Objectives: This paper examines three processes: SSRI antidepressant choice, adherence to treatment guidelines, and long-term health care expenditures associated with antidepressant treatment for patients with a diagnosis of depression.
Methods: Patient records were abstracted from a medical claims database covering employer-provided health care plans. Treatment episodes required a 6-month antidepressant-free prior period; initial treatment with sertraline, paroxetine or fluoxetine; and data on direct medical costs over the 24 months following the initial prescription. The multivariate model of drug selection, patient adherence to antidepressant use guidelines, and cost was subjected to specification testing to rule out the possibility that nonrandom initial antidepressant selection might lead to sample selection bias. Further tests indicated that the results were free of bias due to a possible correlation between antidepressant selection and use of the medication, or because of the endogeneity of use patterns in the process driving cost. However, there was evidence of unobserved variables correlated with both achieving guideline adherent use and expenditures, which might have led to sample selection bias.
Results: Subjects who met the study criteria included 796 initiating therapy with sertraline, 352 with paroxetine, and 882 with fluoxetine. Fluoxetine patients were significantly more likely than sertraline or paroxetine patients to achieve a use pattern that was consistent with guidelines for treating depressive disorder (p < .05). There were no statistically significant differences between the three treatment cohorts in total direct health care expenditures over the 2-year period (p < .05), and depression-related expenditures, other mental health expenditures, and non-mental health care expenditures did not show significant differences across the treatments (p < .05). Natural logged values of antidepressant drug expenditures were predicted to be highest for fluoxetine, followed by sertraline, then paroxetine (p < .01). Predicted log values of mental health expenditures were lower for sertraline relative to fluoxetine.
Conclusions: Fluoxetine patients had the highest likelihood of using antidepressant medication according to treatment guidelines that were developed to assure quality care. This benefit was achieved without incurring greater total health care expenditures.