Assessment of antipsychotic-related risk of diabetes mellitus in a Medicaid psychosis population: sensitivity to study design.