Primary health care performance according to clusters of convergent municipalities in the state of São Paulo

: Objective: To describe the performance of Primary Health Care, according to conglomerates of São Paulo cities that present homogeneous indicators. Methods: This is a descriptive study, based on secondary data extracted from official sources of the Unified Health System, for the year 2018. An analysis matrix was created, with the proposition of performance (access, effectiveness and adequacy) and context indicators (population, health determinants and financing) selected and organized in dimensions and sub-dimensions. Cluster Analysis was used to identify the groups of homogeneous municipalities. Results: 645 municipalities were divided in 6 conglomerates. Clusters 2 and 3 were formed predominantly by small municipalities with greater access to health; cluster 3 has less social vulnerability and greater investment in health. Clusters 1, 4 and 5 were formed by the largest municipalities with less access to health; cluster 4 presents greater social vulnerability, less coverage of private health plans and a greater percentage of health resources; cluster 5 was characterized by greater Gross Domestic Product per capita and greater coverage of private health plans. Cluster 6, formed by the city of São Paulo, was a particular case. Cluster 2 drew attention, as it was shown to have increased coverage, but signaled lower efficacy and adequacy levels. Cluster 3 had the best performance among all clusters. Conclusion: These findings can support regional and municipal management, given the complexity of the territory of São Paulo, pointing to scenarios that demand broader public management initiatives.


INTRODUCTION
The effort to build primary health care (PHC) over almost three decades of the Unified Health System (SUS) requires permanent evaluation and monitoring (E&M) not only of specific processes, but as a whole, in a managerial effort to obtain desirable effects on people's health. During this period, the participation of different spheres of government -federal, state and municipal -in running the SUS (Law No. 8080, of September 19, 1990), as well as in the constitution of a National Primary Care Policy (PNAB) (Ordinance No. 648, of March 28, 2006)  With regard to the process of expanding municipal responsibilities in the operation of PHC (decentralization) and despite advances made by and for the SUS, such as the Program "More Physicians" ("Mais Médicos") and the National Program for Improving Access and Quality of Primary Care 2 , the structural inequality and the heterogeneity of offer, quality and resolution potential of services point to challenges towards territorial equity and the integral attention to the population's health 3 . The most recent change in the federal funding model of PHC is one of the challenges with individualizing approach, being carried out according to criteria such as weighted capitation and payment for performance 4,5 . RESUMO: Objetivo: Descrever o desempenho da atenção primária à saúde, segundo clusters de municípios paulistas que apresentaram indicadores homogêneos. Métodos: Trata-se de um estudo descritivo, com base em dados secundários extraídos de fontes oficiais do Sistema Único de Saúde, referentes ao ano de 2018. Foi elaborada uma matriz de análise, com a proposição de indicadores de desempenho (acesso, efetividade e adequação) e contexto (população, determinantes de saúde e financiamento), selecionados e organizados em dimensões e subdimensões. Para identificar os grupos de municípios homogêneos, foi utilizada a análise de cluster. Resultados: Dos 645 municípios, constituíram-se seis clusters. Os clusters 2 e 3 foram formados, predominantemente, por municípios pequenos e com maior acesso; entre eles, o cluster 3 apresentou menor vulnerabilidade social e maior investimento em saúde. Os clusters 1, 4 e 5, em contrapartida, foram formados por municípios maiores e com menor acesso; entre eles, o cluster 4 apresentou maior vulnerabilidade social, menor cobertura de planos privados de saúde e maior percentual de recursos utilizados em saúde; e o cluster 5, maior produto interno bruto per capita e maior cobertura de planos privados de saúde. O cluster 6, formado pelo município de São Paulo, demonstrou ser um caso particular. Ainda, o cluster 2 chamou atenção. Apresentando maior cobertura, sinalizou menor efetividade e adequação. Entre todos os clusters, o cluster 3 alcançou o melhor desempenho. Conclusão: Os resultados podem subsidiar a gestão regional e municipal, diante da complexidade do território paulista, apontando para cenários que demandam maiores inciativas de gestão pública.
In the state of São Paulo, the federative unit with the largest estimated population (22% of the country's total) and the greatest wealth produced in Brazil in 2019 6,7 , PHC services are distributed across 645 municipalities. The process of expansion of the FHS in the state, which already had an extensive network of health centers at the time of creation of SUS, was slower when compared to other regions of the country. Between 2000 and 2009, however, a significant expansion of the FHS (from 6.31 to 27.96%) took place, being more consistent in municipalities with smaller populations 8 . Until May 2020, the state of São Paulo still had low coverage percentages both for the FHS model (41.48%) and the technological arrangements of PHC (60.19%) 9 .
In this scenario, the E&M of health services' performance characterizes the methodological need to stratify complex realities and base the planning of collective actions in the review and reorientation of PHC 10,11 . From a multidimensional approach to health indicators, Viacava et al. (2004) presented the Health System Performance Assessment Project (Proadess) 12 , where performance was strongly associated with the structure of the system, having the political, social, economic and structural context as dimensions of analysis and considering equity as a transversal feature 13 . Cluster analysis, a rapid and economical sampling methodology, enables the characterization of homogeneous clusters, which helps to identify critical areas, groups of greater health needs and evidence-based practice (decision making) 14 .
Although the literature describes E&M of performance (access, effectiveness and adequacy) proposed by Proadess 15,16 , there are still gaps when it comes to addressing the municipal and regional context of the state of São Paulo. Given the possibility of building health indicators and methodologies that address collective needs to a certain extent, the objective of this study was, therefore, to describe the performance of PHC according to clusters of municipalities in São Paulo that presented homogeneous indicators.

