Goals: The goals of this study are 1) to determine causes and patterns of relapse for a cohort of “revolving door” schizophrenia inpatients, and 2) to assess the feasibility of starting a new psychopharmacologic intervention before discharge, either depot therapy or an atypical antipsychotic. Methods: Consecutive admissions to an acute inpatient unit in New York City were screened for “revolving door” criteria. Patients had to have a primary diagnosis of schizophrenia or schizoaffective disorder and have either 1) two hospitalizations in the last year, or 2) three hospitalizations in the last three years. Patients were then assessed for probable causes of relapse for the index and prior two hospitalizations. Treatment selection, based on this information, was trichotomized to: 1) oral conventional antipsychotic, 2) depot conventional antipsychotic (either haloperidol or fluphenazine decanoate), or 3) atypical antipsychotic (either risperidone or clozapine). Results: Sixty-three out of 131 screened admissions met the above revolving door criteria. They were indeed “revolving”, having an average of 1.3 hospitalizations per year over the last 3 years and were only out of the hospital for five months (median) before index admission. The treatment selection process was hampered by lack of information about events leading to relapse, and by the lack of outpatient participation in the medication selection process. Of the 50 patients with complete histories about precipitants for the index episode, the most common reason for rehospitalization was judged to be medication non-compliance (n = 25; 50%), followed by medication nonresponse (n = 13; 26%). Not surprisingly, medication recommendations were closely linked to the assessed reason for relapse (depot therapy [n = 27; 49%] with medication non-compliance; atypical antipsychotic [n = 20; 37%] with medication nonresponse [X2 = 26.9, p<.001]).These two recommendations were implemented before discharge for about one-half of the cases. Patient refusal was a relatively greater problem for depot recommendation while constraints in the outpatient environment were more problematic for patients recommended for atypical antipsychotics. Conclusions: Medication noncompliance and medication nonresponse, in that order, were judged to be the most common causes of relapse for “revolving door” inpatients. Both depot therapy and atypical antipsychotics were commonly recommended and ultimately accepted by about 2/3rds of patients. Choice between depot and atypical was driven by the assessed cause of relapse. In summary, it seems possible to identify “revolving door” inpatients, and to target specific medication interventions within the time frame of an acute inpatient admission.