中国健康保险欺诈：理论分析与实证研究

Insurance fraud has been the main obstacle of the development of health insurance in China nowadays. It not only violates the utmost good faith principle seriously, but also threatens the safety of medical insurance fund..Based on the investigation and accumulated empirical findings, this paper analyzes the situation of the health insurance fraud in China, studies the inherent characteristics and laws of health insurance fraud and seeks the cause of insurance fraud from the aspect of information asymmetry and medical insurance system. In the light of the fraud caused by moral hazard, a statistical classification model for claim data and an outlier data mining algorithms are built via a serial combination of the rough set and the artificial neural networks. As for the fraud caused by adverse selection, a multiple regression model and Tobit regression model are built. Based on these models, some empirical research on fraud recognition is performed. Then, the actuarial model for measurement the risk of health insurance fraud is built and the basic medical insurance for urban employees is selected as an example to illustrate the application. Finally, some suggestions to health insurance's anti-fraud mainly on the study of technology and optimizing medical insurance system are given to improve the ability of fraud recognition and anti-fraud in health insurance agencies.

现今，保险欺诈已经成为我国健康保险业顺利发展的主要障碍，它不仅严重侵害了保险经营的最大诚信原则，而且还对医保基金安全构成了巨大威胁。.通过对中国健康保险欺诈问题的调查研究并基于已有的实证分析结果，分析我国健康保险欺诈现状及其内在特征与规律，并主要从信息不对称和医疗保险制度解析的视角分析健康保险欺诈产生的原因；针对道德风险引致的欺诈，一方面建立索赔数据的统计分类模型，另一方面将粗糙集与人工神经网络进行串行结合，构建基于粗糙集与神经网络的异常数据挖掘算法；针对逆向选择引致的欺诈，建立复回归和Tobit回归模型；在此基础上，利用上述模型进行欺诈识别的实证研究；以城镇职工基本医疗保险为例，建立精算模型，模拟度量欺诈风险；最后，从易诱发欺诈的制度因素入手，主要在医疗保险制度优化与改进层面和技术手段层面上研究给出健康保险反欺诈的对策建议。研究有助于提升健康保险机构识别欺诈风险和反欺诈的能力与水平。

现今，保险欺诈已经成为我国健康保险业顺利发展的主要障碍，它不仅严重侵害了保险经营的最大诚信原则，而且还对医保基金安全构成了巨大威胁。.本课题通过对中国健康保险欺诈问题的调查研究并基于已有的实证分析结果，分析了我国健康保险欺诈现状及其内在特征与规律，从信息不对称和不完全契约等多个视角研究了健康保险欺诈产生的原因。针对道德风险引致的欺诈，分别采用LOGIT模型、BP神经网络模型、LOGIT-BP神经网络模型进行串行结合，以长期综合健康险、住院医疗保险为例构建基于统计分析与神经网络的异常数据挖掘算法用于欺诈识别；针对逆向选择引致的欺诈，以重大疾病保险为例，建立Tobit回归模型与BP神经网络模型，在此基础上，利用上述模型进行欺诈识别的实证研究；基于新农合欺诈损失数据，借鉴损失分布法原理，建立短期聚合风险精算模型，采用蒙特卡洛模拟方法对健康保险欺诈风险进行度量。最后，从易诱发欺诈的制度因素入手，主要在医疗保险制度优化与改进层面和技术手段层面上研究给出健康保险反欺诈的对策建议。结果表明，研究有助于提升健康保险机构识别欺诈风险和反欺诈的能力与水平。