The purpose of this study was to identify differences in hospitalization incidence and associated economic impact by depressed patients initially treated with a selective serotonin reuptake inhibitor (SSRI) versus those treated with tricyclic antidepressants (TCA). METHODS: A three step regression analysis was conducted utilizing data from Marketscanâ. First, discrete outcome regression methods were used to test for a systematic relation between patient characteristics and initial antidepressant that may be a common factor with hospital use forming the basis of an adjustment for a non-random drug assignment when studying hospital use. The second step was a discrete outcome regression for whether a depressed patient is hospitalized. The final step was a semi-logarithmic regression of hospital expenditures for admitted patients. RESULTS: The only control independent variable that mattered to depression-related hospitalizations were prior admissions. The number of pre-study comorbid conditions was the only significant predictor of drug use. People with more pre-study comorbidities were less likely to have an initial prescription for fluoxetine (t = 3.33, p <0.05). Applying statistical correction for this variable the estimated relative likelihood of hospitalization with fluoxetine versus TCA was 0.47 (95% CI = 0.23 - 0.93) ( t = 2.22, p < 0.05). The estimated hospitalization rates for sertraline and TCA were the same (b = 0.81, t = 0.54, p = 0.59). There was no differential effect of specific anti-depressant on resource use for patients once hospitalized. CONCLUSIONS: Patients initially prescribed sertraline have the same likelihood of hospitalization for a mental health problem as patients prescribed a TCA. Patients initially prescribed fluoxetine are half as likely to be hospitalized as patients initially prescribed TCAs.