Investigate how selection criteria and statistical model specifications affect time to all cause discontinuation (TTAD) comparisons across alternative antipsychotics using retrospective database analyses.

231,635 episodes of antipsychotic therapy were identified using data from the California Medicaid (Medi-Cal) program. A series of regression models were estimated for TTAD that altered selection criteria and the list of independent variables used in the models.

Unadjusted TTAD (in days) for typical antipsychotics, olanzapine, risperidone and quetiapine were 92, 175, 182 and 177, respectively. TTAD achieved by patients using conventional antipsychotics was consistently shorter than TTAD achieved with atypical antipsychotics, but estimates varied from -96 days to -44 days depending on selection criteria and model specification (p<0.0001 relative to olanzapine). TTAD using risperidone or quetiapine appeared to be superior to olanzapine in simple models (+11 to +13 days, p<0.000), while virtually no differences across atypical antipsychotics were found when the analysis was restricted to patients with schizophrenia and more complete model specifications were employed. Specifically, screening for schizophrenia reversed risperidone's advantage over olanzapine from +6 days (p<0.0001) to -1.4 days (p>0.05). TTAD results favoring quetiapine over olanzapine were reversed from +7 days (p<0.0001) to – 0.4 days (p>0.05) when covariates for episode type were included in the model.

Differences in duration of antipsychotic therapy exist across diagnostic groups and episode type. Differences also exist in the diagnostic and episode mix across drugs. Therefore, disaggregated patient samples and expanded model specifications provide more accurate estimates of differences in TTAD.