Developed and developing countries alike have large regional disparities in the provision of medical doctors, with remote and under-served areas often facing a more limited supply of personnel. Retaining doctors in these areas remains a major policy challenge (WHO, 2010). Low retention rates can negatively affect healthcare quality and health outcomes due to understaffed services and discontinuity of care (WHO, 2018). Moreover, as the global demand for healthcare professionals continues to grow, difficulties in doctor attraction and retention are projected to increase, especially in remote areas (Liu et al., 2017, Esu et al., 2021). Yet, causal evidence regarding the extent to which and how doctor supply affects population health remains limited. Most existing studies focus on the population health impacts of increases in the supply of personnel. Very little is known about how doctor exit affects population health outcomes, in particular in resource-scarce contexts, how health systems adapt, and how health production inputs and outputs respond to doctor exit at the market level, in general.

This paper studies the effects of a unique large-scale doctor withdrawal event on population health outcomes, and on health production inputs and outputs by exploiting an unexpected shock to local health systems in Brazil. To overcome a chronic lack of doctors in under-served areas, in 2013 Brazil introduced the More Doctors Program (Programa Mais Médicos – PMM). Over 18,000 doctors served in primary care services in eligible municipalities, roughly half of them being from Cuba. Five years after the introduction of PMM, following the election of Brazil’s former right-wing president Jair Bolsonaro in October 2018, the Cuban government unilaterally withdrew all of the Cuban doctors from the program. The sudden exit led to a major shock to the supply of healthcare where Cuban doctors had been working. Some municipalities were suddenly left without PMM doctors. Other municipalities, where no Cuban doctor had been working, remained unaffected.

We exploit the sharp and unexpected exogenous timing of the Cuban exit from PMM in municipalities that relied more versus less on Cuban PMM doctors at the time of the withdrawal to estimate effects of doctor exit on population health outcomes, health production inputs (such as availability of personnel and health facilities) and outputs (such as number of consultations and medical procedures by complexity level). More specifically, we use an event-study approach to compare outcome changes in affected and unaffected PMM municipalities before and after the Cuban exit, conditional upon time and municipality fixed effects, therefore controlling for baseline heterogeneity in levels and common time trends. We combine different fine-grained administrative microdata sets on healthcare utilization, hospitalization and mortality, as well as on the supply of healthcare facilities, personnel, prescription drug sales and private health insurance into a municipality-by-month panel of data between October 2017 and December 2019. The overall evidence indicates that pre-trends in outcome variables remain irrelevant, while alternative specifications, alternative treatment definitions and robustness checks lend support to our identification strategy.

We find an immediate drop in the number of primary care doctors in affected municipalities from November 2018 on. That reduction is particularly pronounced for municipalities where all PMM doctors were Cuban. From January 2019 on, affected municipalities started to fill PMM vacancies, but only by mid-2019 doctor supply stabilized, almost reaching pre-treatment levels. Next, we document meaningful reductions in primary care utilization. Effects are concentrated in physician duties, and most persist even after the arrival of new doctors. Persistent reduction in service utilization is primarily driven by conditions requiring prevention and ongoing care, such as mental health and chronic diseases, and related risk factors. For example, hypertension consultations decline on average by 20% in the most affected municipalities immediately after the Cuban exit, and remain 14% lower than base levels. Importantly, while there was a strong and persistent decrease in the care of mental health and chronic diseases, service utilization for conditions requiring immediate care, such as infections and maternal-related services, remained relatively stable. Finally, the reduction in healthcare utilization did not translate into any systematic changes in health outcomes even throughout the end of the following year after the Cuban exit. We do not observe any effects on hospital admissions or mortality, neither on aggregate levels nor from specific causes.

Adaptation of local health systems and demand diversion seem to have helped mitigate the negative shock to doctor supply without major adverse repercussions for population health outcomes. We explore different intermediate outcomes that could explain the overall findings. We first focus on adaptation within primary care services. We do not find any changes in the supply of other health professionals nor in task substitution across different professionals. Moreover, we do not observe an improvement in the experience profile of the new doctors who replaced the Cuban professionals. Yet, patients in need of prescription medication for chronic diseases were able to secure their access despite the reduction in access to other related healthcare services, such as consultations. Together with the rapid recovery of consultations for conditions requiring immediate care, the overall evidence indicates a prioritization of primary care services towards urgent and curative needs as well as other protective services.

