169 Antidepressant therapy prescription, do psychologists help? Evidence from Portugal the other hand, the between term shows a significant association between working in local groups with one-unit higher average number of psychologists/100,000 patients and having a lower antidepressant prescription share, by 0.955 pp (CI 95% -1.641, -0.269). When compared to the average prescription share of 26.88 %, this result indicates that GPs in local groups with a higher average number of psychologists by 1/100,000 patients have a prescription share that is lower by 3.6%. Table 2. Within-between random effect models studying the association of psychotherapy supply and antidepressant prescription share. No controls (1) Year dummies (2) GP (3) GP practice (4) Main specification (5) Psychologists/100,000 patients Within-term -0.943*** [-1.334,-0.551] -0.268 [-0.737,0.202] -0.263 [-0.688,0.161] -0.270 [-0.688,0.148] -0.305 [-0.706, 0.0954] Between-term -1.819*** [-2.810,-0.828] -1.828*** [-2.810,-0.845] -1.467*** [-2.132,-0.803] -1.323*** [-2.014,-0.632] -0.955** [-1.641, -0.269] Year dummies No Yes Yes Yes Yes GP covariates No No Yes Yes Yes GP practice covariates No No No Yes Yes Local context * covariates No No No No Yes Constant 30.26*** [27.95,32.57] 30.14*** [26.97,33.31] 25.23 [-35.24,85.71] 47.97 [-16.03,112.0] 22.45 [-48.56,93.46] N 17,210 17,210 17,210 17,210 17,210 R2 within 0.0128 0.0525 0.0697 0.0721 0.0756 Notes: The dependent variable is the share of adults with depression in the GP list that have been prescribed antidepressants. Covariates are included in the model demeaned and centered (e.g. for each covariate two terms are included, one with the mean over the study period and other with the difference from the mean in each year). *Some of the covariates are reported at the municipality level, which corresponds roughly to the local group geographical boundaries. 95% confidence intervals in brackets. Standard errors clustered at the local group level (55 local groups in all models). * p < 0.05, ** p < 0.01, *** p < 0.001 Contrary to the within term, the results for the between coefficient are impacted by the characteristics of GPs, GP practices and local groups, as supported by the reduction of the coefficients with the subsequent inclusion of controls in models (2) and (5). Despite this reduction, fully adjusted model depicts a between coefficient that is statistically significant and larger than the within term, suggesting that even after extensively accounting for patients’ list characteristics and GP/practice observables unmeasured characteristics and processes are dictating that GPs prescribe less in local groups with higher average number of psychologists. Heterogeneity analyses Table 3 reports the heterogeneity analysis conducted by estimating our main model stratified for quartiles of the average number of psychologists in the local group. Results 5
RkJQdWJsaXNoZXIy MjY0ODMw