Pneumonia in Children under 5 Years: Temporal Trends and Spatial Patterns of Hospitalizations in Brazil

ABSTRACT.

This study aimed to identify temporal and spatial patterns in the distribution of hospitalization rates for pneumonia in children under 5 years in Brazil. An ecological study was developed using data from the Unified Health System of hospitalizations for pneumonia in children under 5 years in Brazil from 2000 to 2019. Hospitalization rates per 1,000 children were calculated and Joinpoint Regression analyzed the temporal trends. Different spatial analysis techniques were performed. Annual rates of 25 hospitalizations per 1,000 children were observed in 2000 and of 13.83 per 1,000 children in 2019, with a significant downward trend for the country (annual percentage change = −3.4%; 95% confidence interval: −3.8 to −3.0) and for the regions. There was weak spatial autocorrelation; however, there were regions with high rates of hospitalization in the south region and clusters of low rates in the northeast and southeast. Clusters of areas with high hospitalization rates were observed in areas of favorable socioeconomic conditions and provision of health services in the interior of southern Brazil. There is a decreasing trend in hospitalizations for pneumonia in general; however, there are clusters of high rates in the south of Brazil.

INTRODUCTION

Respiratory diseases are one of the main causes of hospitalization of children and one of the main challenges encountered by health services in terms of detection, diagnosis, and treatment.¹ Among them, pneumonia remains the main cause of morbidity and mortality among children under 5 years of age worldwide, responsible for approximately 2,000 deaths daily, which corresponds to 15% of all deaths of children in this age group.² Of the total number of pneumonia cases, 7% to 13% require hospitalization and advanced care,^3,4 and although pneumonia can be treated with low-cost drugs, < 30% of affected children have access to them.²

Fifteen countries have 74% of the cases of pneumonia in children under 5 years, and Brazil is in 15th position.⁵ In Brazil, approximately 4 million cases of pneumonia in children occur annually,⁶ representing the second cause of hospitalization in the country, accounting for approximately 14% of all hospitalizations.⁷ Although a reduction in mortality from this cause was observed in Brazil between 1990 and 2015, acute lung diseases are responsible for 11% of deaths in children under 1 year, and for 13% among children aged 1 to 4 years.⁸ The main risk factors associated with hospitalization for pneumonia, related to the children, are poor nutritional status, absence of breastfeeding, insufficient weight gain during pregnancy, and low birth weight. The main risk factors associated with hospitalization for pneumonia related to the children’s living conditions are low parental education level, maternal age, smoking in the home, and number of residents in the household, among others.⁹ This disease has negative impacts in developing countries and in regions of high social inequality, where there is a shortage of human and financial resources.¹⁰

In this context, hospitalization for pneumonia can be considered a sensitive marker of deficiencies in the performance of the health system and an expression of social inequalities.¹¹ The incidence of pneumonia is related to less favored social contexts, an insufficient and inefficient healthcare network, low Human Development Index (HDI), among other factors.¹²

To reduce childhood pneumonia mortality in Brazil, one strategy adopted by the Ministry of Health was the implementation of the Pact for Life, the result of strategic and financial efforts of managers to meet the main demands of the Brazilian population, subdivided into some agreed priorities. The objective of the strategies of the Pact is to reduce childhood pneumonia mortality by 20%.¹³ Another strategy that contributes to reducing infant deaths due to this cause is the implementation of pneumococcal vaccines, especially 10-valent vaccines, in the National Immunization Program (NIP). Since 2010, with the introduction of these vaccines into the children’s vaccination calendar, a reduction in the number of hospitalizations and deaths was observed.¹⁴

Pneumonia is on the list of ambulatory care sensitive conditions (ACSC), and the Ministry of Health recommends that this indicator be used as an instrument to assess primary care and/or hospital care at the national, state, and municipal levels. In this sense, the analysis of hospitalizations for pneumonia can help identify areas that can clearly be improved, highlighting health problems relating to pneumonia that require better follow-up and coordination between care levels.¹⁵

In this perspective, studies that identify the temporal and spatial trends of hospitalization for pneumonia are important instruments for understanding the reality of the occurrence of this disease, as well as for enabling decision-making and strategic planning about measures for the prevention and control of respiratory diseases, especially in countries like Brazil, which have a sizeable territorial extension with significant inequalities in income, infrastructure, and availability of health services.¹⁶