STUDY TYPE AND POPULATION
This is a descriptive study based on secondary quantitative data, publicly accessible and obtained from the Health Information Systems (SIS) of SUS, on the performance of PHC in the state of São Paulo, by means of cluster analysis. Units of analysis were the municipalities with convergent indicators of the sanitary, socioeconomic and structural situation of the municipal health system in 2018.

INDICATORS AND ANALYSIS MATRIX
In order to monitor access, effectiveness and adequacy of PHC in different population groups (women, adults, children, and elderly), the methodological model adapted from Proadess 17 was used. To build the municipal health scenario through a set of indicators, the properties of synthesis measures were consulted and the PHC attributes were considered. Considering what was defined for the validity and reliability of basic health indicators 18 , the selection was cautious, aiming at sensitivity, measurability, relevance, cost-effectiveness and integrity of data.
Twenty-five indicators were selected, classified and distributed into dimensions and sub-dimensions of an analysis matrix: • core dimension of (1) Performance, with sub-dimensions of (1.1) Access (two indicators), (1.

STATISTICAL AND DESCRIPTIVE ANALYSIS
The indicators were calculated for each municipality and interpreted in their frequency measures. To identify groups of municipalities with homogeneous analytical characteristics, a multivariate cluster analysis 20 was used, which allowed to identify interdependence between variables characterizing each unit of analysis (municipality).
Initially, the means of 25 indicators were normalized into standard scores (or z-scores), considering the variability for each observation. The quadratic Euclidean measure (similarity) and Ward's minimum variance (Ward's method) were used as parameter for processing the analysis with a hierarchical method. Then, the clusters were visualized and defined by the minimum distance between variables belonging to each cluster (homogeneity) and the maximum distance between them (heterogeneity). To support the final number of clusters and their respective set of municipalities, a dendrogram was consulted. The analyses were performed in the statistical software Stata ® 15.
In the analysis matrix, the indicators were described according to medians corresponding to each cluster. The median, representing 50% of the distribution scale, was preferred due to the distribution of data, with the presence of outliers 21 . Then, the medians of indicators of each cluster were compared with each other and with the median values for the state of São Paulo. The expressions "minor", "intermediate" and "major", scaled from the smallest to the largest median, were used to characterize the behavior of indicators for each cluster, with the values of the state of São Paulo as a comparison parameter. Finally, to characterize and differentiate the distribution of clusters, the municipalities were georeferenced in the MapInfo ® Pro v2019 software.

RESULTS
Six clusters were built out of the 645 municipalities in the state of São Paulo: cluster 1, cluster 2, cluster 3, cluster 4, cluster 5 and cluster 6. The municipalities encompassed by each cluster are described in Supplementary Material 2. Table 1 shows the set of indicators expressed in median value, organized by subdimension and for each cluster, as well as the median value for the state of São Paulo.