Second, we investigate system adaptation and demand diversion beyond primary care services. While there were no significant changes in the use of hospitals and specialists, we observe a suggestive increase in ER utilization for basic ambulatory care as well as in the number of ERs offering such services, particularly in municipalities where all PMM doctors were Cuban. Patients therefore seem to partly substitute treatment at primary care centers with treatment at ER facilities, while municipalities start providing more basic services in these facilities. Importantly, this diversion persists over time. Finally, we do not observe any changes in private insurance coverage in affected municipalities nor detect any spillover effects across municipalities. Adaption therefore comes at the cost of a reduction in preventive care in primary care services and an increase in utilization of ER facilities. We conjecture that this could entail fragmentation and decreased efficiency in healthcare provision, with potential negative long-term consequences for tackling the increasing burden of chronic diseases.

This paper contributes novel evidence to the emergent literature on the impacts of doctor supply, more generally, and of doctor exit, more specifically, on population health. Existing studies document positive associations between the supply of health professionals and health outcomes, but often struggle to establish causality (Anand and Bärnighausen, 2004, Basu et al., 2019). Causally identified studies are scant and document a nuanced picture, especially in resource-scarce contexts. Experimental evidence has recently come from Okeke (2023), who shows that the marginal addition of doctors in primary care services in Nigeria significantly improves health outcomes, while lower-qualified health workers do not have any significant effects. Assessments of the introduction of PMM in Brazil provide quasi-experimental evidence and have gained academic attention both in economics and public health. Similar to other contexts, the country has long faced an unequal distribution of healthcare professionals, both in terms of quantity and quality (Costa et al., 2019). Despite increasingly dense, the literature on the effects of the introduction of PMM has yielded mixed results. For instance, Carrillo and Feres (2019) document that PMM led to the replacement of nurses by higher-qualified primary care doctors, without any effects on infant health. Fontes et al. (2018) and Mattos and Mazetto (2019) find that the introduction of PMM decreased hospitalizations, and Hone et al. (2020) document a small reduction in mortality amenable to primary care. Finally, our paper is also related to studies that quantify the value of medical personnel in producing health (Agha et al., 2019, Fadlon and Van Parys, 2020, Currie and Zhang, 2025, Ginja et al., 2025). For example, Currie and Zhang (2025) show meaningful differences in primary care provider performance within the U.S. Veterans Health Administration, while Ginja et al. (2025) document substantial heterogeneity in GP fixed effects in Norway.

We contribute novel evidence to the literature by exploiting a unique setting in which an exogenous shock to local health systems led to a large-scale exit of doctors at the market level. Importantly, although the diplomatic conflict that triggered the Cuban exit is unique, the local manifestation of the shock is similar to retention failures or abrupt staffing gaps that are common in resource-constrained health systems. Conditions such as high turnover among contract physicians, difficulty attracting doctors to remote areas, and sudden staffing gaps due to administrative or fiscal crises are recurring features of primary care provision in many low- and middle-income countries.

Unlike most studies that focus on increases in doctor supply, we assess the consequences of a sharp reduction in doctor availability. This is relevant because hirings and increases in the availability of doctors are generally followed by challenges in workforce retention, and positive and negative variations in doctor supply do not necessarily lead to symmetric effects on outcomes. Yet, while doctor retention and exit are a widespread policy challenge in all countries, evidence on the effects of doctor exit at the market level remains scarce. Existing studies tend to focus on specific disruptions in patient–provider relationships within developed countries,1 and often overlook effects beyond the directly affected patients and the broader market-level responses. In this paper, we not only examine the effects of a negative shock to the supply of doctors on population health outcomes at the market-level, but also investigate the response in terms of health production inputs and outputs. Our results are particularly informative for resource-scarce contexts, such as in low- and middle-income countries, where patient access to services and provider choice is typically more constrained, and therefore a negative shock to the supply of doctors may potentially lead to the most harmful health effects. To the best of our knowledge, this is also where causal empirical evidence on doctor exit is still absent.

This paper is organized as follows. In Section 2 we describe the institutional background. In Section 3 we detail the data, while in Section 4 we describe our empirical model and discuss identification. Results are presented in Section 5. In Section 6 we discuss potential pathways and explanations behind the results. Section 7 concludes.