Since the 1990s, the public health system in Brazil has functioned under principles of universal access, comprehensive care, equity, and social participation. The Unified Health System (SUS, from the Portuguese acronym) develops actions to prevent diseases, and provide health promotion, health surveillance, treatment, and rehabilitation for all Brazilian citizens, in addition to guaranteeing continuity of care at different levels (primary health care, specialized care, and hospital and urgent and emergency care), thus enabling the expansion of access to health actions and services in the country.¹⁷

The primary source of information for analyzing hospitalizations in the SUS in Brazil is the Hospital Information System of the SUS (HIS-SUS), which contains the records of information from SUS-funded hospital admissions. These data are valid for planning, financial transfers, health service management, and epidemiological and health surveillance investigations.^18,19

In this context, this study aims to identify temporal and spatial patterns in the distribution of hospitalization rates for pneumonia in children under 5 years of age in Brazil in the past 20 years.

MATERIAL AND METHODS

Study design.

This is a population-based ecological study with a mixed design on temporal trends and multiple groups, based on secondary data recorded in the HIS-SUS,¹⁸ from the Informatics Department of the SUS.

Variables.

The outcome variable was hospitalizations for pneumonia of children under 5 that occurred in Brazil in from 2000 to 2019, by place of residence.

Hospitalization rates per 1,000 children were calculated for Brazil, large geographic regions and by Immediate Regions of Urban Articulation (IRUA), according to the following formula:

$$\begin{array}{l}\text{Hospitalization rate}\\ =\frac{\left\{\begin{array}{l}“\text{Number of hospitalizations for pneumonia}\\ \hspace{1em}\text{ in children under }5\text{ covered by SUS}”\end{array}\right\}}{\left\{\begin{array}{l}“\text{Population of children under }\\ \hspace{1em}5\text{ residing in the period considered}”\end{array}\right\}}\times 1000\end{array}$$

The IRUA are structures delimited by the federal government and formed by clusters of urban centers that interact to meet the population’s demands for health services, education, and consumer goods, in addition to the search for work.

The formula expresses the number of cases of hospital admissions for pneumonia in children under 5 covered by the SUS per 1,000 inhabitants, in the population of children under 5 years of age resident in a given geographic space. Population data by geographic area and by age were obtained from demographic censuses and inter-census projections on the website of the Brazilian Institute of Geography and Statistics.²⁰

Units of analysis.

For the analysis of temporal trends, hospitalizations for pneumonia in children under 5 were analyzed in the national territory and in the five main geographic regions (north, northeast, southeast, south, and midwest), according to the political organization of the country.

The IRUA were considered for spatial analyses because such organization aggregates in its construction process issues related to the urbanization and integration processes of the national market, which establish network relationships and strengthen cities and urban agglomerations as fundamental elements for the interconnection of management, infrastructure, and productive activities.²¹ The IRUA, in general, reflect the area where the population lives and where their daily commuting to supply and search for common goods and services such as health services, purchase of consumer goods, and specialized services in general takes place.²² Analyses were conducted per IRUA instead of municipalities to minimize the bias of calculating rates for small areas.

Statistical analysis.

Joinpoint Regression was used for the analysis of temporal trends in hospitalization rates (SUS) for pneumonia in children in the national territory and in large geographic regions. In this model, the annual percentage change (APC) was estimated, and the final selected model was the best adjusted, based on the trend of each segment, checking for statistically significant values (P < 0.05). To quantify the trend in the analyzed cohort of years, the average APC (AAPC) was calculated based on the accumulated geometric average of APC trends, with equal weights for the lengths of each segment during the fixed interval, checking for statistically significant values (P < 0.05). The significance tests used were based on the Monte Carlo permutation method and on the calculation of the annual percentage variation of the ratio, using the logarithm of the ratio.^22,23 The time series analysis was performed in the Joinpoint Regression software program (National Cancer Institute, Bethesda, MD), version 4.6.0.0.²⁴

A univariate spatial analysis was performed with the outcome variable to verify whether there was spatial dependence in the occurrence of hospitalizations for pneumonia in the 482 IRUA in Brazil. For this purpose, the average hospitalization rates for 5 years (2000–2004, 2005–2009, 2010–2014, 2015–2019) were calculated. The analysis was performed using the global Moran index to estimate spatial autocorrelation using the random permutation test with 999 permutations. To verify the distribution pattern of the rates, the local Moran test (Local Indicators of Spatial Association) was applied. For this purpose, a MoranMap was presented to assess the formation of clusters, which were classified as high-high, low-low, low-high, and high-low.²⁵

The global Moran index generates a measure of spatial association existing within the data set. Values vary from −1 to 1; values approaching zero indicate absence of significant spatial autocorrelation of the values with their neighbors, values below 0.50 indicate a weak autocorrelation, values between 0.50 and 0.75 indicate medium autocorrelation, and values above 0.75 indicate strong autocorrelation. Positive values indicate a positive autocorrelation, where the value of what is being evaluated is similar to the values of neighbors, and negative values indicate a negative autocorrelation.²⁶

Rates were calculated for 5 years to minimize random fluctuations in the occurrence of the events.

