Of the 139 eligible physicians, 78.4% (109) participated. The intervention group had 56 physicians and the usual care group had 53 physicians. Because we excluded subjects enrolled in the earlier depression trial (top 15% frequent users of medical services with depression), study PCPs provided care for this sample of 124,893 patients. As some patients were frequent users in one year but did not continue their high use of medical services in the following year, our sample represented 91% of patients cared for by study physicians. Subjects were grouped according to the intervention or usual care status of their physicians with 60,689 in the intervention group and 64,204 in the usual care group. shows the demographic characteristics of the patients and PCPs at GHC and Dean. About 90% of the GHC PCPs were family physicians while almost 70% of Dean PCPs were internists. There were more female and part-time physicians at GHC than Dean.
Demographics of Primary Care Physicians and Patients
In the pre-study period, the intervention physicians had a slightly higher rate of new depression diagnoses (1.84 vs 1.63 visits per 100 visits). The first row of shows that after training, the rate of diagnosing a new depressive episode increased slightly in both groups. In a logistic regression model including main effect of study group (intervention vs usual care), main effect of time (pre vs post), and the interaction of the two, the odds ratio (OR) associated with the interaction term was 1.01 (95% confidence interval [CI], 0.83 to 1.2). In other words, the pre-to-post change was essentially identical in the groups with the 95% CI ranging from 17% lower to 20% higher in the intervention group.
Pre and Post Comparison of Depression Diagnosis, Prescription and Duration of Antidepressant Treatment
As we examined new prescriptions of antidepressant medicines during the pre-study period, intervention physicians prescribed antidepressant treatment a little more frequently than did the usual care physicians (1.82 vs 1.46 prescriptions per 100 visits). The second row of shows that after training, prescription rate decreased somewhat (not statistically significant) for intervention physicians and increased slightly in the usual care physicians. In a logistic regression model including group, time, and the interaction between the 2, the pre-to-post change did not differ significantly between the 2 groups (OR for interaction term, 0.83; 95% CI, 0.69 to 1.03).
The third row of shows, in the pre-study period, about 64% of visits to the intervention physicians and 57% for the usual care group received a new prescription of antidepressant for a newly diagnosed episode of depression. Both groups showed a slight increase in the post-study period. Logistic regression again found that the pre-to-post change did not differ significantly between intervention and usual care groups (OR for interaction term, 0.95; 95% CI, 0.70 to 1.30).
We also used patient level data to assess adequacy of pharmacotherapy (continuous antidepressant treatment for at least 12 weeks within a 114-day period [HEDIS criteria]). The last row of showed that 67.14% (n = 190) of intervention compared to 63.53% (n = 169) of usual care patients had adequate pharmacotherapy in the pre-study period. Once again, the pre- to post- change did not differ between intervention and usual care groups (OR for interaction term, 0.82; 95% CI, 0.43 to 1.55).
For every measure examined (rate of new diagnosis of depression, prescription of new antidepressant medication, or adequacy of pharmacotherapy), we also used group by time logistic regression analyses to test whether effect of the intervention differed between the GHC and Dean sites. For all measures, there was no significant group × time × site interaction, indicating that training effects did not differ across sites.