The goal of this study was to determine if demographic or clinical factors collected at baseline on patients treated with clozapine would increase the risk of having clozapine discontinued for (a) lack of response, (b) side effects, (c) noncompliance, (d) concomitant illness, or (e) death. The subjects were 805 patients treated with clozapine at 96 Department of Veterans Affairs Hospital System facilities. Multiple logistic regression was used to determine if any of the baseline variables predisposed patients to discontinuation from treatment. Factors which were studied include age, race, history of inadequate response to traditional neuroleptics, history of substance abuse, and DSM-III-R Axis I diagnosis. Of the 805 patients started on clozapine 167 (20.7%) were discontinued from treatment. The only significant variable in the logistic regression model was race. This study finds that African American patients are more likely to have clozapine discontinued than non-African American patients, and there is a trend for prior history of inadequate response to traditional neuroleptics to predict clozapine discontinuation. We found no effect of substance abuse or dependence, diagnosis, or age on outcome in the overall patient group. In a post hoc analysis the African American patients had a significantly lower baseline white blood count than the non-African American patients, which could have explained the difference in clozapine discontinuation. The findings of this study support further investigation into the causes of ethnic differences in treatment outcome with clozapine.