For the identification of risk areas for the occurrence of the event, the SCAN spatial scanning technique by IRUA was performed for each 5-year period using the Software SaTScan version 9.6.²⁷ In the present study, the number of hospitalizations for the IRUA per each 5-year period was used as the case file; the population of children under 5 in each 5-year period was used as the population file; and the latitude and longitude of the centroid of each IRUA was used as the coordinate file. For the preparation of the maps, stratification was performed by the relative risk (RR), where an RR equal to 1 (RR = 1) is interpreted as the absence of association between exposure and disease—that is, the risk of falling ill is similar in both groups; a RR greater than 1 (RR > 1) indicates a higher risk in exposed individuals (exposure is associated with the disease and is a risk factor); a RR less than 1 (RR < 1) indicates a lower risk in exposed individuals (exposure is associated with the disease and is a protective factor).²⁸

The SCAN technique examines circular windows in which the area enclosed by the window of analysis, called the z region, can form a cluster if the value found is greater or smaller than expected. For each window examined, the null hypothesis (H0) of absence of cluster in the study region is tested against the alternative hypothesis (H1), considering the region z as a cluster.²⁹ In this case, it is assumed that there is a risk of hospitalization for pneumonia compared with the outside window.

A purely spatial analysis was performed using the Poisson discrete model, determining the following parameters: absence of geographical overlap, maximum cluster size equal to 50% of the population at risk and 999 Monte Carlo replications for the significance test of clusters.^30,31

Assuming the process of social determination of health and disease phenomena, we sought to analyze the spatial association between the average hospitalization rate for pneumonia in the past 5 years (2015–2019) by IRUA and the Municipal HDI (MHDI) and provision of health services. For this purpose, the Skater technique or regionalization method was applied. This method allows grouping smaller areas into larger areas that, in addition to being contiguous, have similarities according to the associated attributes, and constitutes an analysis of clusters that takes into account the spatial location of objects.³² The evaluation period took into account the past 5 years because it is more recent data, thus providing a better evaluation of the current period.

The regionalization of areas by the Skater method starts with the construction of the graph, which consists of a diagram composed of points (spatial units with the respective data), from which the neighbors are connected to each other by lines that represent the spatial units that are interrelated. In the Skater algorithm, the assembly of this graph is done by a proximity matrix, and a value is assigned to the edges according to the dissimilarity (distance between the corresponding attributes or values). The Minimum Generation Tree is assembled from this graph, and the regionalization problem becomes equivalent to the partitioning of the graph to achieve the best similarity between regions.^33,34

In this technique, all variables were standardized by the Z score due to the impact that different scales can have on the dissimilarity function and the sum of squares within the clusters. The formation of five clusters was used. For the analysis of the formed clusters, the sums of squares within each cluster and between clusters were analyzed to show the difference between them.

For this analysis, the median values of the independent variables represented by the contextual indicators of the IRUA were presented: (V1) MHDI, (V2) coverage of the Family Health Strategy (FHS), (V3) rate of FHS physicians per 100,000 inhabitants, (V4) rate of pediatricians per 100,000 children, and (V5) rate of pediatric beds per 100,000 children. The MHDI for 2010 was obtained from the Human Development Atlas in Brazil of the United Nations Development Program. Information on the offer of services and health professionals were collected on the web pages of the SUS Informatics Department,¹⁸ from the base of the National Registry of Health Establishments,³⁵ having as reference the month of December 2019.

For the production of thematic maps, the cartographic base of Brazil by IRUA was obtained from the Brazilian Institute of Geography and Statistics website,²¹ and to perform these analyses, the GeoDa software 1.14.0.24 of August 2019³⁶ and the SaTScan version 9.6,²⁷ and the Quantum GIS version 2.18.2³⁷ were used.

