OBJECTIVES: To assess the demographic, diagnostic and cost differences at baseline among Medi-Cal patients with schizophrenia using olanzapine, risperidone, or typical antipsychotics. METHODS: Ambulatory patients with schizophrenia, who initiated antipsychotic drug therapy on olanzapine, risperidone or typical antipsychotics, after a break of more than 30 days from their previous antipsychotics, were identified. Of these, patients with data available for at least six months prior to and 12 months following the initiation date of these medications during March 1998-August 2001 were included. Baseline demographic, diagnostic and cost variables were compared among these three groups. RESULTS: The final sample comprised of 8022, 4909, 4867, olanzapine, risperidone and typical (O;R;T) patient episodes respectively. The episodes for which typical antipsychotics were initiated had the highest total costs in the six months prior to drug initiation (mean (SE): O=$4,224 (11,219); R=$4,339 (11,217); T=$5,403 (14,786); p <0.0001). Patients initiated on olanzapine had been more treatment resistant, as indicated by differences in use of more than two antipsychotics previously (O=42.97%; R=37.10%; T=40.13%; p <0.0001), prior clozapine use (O=1.11%; R=0.71%; T=1.07%; p <0.07), prior use of depot antipsychotics (O=3.96%; R=2.55%; T=2.16%; p <0.0001) and dose of prior antipsychotic (mean chlorpromazine equivalents (SE): O=228.1 (662); R=161.2 (494); T=130.5 (474); p <0.0001). Patients with episodes treated with typical antipsychotics had a higher proportion of comorbidities including muskoskeletal system disorders, anxiety disorders, and circulatory system diseases. Olanzapine had relatively few older and younger initiators (age < 20 or age > 65), compared to risperidone (O=13.67%; R=24.69%; T=18.90%; p<0.0001). CONCLUSIONS: Significant cost, demographic and diagnostic differences at baseline existed among patients initiating risperidone, olanzapine, and typical antipsychotics. Specifically, olanzapine was used for more treatment resistant patients with schizophrenia. Non-randomized studies comparing treatment and cost outcomes for these medications need to account for the differences.