Methodological problems of prospective natural trials are discussed. The application of adequate methods is demonstrated by analysing the effects of antipsychotic medical treatment on the probability and the costs of inpatient episodes of 307 patients with schizophrenia for 2.5 years. Selection bias is controlled by the propensity score method. Effects of medical treatment on the incidence of inpatient episodes are analysed by means of a random-effect logit model. Effects of medical treatment on the costs of inpatient episodes are analysed by means of a random-effect Tobit model. Results of the study show that antipsychotic medical treatment reduces the probability and the costs of inpatient treatment irrespective of the type of neuroleptic drugs. The hypothesis of a higher effectivity of atypical neuroleptics in comparison to conventional antipsychotics regarding the reduction of the incidence and the costs of inpatient treatment was not supported. Beyond the effects of medical treatment the study results indicate that improving social support and competence for coping with life events as well as providing community mental health services may reduce the risk and the costs of inpatient episodes.