OBJECTIVES: This retrospective study aimed to compare health care utilization associated with duloxetine versus other antidepressants for patients with major depressive disorder (MDD) in the Veterans Affairs (VA) health system. Duloxetine was not on the VA national drug formulary. METHODS:  The electronic medical records from October 2004 to October 2008 were extracted from the VA Veterans Integrated Service Network 16 data warehouse. All patients were treated with either duloxetine monotherapy or other antidepressants (non-duloxetine) over the study period. The first dispense date of the index agent was defined as the index date. All patients must have at least 1 prior MDD diagnosis (ICD-9-CM: 296.2 or 296.3), but no prior diabetes (ICD-9-CM: 250.xx) or bipolar disorder (ICD-9-CM: 296.4x-296.8x) diagnosis. The non-duloxetine patients were matched to the duloxetine patients via propensity scoring (1:1 ratio), controlling for differences in demographics, comorbidities, prior opioid use, and pain levels in the 12 months pre-index period. Paired t-tests were used to compare health care utilization over the 12 months post-index period including outpatient visits, emergency department (ED) visits, hospital admissions and length of hospital stay. Chi-square tests were used to compare percentages of patients with hospitalization or ED visit. RESULTS: The study sample included 878 patients (duloxetine: n=439; non-duloxetine: n=439) with comparable baseline characteristics. Compared with duloxetine patients, non-duloxetine patients on average had 12.0 more outpatient visits (95% Confidence Interval [CI]: 6.4-17.5, p<0.0001), 0.16 more hospital admissions (CI: 0.07-0.26, p=0.001), and 0.79 more hospital days (CI: 0.17-1.41, p=0.013). Additionally, a higher percentage of non-duloxetine patients was hospitalized (17.8% vs. 10.9%, p=0.004). No group difference was found in ED visits. CONCLUSIONS: Controlling for cross-cohort differences, VA patients who were treated with duloxetine were found to be associated with lower health care utilization than those treated with other antidepressants.