Olanzapine and risperidone are two commonly prescribed atypical antipsychotics for schizophrenia. Prior studies have shown inconsistent results in terms of advantage in cost saving in prescribing these agents. Our preliminary analysis showed that a small percentage of intensive healthcare utilizers had substantial impact on healthcare costs. This study analysed the cost effects of olanzapine and risperidone among those who had intensive utilisation of medical care prior to drug initiation, and the relationship between the choice of the two drugs and patients' co-morbid condition.We retrospectively investigated patients first treated with either risperidone or olanzapine between 1 April 1999 and 31 March 2000. According to patients' medication history during the 6 months prior to initial prescription of a study medication we categorised patients into three groups: (i) not receiving olanzapine or risperidone; (ii) not receiving any atypical antipsychotic agent; or (iii) not receiving any antipsychotic agent. We then compared the difference in cost saving by type of care in the 10% most expensive patients through bivariate and multivariate analyses. Based on the records of 18 499 patients with schizophrenia prescribed either olanzapine or risperidone between 1 April 1999 and 31 March 2000 we defined intensive users of healthcare as those who incurred an annual total cost in the top tenth percentile. We measured co-morbidity by number of diseases, and healthcare costs ($US, 1998-2001 values) in the year prior and the year after treatment initiation in six categories of care (inpatient medical/surgical, inpatient psychiatric care, other inpatient, outpatient psychiatric care, other outpatient and outpatient pharmacy).The top 10% most expensive patients accounted for about half of the total cost of the entire cohort and had nearly a 40% cost reduction for the year after treatment initiation versus the prior year, while the entire cohort increased cost between 2% and 12%. Compared with those prescribed olanzapine, patients prescribed risperidone had more medical co-morbidities, higher inpatient medical/surgical costs and lower psychiatric costs. Patients taking olanzapine had greater cost reduction in inpatient psychiatric care, whereas those taking risperidone had greater reduction in inpatient medical/surgical care.Among the top 10% most expensive patients, olanzapine and risperidone treatments were associated with comparable cost reductions in inpatient care. The choice of agent was associated with patients' co-morbid condition and was correlated with cost reduction in inpatient medical/surgical or psychiatric care.