161 Antidepressant therapy prescription, do psychologists help? Evidence from Portugal with a GP, the ratio of nurses to GP in the practice and the practice incidence rate of depression (i.e. the share of yearly diagnosed cases among the patients enrolled in the practice). Additionally, we controlled for the local context in which the practice was located by considering whether it is part of an integrated care scheme (Unidades Locais de Saúde - ULS) but also by accounting for municipality-level characteristics such as the population density, average earnings of employees, unemployment rate and the degree of urbanization (cities, towns or suburbs and rural areas5). Empirical approach To examine how the GP antidepressant prescription changes with different levels of supply of psychological therapy we start from a (pooled) baseline model: Yijl =α0 +α1Pl +X′ ijlα2 +Z′ jlα3 +εijl (1) where Yijl represents the antidepressant prescription rate of GP i, working in practice j , that belongs to local group l. Pl is the measure of psychotherapy supply facing all GPs at local group l; Xijl is a vector of GP characteristics and patients list characteristics, that accounts for observed differences in average medical need, in the absence of individual patient data; Zjl is a vector of GP practice characteristics, and εijl is the error term. In this baseline model the coefficient of interest, α1, measures the association between antidepressant prescription and psychological therapy supply. The estimation of this association ( α1 ) might be biased by unobserved provider characteristics such as GP characteristics that systematically influence their decision-making in treating depression (e.g., knowledge, beliefs, etc.), and unmeasured factors driving psychologists hiring by local groups. While mental health need should explain a great part of psychologists’ supply, other factors such as clinical leadership and the maturity of mental health services integration in each local group might also play a strong role. These could, for example influence GP and psychologist sorting into the local groups. One other potential threat to the robustness of α1 might arise from reverse causality, namely when no psychologists would be hired because GPs prefer to prescribe antidepressants than refer patients to psychotherapy. Therefore, to better investigate what is driving the association between antidepressant prescription and psychological therapy supply captured by α1, we decomposed this relationship by examining two types of variation. On the one hand, the variation in the number of psychologists faced by each GP in a local group overtime (within term). On the other hand, the variation in the average number of psychologists available to GPs across 5 Municipality urbanization classification according to Eurostat 2011. Available at: https://ec.europa.eu/eurostat/ statistics-explained/index.php?title=Degree_of_urbanisation_classification_-_2011_revision (visited in Jan 2023) 5
RkJQdWJsaXNoZXIy MjY0ODMw