Effect of More Doctors (Mais Médicos) Program on geographic distribution of primary care physicians

Resumo Este artigo avalia o efeito do Programa Mais Médicos (PMM) na distribuição de médicos na atenção básica no Brasil. As técnicas de análise de dados espaciais, curva de Lorenz e coeficiente de Gini foram empregadas para avaliar a distribuição geográfica dos médicos antes e após a implementação do PMM (2012 e 2016). Para tanto, foram utilizados dados secundários de 5.564 municípios. Os resultados indicaram que a distribuição de médicos se tornou mais igualitária após a implementação do PMM. Entre 2012 e 2016, o coeficiente geral de Gini diminuiu 11%, passando de 0,255 para 0,227. No âmbito estadual, uma tendência estatisticamente significativa em direção a uma distribuição mais igualitária de médicos foi encontrada em 21 estados brasileiros. No entanto, ainda persiste uma diferença substancial na distribuição de médicos na atenção básica, especialmente entre os estados, com o coeficiente de Gini variando de 0,093 a 0,341 em 2016. Palavras-chave Médicos, Atenção Básica à Saúde, Disparidades em Assistência à Saúde, Mão de Obra em Saúde Abstract This study assesses the effect of More Doctors Program (Programa Mais Médicos PMM) on the equality in the distribution of primary care physicians (PCPs) in Brazil. Spatial data analysis, Lorenz curve and Gini coefficient were used to evaluate the geographic distribution of PCPs before and after the implementation of PMM (2012 and 2016). Data from 5,564 municipalities were used in the analyses. The results indicate that the distribution of PCPs has become more equal after PMM implementation. Between 2012 and 2016, overall Gini coefficient decreased by 11% from 0.255 to 0.227. At the state level, a statistically significant trend towards a more equal distribution of PCPs was found in 21 out of 26 Brazilian states. However, there still remains a substantial difference in the level of equality in PCP distribution, especially across states, with Gini coefficient ranging from 0.093 to 0.341 in 2016.


introduction
The geographic distribution of physicians has been a persistent issue worldwide 1,2 . Despite many countries facing the uneven distribution of these professionals, this is a matter of great concern for low-and middle-income countries, since the population already have limited access to health care 3 . In general, physicians tend to concentrate mainly on wealthier and larger urban areas in the country. The lack of physicians in rural, poor and peripheral regions implies in significant disparities in health outcomes between the populations 2,4 .
Over the past decades, different strategies have been adopted in order to attract and retain these professionals in underserved and high need areas 3,5,6 . In Brazil, despite great progress towards a more equitable access and utilization of health care [7][8][9] , there still remains a marked unequal distribution of physicians across national territory. The unbalanced distribution of physicians is noticed even when compared the numbers between states. For example, Rio de Janeiro has 3.75 physicians per 1,000 population (61,346 physicians and 16,369,179 population in 2014), while Maranhão, the poorest state in Brazil, has 0.79 physician per 1,000 population (5,396 physicians and 6,794,301 population in 2014) 10 . Commonly, areas with lack of physicians are also areas with the largest population in need, intensifying the unequal access to health care 11 .
There are two major reasons why the unbalanced distribution of physicians across areas is a concern to policy makers and researchers. First, in countries with system founded on the principles of equality and social justice, there is an issue when part of population has limited access to health care. Second, an extensive literature indicates the link between physician availability, access to medical care and health status 12,13 . Clearly, these impacts are considerable for primary health care. In general, primary care physicians (PCPs) are more equally distributed than specialists, but maldistribution of PCPs also occurs in many countries 14 .
In recent years, Brazil has implemented new national policies to face the geographical maldistribution of health workforce, especially targeting primary care physicians. In 2013 was launched the More Doctors Program (Programa Mais Médicos -PMM). It was set up to reduce the shortage of physician in vulnerable areas and regional disparities in access to health 15 .
The positive impact of PMM on health indicators has been shown by an increasing number of studies 16,17 . However, there is a lack of evidence of the effect of the program towards a more equal distribution of physician at national level. Some studies have address this topic, but at the regional/local level or in a selected group of municipalities 18,19 . The present study asses the effect of More Doctor Program on the geographic distribution of PCPs and its contribution to a more equal distribution of PCPs in Brazil.

