Disease Prediction And Transfer Behavior Analysis Based On Reimbursement Data

Health insurance plays distinctly important role in health care. The reim-bursement data includes disease names, costs, visit time, visit site, and demo-graphic information etc.. By digging into the reimbursement data,many things can be done, such as analyzing the pathogenesis of diseases, predicting the devel-opment trends of the diseases, analyzing the visiting modes, assessing the effect of policies in practice. It will provide practitioners in medical field with enlighten-ment for further research. It will also remind people of the precautions facing with diseases. And lastly, from the perspective of policies, it will function as important reference for medical managers.This paper mainly explore the reimbursement data from two perspectives.The first is the potential relationships among diseases. By analyzing similar his-tory of patients' visits, associations between diseases can be found. This will provide guidance for the control and treatment of the diseases. Secondly, as the data is full sample data, this paper explored the whole treatment process in the observation period of the patients. This can be used to seek for the affecting factors of treatment behaviors.Traditional prediction methods usually focus on the prognoses of one or sev-eral particular diseases. These are generally broad research and will not provide personal predictions. From another perspective, traditional research take a long time to observe and record the patients' conditions during the experiments. It would consume much time and money. The CAC method, as provided in this paper, combining several methods in data mining, can be used to predict multiple diseases. It can also provide personal prediction for every patient. This work only utilize disease names and several demographic variables to finish this task.The paper predict for acute patients and chronic patients respectively according to their characteristics. The results show that for 71% of acute patients and 82%of chronic patients, their future conditions are predictable. This method that can be used to make a multiple diseases and personal prediction is rarely found in literature. The result is improved compared with few known research works. A case study was made with the proposed method on testing data set. The results show that the actual situation of the patient is highly in accordance with our predictions.The visiting behavior that is explored here is mainly the transfer behavior of patients when seeing a doctor. The medical system in our country is a huge and complicated hierarchical one. In theory, the choices of hospital for patients is not limited. In practice, however, for different medical groups, there are some issues like special policies, hospitals factors and personal factors that will influence the choice. Under the comprehensive condition, how will the patients choose the hospitals, and how will different policies have different influences on the choices,are concerned problems of the government, the hospitals and the patients them-selves. This paper categorize the different transfer behaviors and put forward a clear definition for it. Based on the definition, new transfer models were proposed and transfer modes of the patients were analyzed. Main factors that will influence the transfer behaviors were found. Econometric methods were used to analyze the effect of each factors. Some useful laws that can be used as guidance for the practice were found. Together with the huge and unique health care system in our country,the transfer behavior is extremely complicated. This work will have a certain inspiration in related areas and can provide new ideas for later research.