Alleviating moral hazard in health insurance: evidence on commercial medical expense insurance fraud in China

YAO Yi SUN Qixiang LIN Shanjun FAN Qingzhu

【Abstract】Health insurance fraud is an extreme form of moral hazard. It is widely accepted that health insurance fraud increases inefficiency and inequality in the financing of health care expenditures and hinders the development of the health insurance industry. Fraud contributes to rising health insurance costs and results in significant social welfare loss. According to the Global Health Care Anti-Fraud Network, health insurance fraud is a worldwide problem suffered by both developed countries and developing markets. It is estimated that the annual total cost of health insurance fraud could be USD 260 billion, or 6% of global healthcare spending. In a developed market, major stakeholders work together to develop an advanced fraud detection system using abundant data and predictive analytics to provide efficient fraud management. In an emerging country such as China, the typical procedure to detect health insurance fraud still follows simple guideline criteria such as claim amount thresholds and largely relies on the experience and resources of individual claim adjusters to perform manual investigations. Both the efficiency and accuracy of China’s fraud detection would be improved dramatically with an automated fraud detection system. Despite the urgent need, we are aware of no study focusing on health insurance fraud in commercial medical expense insurance in China. We filled this gap in the literature and provided evidence on the factors that predict health insurance fraud in China. We built a theoretical model to analyze fraudulent behavior and applied it to China’s commercial health insurance market. With a dataset acquired from a leading company, we proposed several forms of a discrete choice model to identify the predictive factors of fraudulent claims. We used a probit model as our baseline and considered a probit model with weighted exogenous sampling and maximum likelihood estimation to address the oversampling of fraudulent cases and a probit model adjusted for the omission error of mislabeling a fraudulent claim as a legitimate claim. Our results show that several factors, including the hospital’s qualification, total medical expenditures and policy renewal status predict potential fraudulent claims for further investigation. To check the adequacy of our models, we reported the classification results. We chose the threshold for labeling a claim as fraudulent using a grid search framework. The in-sample and out-of-sample classification results show that more than 80% of the fraudulent cases can be detected while keeping the overall rate of correctly classified cases close to 60%. We have contributed to the literature in the following ways. First, we are the first empirical paper to use data on commercial medical expense insurance from China as a sample to detect fraud. This focus on China is in contrast to the literature, which often uses social health insurance data from developed markets. Second, we consider multiple forms of a discrete choice model, which is a type of supervised model. The predictive power is improved as a result of addressing the oversampling problem and omission error. Third, the literature focuses on detecting fraud from the supply side using the healthcare provider as the unit of analysis. We focus on the demand side of fraud by using a rich individual-level dataset to study the predictive factors of health insurance fraud. Our analysis helps insurers in China evaluate claims better and improve the efficiency and accuracy of claim management. In terms of policy implications, we have proposed the following. First, the government should improve the laws and legislations related to health insurance fraud and further increase the penalties for fraud. Second, the government should encourage and facilitate cooperation between the social and commercial health insurance industries in terms of data sharing and management to increase the efficiency of fraud detection. Third, the government should establish an anti-fraud entity to coordinate the work of the insurance industry, NGOs and regulatory committees.

【Keywords】 moral hazard; fraud prevention; commercial medical expense insurance;

【DOI】

Download this article

This Article

ISSN:0577-9154

CN:11-1081/F

Vol 55, No. 06, Pages 189-206

June 2020

Downloads:0

Share
Article Outline

Abstract