Brazilian strategies to deal with the maldistribution of physicians
Since the military dictatorship period, Brazil has designed programs aimed at attracting and retaining health professionals in underserved areas. Although these programs range in size, scope and implementation path, the common purpose is to address the problem of lack of health professionals. Some of these programs had specific actions to recruit and attract physicians.
The Program for Internalization of the Unified Health System (Programa de Interiorização do Sistema Único de Saúde -PISUS) was created in 1993. The PISUS was a cooperation of the three branches of government targeting municipalities without health services. It had the goal of retaining at least one physician with residence in the selected municipality; besides the installation of a small healthcare center with basic equipment and healthcare team. Due to the lack of political support, the program was extinguished even before it was effectively implemented 20 .
The Program for Internalization of Health Work (Programa de Interiorização do Trabalho em Saúde -PITS) emerged in 2001. PITS used financial incentives and professional qualification to encourage physicians and nurses to move to selected municipalities. Initially, it was selected only municipalities with population of up 50 thousand inhabitants, no Family Health Program coverage, high infant mortality rate (80 per 1,000 live births), low indicator of medical consultation and participation in the Active Community or Alvorada Project. The Program selected 151 municipalities. In the second year of the program, some of the eligibility criteria of the municipalities were changed and 216 municipalities were selected. In additional to unfilled vacancies, the program registered the desistance of some professionals and municipalities in a few months. PITS ended in 2004 20 .
In 2010, the federal government adopted new strategies to address the problem of physicians' distribution. The Law No. 12,202/2010 decreed that physicians, who had their courses funded by the Fund for Financing Student in Higher Education (Fundo de Financiamento ao Estudante do Ensino Superior -FIES), could have their debts reduced if they worked on the Family Health Program located in areas with difficult retention of professionals 21 .
In 2011, another initiative was the Primary Care Professional Valorization Program (Programa de Valorização do Professional da Atenção Básica -PROVAB). It was created to encourage physicians, nurses and dental surgeons to move to underserved municipalities or areas with vulnerable population. A postgraduate course in family health and additional score of 10% on medical residency tests were offered to health professionals 22 . Initially, the program attracted 380 physicians; in subsequent years it registered more than 3 thousand physicians. Even so, it did not meet all the shortage recorded by the municipalities 23 .
In January of 2013, the National Front of Brazilian Mayors (Frente Nacional de Prefeitos do Brasil -FNP), claiming support from the federal government, launched the campaign "Where is the Doctor" (Cadê o médico?). In October of 2013 was instituted the More Doctors Program (Programa Mais Médicos -PMM). It was created to reduce the shortage of physician in the priority SUS areas, minimize inequalities and strengthen the provision of primary care 15 . The actions of the program follow 3 bases: emergency supply of physicians; expansion and improvement of the infrastructure of primary healthcare centers and; creation of new medical schools and qualification of medical workforce 23 . PMM and PROVAB coexisted in parallel until the beginning of 2015, when PROVAB became part of the PMM.
PMM followed a combined set of criteria to classify the priorities regions, including: areas with high percentage of the population living in extreme poverty; low human development index; municipalities located in semi-arid and Amazon region; areas with indigenous and Maroon population and; areas with difficult to recruit and retain workforce 15 . After two years, the program recorded the participation of 18,240 physicians (Brazilians and foreigners) in 4,058 municipalities and 34 indigenous health districts 24 .

Methods
This study investigates the effects of PMM on geographic distribution of primary care physicians and uses two years for comparison purposes, 2012 and 2016 (before and after PMM im-plementation). The analysis includes data from 5,564 municipalities.