CLUSTERS' CHARACTERIZATION
•  Effectiveness subdimension, the post-neonatal mortality indicator stood out, with a median equal to 0, and the highest percentage of hospitalizations for ACSC among clusters and compared to the state of São Paulo. In the Adequacy subdimension, there was a higher proportion of cesarean deliveries among all deliveries and among deliveries performed within SUS (total deliveries and deliveries-SUS) and a lower ratio of cervical cancer screening exams in women aged 25 to 64 years (25-64 years). The Context dimension had the sectors with higher vulnerability, higher percentage of population benefiting from private health plans, lower total health expenditure per capita, and higher percentage of budgetary resources used in health; • Cluster 2: made up of 62 municipalities, it had the second smallest population.
The Access subdimension showed the second largest coverage of PHC and oral health teams. In the Effectiveness subdimension, it had the lowest rates of hospitalizations for asthma in children under 10 years of age and all ages (0-9 years and all ages) and for acute respiratory failure in children under 5 years, as well as the highest rates of infant, neonatal and post-neonatal mortality, the highest proportion of newborns born to adolescent mothers under 20 years of age, and the second highest percentage of hospitalizations for ACSC. The Adequacy subdimension had a higher proportion of cesarean deliveries (total deliveries and deliveries-SUS) and a higher ratio of cervical cancer screening exams in women (25-64 years). In the Context dimension, there was a higher proportion of elderly people, sectors with high vulnerability, the lowest percentage of population benefiting from private health plans, the highest total health expenditure per capita and the lowest percentage of budgetary resources used in health; • Cluster 3: formed by 86 municipalities, it had the smallest population. The Access subdimension showed the highest coverage by PHC and oral health teams. In the Effectiveness subdimension, the highlights were hospitalization rates for asthma (0-9 years and all ages), infant, neonatal and post-neonatal mortality, and syphilis detection in pregnant women-all with medians equal to 0-, the highest hospitalization rate for stroke in individuals aged 30 to 59 years, higher rate of hospitalization for acute respiratory failure in children, and the lowest proportion of low-birth-weight newborns. The Adequacy subdimension had the highest proportions of cesarean deliveries (total deliveries and deliveries-SUS), the highest percentage of pregnant women who had seven or more prenatal consultations, and the most reasons for cervical and breast cancer screening tests in women (25-64 years, 50-69 years, respectively). The Context dimension had the highest proportion of elderly people, absence of a median in sectors with high vulnerability, the lowest percentage of the population benefiting from private health plans, and the highest total health expenditure per capita; • Cluster 4: formed by 130 municipalities, it had the largest population. The Access subdimension had the second lowest coverage of PHC and oral health teams. The Effectiveness subdimension had the lowest rate of hospitalization for stroke (30-59 years), the highest proportion of newborns born to adolescent mothers (<20 years), the highest detection rate of syphilis in pregnant women and the lowest percentage of hospitalizations for ACSC. The Adequacy subdimension had a lower proportion of cesarean deliveries (total deliveries and deliveries-SUS) and the lowest ratio of mammography for breast cancer screening in women (50-69 years). The Context dimension had the lowest proportion of elderly people, the highest percentage of sectors with high vulnerability, the lowest per capita gross domestic product (GDP), the lowest percentage of the population benefiting from private health plans, the lowest total expenditure on health per capita, and the highest percentage of budgetary resources used in health; • Cluster 5: formed by 145 municipalities, it had a large population. In the Access subdimension, intermediate results for coverage of PHC and oral health teams. The Effectiveness subdimension had the lowest rates of hospitalization for stroke (30-59 years) and for acute respiratory failure in children (<5 years), lower post-neonatal mortality rate, higher syphilis detection rate in pregnant women and lower percentage of hospitalizations for ACSC. The Adequacy sub-dimension had a lower percentage of pregnant women who had seven or more prenatal consultations. The Context dimension had a lower proportion of elderly people in the population, sectors with high vulnerability, higher GDP per capita, higher percentage of population benefiting from private health plans, and lower percentage of budgetary resources used in health; • Cluster 6: formed by the city of São Paulo, it had the largest population in the state.
The Access subdimension had the lowest coverage of PHC and oral health teams. The Effectiveness subdimension had the highest rate of hospitalization for asthma (0-9 years and all ages) and acute respiratory failure in children (<5 years), the lowest proportion of newborns born to adolescent mothers (<20 years), the highest proportion of low-birth-weight newborns, the highest rate of syphilis detection in pregnant women and the lowest percentage of hospitalizations for ACSC. The Adequacy subdimension had the lowest proportion of cesarean deliveries (total deliveries and deliveries-SUS), the lowest percentage of pregnant women who had seven or more prenatal consultations, less reasons for cervical and breast cancer screening tests in women (25-64 years and 50-69 years, respectively). The Context dimension had a lower proportion of elderly people in the population, a higher percentage of sectors with high vulnerability, the highest GDP per capita, the highest percentage of the population benefiting from private health plans, the lowest total health expenditure per capita, and the lowest percentage of budgetary resources used in health.

CLUSTERS' SPATIAL CHARACTERIZATION
Some of the clusters had a marked geographic distribution in the state of São Paulo: the municipalities in cluster 1, the largest, appeared in almost the entire territory, except for the southern region, where cluster 4's municipalities were predominant; the municipalities in  cluster 2 were from the central and northwest regions of the state, being practically absent in the east and the south; cluster 3, in turn, had municipalities from the northwest region; the municipalities in cluster 4 were predominantly from the southern region and the northern coast; the municipalities in cluster 5, however, were mainly metropolitan areas such as São Paulo, Baixada Santista and Campinas, in addition to municipalities in the regions surrounding São José dos Campos, Piracicaba, Ribeirão Preto and Marília (Figure 1).

