Abstract
In order to plan the appropriate number of beds in a regional medical plan, it is essential to estimate accurately the number of future patients. Therefore it is necessary to clarify what kind of patient chooses which medical institution for each residential area. In this study, we propose a method to clarify the structure of the patient' hospital selection from the large regional medical information: Community health care data bank. When a patient chooses a hospital visit, we can assume that the patient considers not only the utility of candidate institutions, but also the distances from his/her home to these institutions. Furthermore, the influence of the latter might be different by their attributes, such as their gender and age. Then we formulate their utility structure by the multinomial logit model. Using the model estimated from a large dataset of receipts, we can estimate the selection probability of patient. We estimate an example model in a prefecture. First, we extract data of 10 major hospitals from the data bank, and we estimate the utility model for cancer patients. From the coefficients of the estimated model, we reveal that the patient who is away from all hospital tends to choose the largest university medical institution, and that the high age patient has a large probability to go to a hospital near his home. If we are able to clarify the structure of patients' share, combining with future population composition, we can estimate the future number of patients of each medical institution.