Data
Primary care physician density is defined as the total number of PCPs per 10,000 population in each municipality. The PCP data source is the Information Department of Brazil's Unified Health System (DATASUS) from the Brazilian Ministry of Health. The TabWin 32 (Tab for Windows) program was used for collect the data. Primary care physicians include physicians from the Family Health Strategy (FHS) and Primary Health Teams (PHT). It was included only PCPs who work in healthcare centers designed to provide primary health care to the population (Centro de Saúde/Unidade Básica de Saúde, Posto de Saúde and Centro de Apoio a Saúde da Família -CASF). Physicians working in hospitals are not included. In Brazil, it is known that physicians can work different number of hours and/ or in more than one municipality. So, for a more accurate number of the provision of this service and avoid an overestimation, the total number of PCPs is calculated dividing the total hours worked weekly in the municipality by 40.

Analysis
The study uses different approaches to analyse the inequality in the distribution of PCPs. First, it was employed spatial data analysis, where the geographic distribution of PCPs and patterns of spatial association (clusters) were displayed. The maps of geographic distribution provide some indicators regarding the inequality that exist across Brazilian municipalities. The cluster maps detect patterns of spatial dependence between neighbouring municipalities. Spatial dependence (or spatial autocorrelation) can be interpreted a systematic pattern in the values recorded in geographic space 25,26 . To identify local patterns of spatial association, it was calculated the Local Indicator of Spatial Association (LISA). In formal terms, LISA statistic can be expressed as follows 27 : Where x i,t represents the observation in municipality i and year t; μ t corresponds the mean of the observation across municipalities in year x i,t -μ t m 0 t and the summation over j includes only neighbouring values of j.
Then, the Lorenz curves and Gini coefficient were used to evaluate the level of inequality in the distribution of PCPs. These measures have been widely employed to analyse the distribution of physicians [28][29][30] . Lorenz curve illustrates the entire distribution and allows to compare how the structure of inequality change over time. Lorenz Curves is displayed with cumulative population proportion on the horizontal axis and cumulative proportion of PCPs on vertical axis. The diagonal line represents the complete equity. This study follows the formal definition of Lorenz curve given by Jann 31 .
The Gini coefficient is derived from the Lorenz curve and it is calculated by the ratio of the area between the Lorenz curve and the diagonal line, and the whole area bellow the diagonal line. The closer the Lorenz curve is the equity line (diagonal), the smaller is Gini coefficient. It is also possible to compare two Lorenz curves. If the curves do not cross, the Lorenz curve closer to the diagonal line has the lower Gini coefficient. Gini coefficient ranges from 0 (perfect equality) to 1 (perfect inequality) 32 .
The following equation was estimated to calculate the time trend in the Gini coefficient: Where G i is the Gini coefficient for the unit of analysis i and ε i denotes the error term. The equation (1) was estimated for Brazil and each state (26 states) and region (5 regions).

Results
The number of PCPs has increased considerably after the implementation of More Doctor Program. The average number of PCPs in Brazilian municipalities increased from 3.45 per 10,000 population in 2012 to 3.93 per 10,000 population in 2016. Figure 1 shows the kernel density plots of PCPs before and after PMM. There was an increase of PCPs in municipalities with shortage of these professionals. In 2012, there were 43 municipalities without PCP and 952 municipalities with less than 2 PCPs per 10,000 population. In 2016, there was an expressive reduction of municipalities with a low number of PCPs. It was recorded 13 municipalities without PCP (70% decrease) and 500 municipalities with less than 2 PCPs per 10,000 population (47% decrease).
Brazilian municipalities have an expressive uneven distribution of physicians. The municipal number of PCPs ranged from 0 to 16.2 PCPs per 10,000 population in 2012 and from 0 to 16.1 PCPs per 10,000 population in 2016. The municipalities with shortage of physicians are located in all Brazilian regions. However, there is a massive concentration of these municipalities in the North, the poorest region of the country. Figure  2 displays the spatial distribution of the municipalities below and above the average number of PCPs. North region is clearly the area with the lowest density of PCPs and largest spatial concentration in both years (Maps (A) and (B)). Map (C) shows the municipalities that increased the number of PCPs after PMM implementation. In general, although the same municipalities remain below average in both years, it is noted that they received more physicians in 2016. The distribution of PCPs has become more equal after PMM implementation in 4 (North, Southeast, Northeast, South) out of 5 regions.
In North, Gini coefficient reduced by 23% from 0.300 in 2012 to 0.230 in 2016. North is the region where was identified a large low-low cluster, as showed in Figure 3.
It was identified a statistically significant trend towards a more equal distribution of PCPs in 21 out of 26 states, with Gini coefficient change ranging from -37% to -6%. The distribution of PCPs has become more unequal in 3 states, but only Goiás exhibited a statistically significant trend toward increasing inequality (p<0.05). Despite substantial improvements in PCP distribution, there is still an expressive difference in the level of equality in physician distribution in 2016, especially across states. São Paulo is the state in which PCPs are most unequally distributed (0.341), while Piauí is the most equal state (0.0093).

