OBJECTIVES To investigate the relationship between compliance and first-year post-treatment cost for patients with schizophrenia. METHODS Commercial health plan data from July 1, 2003 to June 30, 2006 were used to identify patients with schizophrenia who initiated treatment using a typical or atypical antipsychotic, a mood stabilizer or an antidepressant. Episodes were divided into three categories: restarting treatment after a break in drug therapy >15 days with the drug used in the previous episode; switching therapy with or without a break in treatment; and augmentation therapy. First observed episodes were excluded due to uncertainty concerning the patient's prior treatment history. A total of 21,872 episodes were included in ordinary least squares (OLS) regression models of post-treatment cost as a function of alternative definitions of compliance adjusting for age, gender, geographic region, drug use history, prior medical care use, schizophrenia diagnosis and co-morbid medical conditions. RESULTS Adding a second medication within one year is the key factor in the relationship between compliance and one-year costs. One year of uninterrupted therapy on the initial drug without adding a second medication increased drug costs by +$3314, while reducing medical costs by $3919 (p<0.0001 for both estimates) and total cost by $605 (p>0.05). One year of therapy on the initial drug augmented with a second medication increased drug costs (+$5062) and medical costs (+$3477) compared to non-compliance (p<0.0001 for both estimates). One year of uninterrupted therapy on a second drug after terminating the initial therapy was associated with the highest costs: +$3,955 for drugs and +$8,902 for medical services (p<0.0001 for all estimates). Analyses conducted separately by type of episode confirmed these patterns of cost estimates. CONCLUSIONS Compliance is associated with significant reductions in medical costs only if the patient is maintained on their initial drug regimen. Changes in therapy are associated with higher medical costs.