چارچوب پروسه استخراج برای تشخیص کلاهبرداری و سوء استفاده از خدمات بهداشتی

People rely on government-managed health insurance systems, private health insurance systems, or both to share the expensive healthcare costs. With such an intensive need for health insurances, however, health care service providers' fraudulent and abusive behavior has become a serious problem. In this research, we propose a data-mining framework that utilizes the concept of clinical pathways to facilitate automatic and systematic construction of an adaptable and extensible detection model. The proposed approaches have been evaluated objectively by a real-world data set gathered from the National Health Insurance (NHI) program in Taiwan. The empirical experiments show that our detection model is efficient and capable of identifying some fraudulent and abusive cases that are not detected by a manually constructed detection model.

Healthcare has become a major focus of concern and even a political, social, and economics issue in modern society. The medical expenditure required to meet public demand for high-quality and high-technology services is substantial. This phenomenon is likely to become more widespread and more intense due to the increasing average lifespan and decreasing birth rates of humans in many societies. People rely on health insurance systems, which are either sponsored by governments or managed by the private sector, to share the high healthcare costs. Such an intensive need for health insurance has resulted in fraudulent and abusive behavior becoming a serious problem. According to a report (Health Insurance, 1992) published by the General Accounting Office in the US, healthcare fraud and abuse costs the US as much as 10% of its annual spending on healthcare, representing US$ 100 billion per year. Similar problems have been reported for the health insurance programs of other developed countries (Lassey, Lassey, & Jinks, 1997). The above figures indicate that detecting healthcare fraud and abuse is imperative. Detecting healthcare fraud and abuse, however, needs intensive medical knowledge. Many health insurance systems rely on human experts to manually review insurance claims and identify suspicious ones. Most of the computer systems that are intended to help detect undesirable behavior require human experts to identify a set of features so as to develop the core of detection models. This results in both system development and claim reviewing being time-consuming, especially for the large government-sponsored national insurance programs in countries such as France, Australia, and Taiwan. In this research, we propose a process-mining framework that utilizes the concept of clinical pathways to facilitate the automatic and systematic construction of an adaptable and extensible detection model. We take a data-centric point of view and consider healthcare fraud and abuse detection as a data analysis process. The theme of our approach is to apply process-mining techniques to gathered clinical-instance data to construct a model that distinguishes fraudulent behaviors from normal activities. This automatic approach eliminates the need to manually analyze and encode behavior patterns, as well as the guesswork in selecting statistics measures. The proposed framework is evaluated via real-world data to demonstrate its efficiency and accuracy. This paper is organized as follows. Section 2 examines in more detail the problem of healthcare fraud and abuse. Related research efforts are also reviewed. Section 3 presents the process-mining framework of our research. 4 and 5 describe in detail the methods for building the detection model. Section 6 presents the results of an evaluation of the detection power of the model on a real-world data set gathered from the National Health Insurance (NHI) program in Taiwan. Section 7 concludes the current work and discusses the directions of future research.

In this research, we have outlined a framework that facilitates the automatic and systematic construction of systems that detect healthcare fraud and abuse. We investigated the mining of frequent patterns from clinical instances and the selection of features that have higher discrimination power. The proposed approaches have been evaluated objectively using a real-world data set gathered from the NHI program in Taiwan. The empirical experiments show that our detection model is efficient and capable of identifying some fraudulent and abusive cases that are not detected by a manually constructed detection model. This work could be extended in several directions. First, the handling of noisy data in this context remains a challenging problem. Second, there are many cost factors in healthcare fraud and abuse detection, and so building detection models that can be easily adjustable according to site-specific cost policies is important in practice. Finally, we believe that it is beneficial and natural to integrate a healthcare fraud and abuse detection system with a cost-restricted system, so that the detection system can communicate with the cost-restricted system when determining the appropriate actions to take.