OBJECTIVE: Assess diabetes incidence for individuals treated with antipsychotics. METHODS: Medical/pharmacy claims from a large insured population were analyzed to identify 5631 individuals initiated on typical (n = 2326) or atypical (n = 3305) antipsychotics between October 1, 1996 and December 31, 1998; had no antipsychotic use; had no diabetes diagnosis and/or antidiabetic use in the year prior. Patients were followed to the first event including: new-onset diabetes; antipsychotic augmentation/switch/discontinuation; insurance disenrollment; end of the data. Diabetes was defined as receiving two diabetes diagnoses (ICD9-CM: 250.xx) and/or one antidiabetic prescription. Additionally, incidence of treated diabetes (received antidiabetic prescription) was examined. Cox proportional hazard models were used to compare the hazard ratios (HR), controlling for potential confounding factors. RESULTS: Crude diabetes incidence was 1.76% and 1.97% for typical and atypical cohorts, respectively (2.04% and 1.95% for olanzapine- and risperidone-treated patients, respectively). Rates of treated diabetes were 1.72% for typical and 1.94% for atypical cohorts; 1.95% for both olanzapine and risperidone cohorts. Among the treatment groups, adjusted HR (95% confidence interval): atypical-versus-typical: 1.13 (0.75-1.69); olanzapine-versus-typical: 1.14 (0.67-1.95); risperidone-versus-typical: 1.12 (0.72-1.75); and olanzapine-versus-risperidone: 0.96 (0.56-1.65). Similar results were observed for treated diabetes incidence. CONCLUSION: Diabetes incidence was not statistically significantly different for atypical versus typical, although typical-treated patients had numerically lower incidence. There was no significant difference between olanzapine and risperidone.