RESULTS

From 2000 to 2019, there were 5,777,348 hospitalizations for pneumonia of children under 5 in Brazil. The annual rates of hospitalization for this disease show that Brazil registered a rate of 25.08 hospitalizations per 1,000 children in 2000 and 13.83 per 1,000 children in 2019 (Figure 1). The regions with the highest median hospitalization rates were the Midwest (21.71/1000 children), South (21.19/1000 children) and North (20.39/1000 children), all with values above the national median (18.36/1000 children) (Figure 1).

Figure 1.

Temporal distribution of hospitalization rates for pneumonia in children under 5 years of age (per 1,000 children) in the Unified Health System in Brazil and regions, 2000–2019.

Hospitalization rates for pneumonia in children showed a significant downward trend in the period from 2000 to 2019 in Brazil (AAPC = −3.4%; 95% confidence interval [CI]: −3.8 to −3.0) and for the northeast, south, southeast and midwest, with a more pronounced reduction in the midwest region (AAPC = −4.6%; 95% CI: −5.1 to −4.0). For the north region, there was a significant downward trend between 2011 and 2016 (APC = −9.2%; 95% CI: −14.7 to −3.4), followed by a period of increase, but without statistical significance for the analysis of the entire period (Table 1).

Table 1.

Temporal trend of hospitalization rates for pneumonia in children under 5 years of age in the Unified Health System from 2000 to 2019 for Brazil and its geographic regions

Geographic area	Seg.	Initial year	Final year	APC (95% CI)	P	AAPC (95% CI)	P
North region	1	2000	2011	0.9 (−0.4 to 2.2)	0.2	−0.7 (−2.8 to 1.4)	0.5
	2	2011	2016	−9.2 (−14.7 to −3.4)	< 0.05
	3	2016	2019	8.7 (−1.5 to 19.9)	0.1
Northeast region	1	2000	2019	−4.3 (−5.1 to −3.6)	< 0.05	−4.3 (−5.1 to −3.6)	< 0.05
Southeast region	1	2000	2019	−2.9 (−3.3 to −2.6)	< 0.05	−2.9 (−3.3 to −2.6)	< 0.05
South region	1	2000	2019	−3.1 (−3.5 to −2.8)	< 0.05	−3.1 (−3.5 to −2.8)	< 0.05
Midwest region	1	2000	2019	−4.6 (−5.1 to −4.0)	< 0.05	−4.6 (−5.1 to −4.0)	< 0.05
Brazil	1	2000	2019	−3.4 (−3.8 to −3.0)	< 0.05	−3.4 (−3.8 to −3.0)	< 0.05

CI = 95% confidence interval; Seg. = Segment. Source: Hospitalization Information System of the Unified Health System (SUS), SUS Informatics Department (DATASUS).

The spatial analysis of hospitalization rates showed a weak spatial autocorrelation when analyzed by the global Moran values in the 5-year period studied (I < 0.3). However, the maps showed the formation of clusters of IRUA with high rates of hospitalization, surrounded by neighbors with high rates of hospitalization (high-high) in the midwest region, from 2000 to 2004, and in the south region in all 5-year periods, despite the slight reduction in the high-high clusters over the 5-year period in this region. Also noteworthy is the formation of clusters of low hospitalization rates surrounded by neighbors with low rates (low-low) in the northeast and southeast regions in all 5-year periods; it is noteworthy that there was an increase in Low-Low clusters throughout the 5-year period (Figure 2).

Figure 2.

Moran Map and global Moran values of hospitalization rates for pneumonia in children under 5 years of age in the Unified Health System for 5-year periods from 2000 to 2019 per Immediate Regions of Urban Articulation in Brazil.

The RR maps showed that over the analyzed period, the south and midwest regions presented a higher risk of hospital admissions, with RR values > 1. However, a reduction was observed in the south region during the 5-year period evaluated. In the northeast and southeast regions, the RR was < 1, indicating that they were areas of low risk for the occurrence of the event. The RR values also showed a reduction when analyzing each 5-year period in these regions. On the other hand, there was an increase in RR over the 5-year periods in the north region (Figure 3).

Figure 3.

Relative risk map for the occurrence of hospitalization for pneumonia in children under five years of age in the Unified Health System for 5-year periods from 2000 to 2019 per Immediate Regions of Urban Articulation in Brazil.

