OBJECTIVES: Many pharmaco-economic studies have sought to identify the cost-effectiveness of different SSRI antidepressants. In order to replicate randomized controlled trials, many of these studies have reduced the SSRI-user population to small, often unrepresentative numbers. This study seeks to inform health plan decision-makers by studying a one-year cross-sectional population of SSRI users. This population-based perspective is useful in understanding natural differences in SSRI use. METHODS: We identified 27,134 SSRI users in 1997 in a large, mixed model HMO. For each user we identified total SSRI days, SSRI cost, total pharmacy cost, total health plan cost using charges, paid amounts, and a standardized cost algorithm, and comorbidities using the Charlson comorbidity index. Statistical methods appropriate to the highly skewed distribution of health care expense were used to calculate adjusted mean expense for users of each product. RESULTS: SSRI users included children and adults. Unadjusted total paid charges were lowest for users of fluoxetine. However, both the paroxetine and sertraline populations were significantly sicker as measured by the Charlson comorbidity index (difference = 0.11, p<0.05). Adjusted for age, gender, and comorbidities, the mean paid charges were 15% higher for paroxetine (p<.001) compared to fluoxetine and not significantly different for sertraline compared to fluoxetine. When total pharmacy charges are added to the non-pharmacy charges, the paroxetine is 20% and sertraline is 27% lower than fluoxetine. Users of paroxetine are about 50% more likely to switch among the SSRIs. Switchers have 34% higher total charges. CONCLUSIONS: The use of SSRIs represents an important and growing portion of medical expense. This paper illustrates that total medical care cost should be used in making population-level treatment choices.