The associations between psychosis, antipsychotic drugs and diabetes mellitus have not been precisely defined but it has been repeatedly suggested that atypical antipsychotics are more likely to give rise to diabetes than are conventional drugs. This belief is largely based on healthcare database analyses which, in part, rely on the assumption that all cases of diabetes are identified in practice. We examined records of 606 hospitalized patients receiving antipsychotic treatment and found an apparent prevalence of diabetes and impaired fasting glucose of 6.4%. From this sample of patients, we investigated 166 patients (fasting blood samples) who were not known to have any disorder of glucose homeostasis and identified 10 cases of impaired fasting glucose and nine cases of diabetes mellitus (11.4% of those tested). Nine of these cases had documented evidence of previous testing for diabetes. Apparent prevalence of diabetes and impaired fasting glucose was 16.9% in those tested in practice or the study. Diagnosis was significantly associated with age [odds ratio (OR) 1.02] and treatment duration with current drug (OR 1.01). Adjusted ORs of diagnosis were not significantly different for any atypical antipsychotic compared with conventional drugs. It is concluded that there was a clinically significant prevalence of undiagnosed diabetes and impaired fasting glucose in those individuals receiving antipsychotics. Importantly, database analyses may underestimate the true prevalence of diabetes in similar populations and erroneously ascribe increased risk to certain drug treatments.