavor of healthcare insurance claims. While the overall benefit of ICD-10 throughout the healthcare industry is unquestionable, some experts believe healthcare fraud detection and prevention could experience an initial drop in performance due to the implementation of ICD-10. We aim to quantitatively test the validity of this concern regarding an adverse transitional impact. This project explores how predictive fraud detection systems developed using ICD-9 claims data will initially react to the introduction of ICD-10. We have developed a basic fraud detection system incorporating both unsupervised and supervised learning methods in order to examine the potential fraudulence of both ICD-9 and ICD-10 claims in a predictive environment. Using this system, we are able to analyze the ability and performance of statistical methods trained using ICD-9 data to properly identify fraudulent ICD-10 claims. This research makes contributions to the domains of medical coding, healthcare informatics, and fraud detection.

Share

COinS