DISCUSSION
Our results allowed us to describe and discuss the different performance standards of PHC in the state of São Paulo in 2018 by means of cluster analysis.
In the analysis of access, based on population characteristics and considering that a medical or oral health professional, working 40 hours a week, should cover a population of three thousand inhabitants 22 , clusters 2 and 3 were predominantly made up of municipalities with low and high coverage; clusters 1, 4 and 5, on the other hand, have larger municipalities and lower coverage; cluster 6, formed exclusively by the city of São Paulo was a particular case, with the largest population in Latin America (11,753,659) and the lowest coverage. Inequalities in access corroborated the contextual analysis: cluster 3 had low social vulnerability and high investment in health; cluster 4 had with lower coverage, high social vulnerability and less investments in health; cluster 5 had a higher GDP per capita and high coverage of private health plans.
Small municipalities stand out for the greater coverage of PHC, in view of the turnover of professionals who are part of the health teams and lower coverage of private health plans 23 . In addition, in most of these municipalities, the economy is fragile by nature, reflected as difficulty in paying for basic services to the population 24 . For the larger cities, the clusters with the highest population medians showed characteristics that elucidated the effects of the urbanization process and the social problems in these locations: low coverage of the FHS, limitations of effectiveness and resolution capacity of services, socioeconomic inequality, vulnerability, and a fragmented care network 3,25 .
Although regionalization drives the establishment of Health Care Networks to strengthen, in a universal and equitable way, the integrality guaranteed by the SUS 26 , the structural conditions must be reversed, given the deficit in the collection of the smallest municipalities and the chronic federal underfunding 4,27 , to enable the coordinating character of PHC. In the reality of the largest centers, in addition to ensuring the financial sustainability of a regionalized PHC model, the modernization of E&M is essential to overcome, in line with urban and population development, the constraints in the expansion of the FHS in geopolitical territory 28 .
In the effectiveness analysis, the clusters helped to visualize the consensus in the scientific literature: greater PHC coverage improves the overall health condition of the population 29,30 . Cluster 3, from this perspective, presented the best performance for the set of indicators; the city of São Paulo (cluster 6), almost in ambivalence, presented the worst performance. The behavior of the indicators, in all clusters, put the spotlight on the debate around maternal and child health and chronic health conditions of adults and the elderly, with emphasis on the metropolitan regions [31][32][33][34] .
However, cluster 2 draw attention, as it had municipalities with greater coverage of PHC and serious performance issues distributed in the state of São Paulo. With the highest infant, neonatal and post-neonatal mortality rates, the results of cluster 2 indicate a direct relationship with the socioeconomic conditions of these municipalities 35 . The greater number of hospitalizations by ACSC, in turn, pointed to a deficiency in the integration and quality of care 36 , with greater coverage of PHC not achieving satisfactory results. Thus, it is essential to understand the peculiar factors of municipal health systems in the regional contexts of the state of São Paulo, especially in the northwest region.
In the analysis of adequacy, the higher proportion of cesarean deliveries in smaller municipalities was added to the discussions on quality of prenatal care, bearing in mind the criticisms of the surgery, as well as clinical/organizational conducts established in health units 15,37,38 . In this sense, the results brought to light the importance of taking action to deal REV BRAS EPIDEMIOL 2022; 25: E220017 with teenage pregnancy, with a possibility to access to sexual and reproductive health educational policies 39 . The detection and treatment of precursor lesions of cervical and breast cancer in women, especially in the largest centers, should be expanded in a timely and targeted manner 40,41 .
In an effort to quantify, describe and present scenarios and health needs, we found inconsistencies and incoherencies in PHC performance indicators as quality criteria. The overall performance, despite limitations of secondary quantitative data from different SIS, was unsatisfactory and marked by health diversities, which homogeneously indicated sanitary and structural conditions that should be prioritized. The importance of accelerating the expansion of the FHS in the state of São Paulo is reinforced to improve the health of the population, especially in the largest centers, but the inadequacy of PHC coverage requires qualitative interventions to reorient the practice of health care when associated with programmatic vulnerabilities.
Finally, the analysis of performance by convergence stratified inequities in health for which the PHC un-funding policy, especially from 2020 onwards, tends not to cover, and it is up to the State to train and structure municipal and regional management to routinely promote prompt E&M for decision-making and budget restructuring. In a macro-political view, it is a priority to invest in the adequacy of the democratic model of PHC, in the access to technical-scientific advances and in the elaboration of a PNAB that enables the continuity of management, work and health care, with focus on the FHS and on the sense of social security and citizenship.