156 Chapter 5 Several mechanisms can be outlined both in favour and against this theoretical hypothesis that an increased supply of psychological therapy leads to the decreased prescription of antidepressants. As in other countries, Portuguese GPs work with large patient lists and limited appointment time. This means that shifting care to other practitioners may have considerable efficiency gains, especially when it does not have direct budget implications for GPs and when there is no competition for patients on the quality of care, as in the Portuguese case. With psychotherapy being the first-line recommendation for mild depression and the accumulation of evidence supporting its efficacy versus careas-usual but also pharmacotherapy [28–31], it would be reasonable for GPs to substitute antidepressants prescription by a referral to mental health workers working in primary care. This substitution would also free GP´s time for additional activities, as the (initial) prescription of antidepressants requires monitoring and a more intensive follow-up than a referral. On the other hand, GP surveys and qualitative studies suggest that treatment decisions are significantly influenced by GPs’ beliefs, attitudes, and personal experiences [22, 23, 32]. Practical aspects such as medication fitting more easily into the routine of clinical practice might also influence the choice [33]. Literature also shows that while GP perception of psychotherapy effectiveness is high and has increased over the years, GPs still prefer to use it in combination with antidepressants, rather than psychotherapy alone [14, 23, 34]. Faced with the lack of guidelines consensus on first-line treatment for moderate depression, GPs may prefer to use the increasingly available psychotherapy to combine it with antidepressants for higher severity patients, instead of using it as a substitute for medication in milder cases. We explored the nested and longitudinal nature of our data to study the variation in the number of psychologists - a proxy of psychological therapy supply – by decomposing it in within GP and the between GP variation. The within variation shows how changes in the number of psychologists in each local group led to changes in prescription by GPs exposed to these changes overtime, between 2015 and 2018. As in a fixed effects frameworks, the within estimate obtained from our model is not affected by omitted variable bias due to unobserved time-invariant confounders, thereby providing stronger evidence about the causal relationship of interest [35]. Such biases are absorbed by the between term, which captures the variation in the number of psychologists between GPs working at different local groups [36, 37]. Importantly, the between estimate is influenced by unmeasured factors related to both the number of psychologists and GPs’ prescription behaviour. The potential bias of the between GP coefficient makes its comparison with the within GP coefficient informative. A large difference between these terms suggests an important role of unmeasured characteristics and processes in the relationship observed between the supply of psychological therapy and antidepressant prescription. We conducted heterogeneity analyses to further explore the mechanisms at play in the relationship of interest. Firstly, we explored heterogeneity in the within term by the average level of psychologists available to each GP during the study period, which
RkJQdWJsaXNoZXIy MjY0ODMw