OBJECTIVES:   To assess the all-cause medication discontinuation rate in matched cohorts of patients with schizophrenia at risk of nonadherence who were initiated on depot or oral antipsychotics and followed over 12 months. METHODS:  At study entry, outpatients with schizophrenia from Australia, Mexico, Romania, and Taiwan were switched, due to clinician-perceived medication non-adherence risk, from their current oral antipsychotic to either a depot or different oral antipsychotic in this 12-month, prospective observational study.   Patients were compared on all-cause medication discontinuation rates, defined as a switch from the initiated medication or its augmentation with another antipsychotic. Patients initiated on depot were matched with those initiated on any oral antipsychotic, using full optimal and nearest neighbour (greedy) matching algorithms. The Rank-based Mahalanobis metric was chosen as the distance based on propensity score plus other relevant covariates, with country and antipsychotic class used for exact matching. RESULTS:  Based on the full optimal 1:1 matching, only 40 of the 43 original depot initiators were matched to a corresponding oral initiator. After matching, there were no statistically significant differences between the depot and oral cohorts on any of the study entry covariates examined. During the 12-month study, 20% of depot patients discontinued their initial medication, compared with 40% of oral patients (survival curve comparison, p=0.025; Hazard Ratio=0.33 [0.12, 0.92], p=0.033, NNT=6 [3,-34]). No statistically significant differences were found between the two groups on other assessed outcomes, including hospitalization, length of stay, and quality of life and disease severity measures. CONCLUSIONS:  Systematic matching of patients initiated on depot with those initiated on oral antipsychotics showed that the oral-initiated patients were statistically significantly more likely to discontinue their medication. Findings highlight the importance of systematic matching of patient cohorts when comparing treatment outcomes in observational studies.