Discussion
This paper addresses the problem of distribution of physicians in the Brazilian municipalities. Since the 70's, the government has adopted strategies to recruit and retain physicians in underserved areas and regions with shortage of health professionals. In general, these strategies have been targeted towards primary care and focused to ensure the equitable access of health service around the country 20 .
In 2013, More Doctors Program was instituted to provide physicians to remote and deprived areas. This study presented an overview of the effect of More Doctor Program on the distribution of PCPs and assessed its contribution to a more equal distribution of PCPs in Brazil. Between 2012 and 2016, overall Gini coefficient decreased by 11% from 0.255 to 0.227. At the state level, a statistically significant trend towards a more equal distribution of PCPs was found in 21 Brazilian states.
It was identified only one study that had applied the Gini coefficient to evaluate inequality trends in the distribution of health workers in Brazil. Sousa et al. 33  These coefficients are larger than the found in this study. However, they measured the inequalities in the distribution of all physicians, including specialists, which it is much more concentrated than primary care physicians. The ratio of specialists to generalists in Brazil is 1.67, with a significant difference between the regions, for example, the ratio of specialists to generalists is 2.27 in South and 1.06 in North 34 .
The results are consistent with previous studies, which show that the recent increased in the  level of PCPs has contributed towards a more equal distribution of PCPs in Brazil. The More Doctor Program has been succeeded in providing physicians to vulnerable and deprived areas 18,35,36 , including semi-arid zone 37 , indigenous areas 19 , Maroon communities 38 and rural Amazon 39 .
After the implementation of PMM, the number of municipalities without PCP decreased by 70% and the number of municipalities with less than 2 PCPs per 10,000 population decreased by 47%. Girardi et al. 40 identified that the number of municipalities with shortage of physicians decreased by 35% after implementation of the program, 1,200 municipalities in 2013 and 777 municipalities in 2015. They used the Primary Healthcare Physicians Shortage Index, which takes into account, besides the number of physicians, the infant mortality rate, the access to health service (distance) and percentage of poor households.
The cluster analysis evidence the uneven distribution of these professionals faced, especially, in the North region. This is the region with highest proportion of municipalities with scarcity of physicians 40 . The large low-low cluster may indicate that these municipalities and its neighbours follow similar policies, and may also have closely resembling regional characteristics. After the implementation of PMM, North was the region with largest drop in Gini coefficient. This expressive reduction of scarcity of physician in the North has been reported in the literature 36,40 .
Despite the lowest Gini coefficient, Northeast is the region that received the largest number of PMM physicians 41 . This is explained by poor social and health indicators in the region, also criteria considered by the program 19 . In general, the limited access to healthcare is often associated with poor socioeconomic indicators, such as lower income and higher levels of social vulnerabilities 42 . This combination of limited access to health care and poor socioeconomic indicators contributes to aggravate the health status of the population.
In conclusion, findings in this study evidence that the implementation of PMM has contributed to reduce inequalities in the distribution of PCPs in Brazil. This experience shows the relevance of policies oriented at promoting a more equal distribution of physicians, especially in a context with notable disparities in access to healthcare. However, although the recent increase in the number of PCPs has led to systemic changes, there is still substantial difference in the level of equality in PCP distribution across geographic regions.