This article presents a series of models for explaining hospital utilization. The models belong to the multinomial logit class, which relate the probability of hospital selection to factors that include travel time between patients and hospitals, hospital attractiveness factors, physician characteristics, and patient characteristics. Separate models for different population groups (sex, age, socioeconomic classes) and medical services (general medicine, obstetrics, pediatrics, psychiatry, and general surgery) are estimated and compared with the general hospital model. The results, which are based on an extensive data base and new estimation procedures, are superior to those of past studies. A number of differences between utilization patterns for the various populations are also observed.