OBJECTIVE: To compare prescription and medical resource utilization by patients in Maryland Medicaid plans who are initiated on thiazolidinediones (TZDs) versus those initiated on metformin. METHODS: This study includes a prospective non-concurrent analysis of prescription and medical claims for Maryland Medicaid patients who were initiated on metformin or TZDs between June 15, 2000 and June 15, 2002. We ran univariate, bivariate, and multivariate models, examining the associations between the likelihood of being initiated on a TZD and age, gender, race, county of residence, coverage group, and days supply, as well as variables to assess resource utilization: drug count, total pharmacy claims, unique diagnoses count, and total medical claims. Logistic regression models were used to assess the combined effect of all utilization variables on the likelihood of incident use of TZDs or metformin, adjusting for demographic variables. We also tracked the use of metformin and TZDs in this population. RESULTS: The sample of 4440 patients was mostly female (71.19%), older (52.58% over 49 years old) and African American (58.84%). Patients with higher numbers of unique diagnoses (OR = 1.6, p = 0.0071), prescription claims (OR = 1.5, p = 0.0813), and unique drugs (OR = 1.7, p = 0.028) were more likely to have received TZDs first line. Those patients with higher numbers of medical claims (OR = 0.5, p = 0.0029) were less likely to have been started on TZDs. Among patients started on metformin, 84% did not have subsequent use of TZDs. Among those started on TZDs, 82% did not use metformin. CONCLUSIONS: Results show that patients initiated on TZDs are more likely to have subsequently higher utilization of prescription resources, even after adjusting for demographic variables. These patients were also likely to use less medical resources than metformin initial users.