Healthcare insurance fraud influences not only organizations by overburdening the already fragile healthcare systems, but also individuals in terms of increasing premiums in health insurance and even fatalities. Identifying the behavioral characteristics of fraudulent claims can help shed light on the development of artificial intelligence and machine learning technologies to detect fraud in health information system research. In this paper, a theoretical model of medical insurance fraud identification is proposed, which characterizes the judgment variables of fraud from the three dimensions of time, quantity, and expenses. The model is verified with large-scale, real-world medical records. Our study shows that, in comparison with claims made by normal people, fraudulent claims usually have a greater frequency of hospital visits, and more medical bills, accompanied by higher amounts of medical expenses. An interesting discovery is that the price per bill for fraudulent cases is not statistically different from the normal cases.