164 Chapter 5 We also conducted four additional analyses to support the robustness of our findings. We started by examining whether psychologists’ supply was associated with changes in prescription but not with the depression diagnosis. Because the capacity and ability of GPs diagnosing depression may be affected by the presence of more psychologists, this could distort the findings of our outcome, which measures an antidepressant prescription share standardised to depression prevalence. For examining this possibility, we replaced the dependent variable in Model 2 with the number of newly diagnosed cases of depression at the GP level. Second, we examined whether the variation in the number of psychologists could be partially a response to GP patterns in antidepressants prescription. Based on the concept of reverse causality it could be that no (more) psychologists would be hired because GPs would not refer patients to psychotherapy anyway. To exclude this hypothesis, we estimated whether changes in the number of psychologists were associated with former antidepressant prescription shares ( Yit−1 and Yit−2 ). Third, we estimated our baseline model excluding the variables of physician list size and the utilization share of the GP visits, to avoid potential endogeneity if diagnosis capacity would be improved due more interaction and closer relationship between GP and patients. Furthermore, excluding the utilization share helps us to addressing the concern that amount of GP visits may be affected by the depression treatment decision and thus bias the association with psychotherapy. Last, we investigated the robustness of our results using a balanced panel of GPs in practice during the 4 years of the study period. RESULTS Descriptive statistics From 2015 to 2018 our population follows the prescription decisions of 5,359 unique GP (n=17,210 observations) in 888 unique GP practices grouped in the 55 local groups that exist in Portugal. The smallest local group provided care to 30,493 patients and the largest to 386,027 patients in 2018. Table 1 summarizes the characteristics of the study population. The share of antidepressants prescribed to adults identified as having depression was on average 27%. The standard deviation (SD) figures suggest that there is limited variation over time (SD 3.97) but a considerable variation between GP (SD 14.6, ranging from 0% to 96% - Table 1). Most GPs were female (73%) and worked in the same GP practice for more than 5 years (60%), having an average of 1,710 patients on their patient list, from which 69.9% effectively used services. The most prevalent conditions were hypertension (22.7 % of the total population in the GP list), obesity (9.2%) and diabetes (8.2%). About 47.6% of the patients on the list were exempted from co-payments. GP practices would more frequently follow the model of PHCU (38.7%) and FHU-B (34.7%) and had operated on average for 7.3 years. The proportion of patients in the practice without GP varied substantially from 0 to 77.8%, and the ratio of nurses to GPs was on average 1.12. The incidence rate of depression at
RkJQdWJsaXNoZXIy MjY0ODMw