The analysis using the Skater technique between the hospitalization rates for pneumonia and their relationship with the MHDI and the provision of health services identified two C3 and C5 clusters composed mainly of the south region, with high rates of hospitalization in areas with MHDI > 0.70, FHS coverage rate > 90%, rate of FHS physicians per 100,000 inhabitants > 0.16, and rate of pediatric beds per 100,000 children > 0.40, with the highest rate among the formed clusters. However, the low rate of pediatricians per 100,000 children in these two clusters was noteworthy (Figure 4).

Figure 4.

Skater map for analysis of hospitalization patterns for pneumonia in children under 5 years of age and their relationship with indicators of health service provision and with the Municipal Human Development Index from 2015 to 2019 according to the Immediate Regions of Urban Articulation in Brazil. The total sum of squares was 2,886; sum of squares within clusters was as follows: (C1) 857.147, (C2) 744.343, (C3) 119.073, (C4) 73.076, and (C5) 72.0202.

There was also the formation of a C4 cluster that presented the lowest rates of hospitalization for pneumonia and the worst indicators of provision of services and health professionals at the same time, with a low rate of FHS coverage, low supply of FHS physicians and pediatric beds, but the highest rate of pediatric doctors. This cluster comprises 23 IRUA in the state of São Paulo (Figure 4).

Cluster C1 is composed mostly of the northeast region, where there was one of the lowest rates of hospitalization, with good coverage rates of the FHS, the highest rate of FHS physicians and supply of pediatric beds with acceptable values. However, the northeast is the region that presented the lowest MHDI and the lowest rate of pediatric doctors. In turn, Cluster C2, corresponding to the north and midwest regions of the country, had low hospitalization rates and a good offer of services and professionals (Figure 4).

DISCUSSION

The results of the present study show that there was a downward trend in hospitalizations for pneumonia in children under 5 in Brazil during the analyzed period and that the hospitalization rates were unevenly distributed among the regions. The IRUA of the midwest and south regions had the highest median hospitalization rates, and the lowest rates were observed in the northeast.

Spatial patterns in the distribution of hospitalization rates for pneumonia in children under 4 years of age from 2009 to 2013 in municipalities of the state of São Paulo, in the southeastern region of Brazil, also showed a reduction in the average hospitalization rates of 25.46 per 1,000 children (in 2009) to 20.46 per 1,000 children (in 2013), being considered an important reduction for the period.⁵

In Korea, researchers identified the rate of hospitalization rate for pneumonia in children aged 0 to 14 years to be 325.3 per 10,000 inhabitants and identified a high geographic variation, with a Moran index of 0.6.³⁸ In Ethiopia, they found that acute respiratory infection among children under five showed spatial variations across the country, with a Moran index of 0.34.³⁹

The decline in hospitalization rates over the 20 years analyzed may be a result of the strengthening of primary health care (PHC) with the strengthening of the FHS through the National Policy of Primary Care⁴⁰ in 2006, and greater access to outpatient health levels, as well as the advent of the Pneumococcal vaccines PCV 7 and PCV 10. Studies state that PHC has the capacity to create strategies to favor health systems, improve health and equity, in addition to providing greater efficiency to services and lower costs.^41–43

The healthcare model adopted by SUS emphasizes the FHS and PHC principles, in which the professionals’ performance is guided by the bond and responsibility of users, generating efforts that help in the use and coordination of the provision of services through Health Care Networks.⁴⁴ The programs offered by the FHS in child care, such as immunization, attention to prevalent diseases, nutritional monitoring, prenatal care and family planning, should be priorities to prevent hospitalizations of children for pneumonia. Appropriate and qualified practices contribute to for the reduction of pathologies that are considered ACSC.⁴⁵

It is also necessary to highlight the changes that have taken place in the social determinants of health in Brazil and the implementation of government policies for child healthcare, with a significant and favorable impact on various indicators of child health in the country. Currently, Ordinance No. 1,130 is in effect, establishing the National Policy for Comprehensive Child Health Care and emerged with the objective of promoting and protecting child health in various segments, granteeing care up to age 9 through actions such as breastfeeding, comprehensive and integrated care during pregnancy, and special attention to early childhood and to the most vulnerable populations, with a view to reducing morbidity and mortality and also enabling an environment that facilitates life, with conditions worthy of existence.⁴⁶

In addition, the NIP had a significant effect in child healthcare with a view to introducing vaccines such as pneumococcal PCV 7 and PCV 10 in the immunization calendar in Brazil in 2002 and 2010, respectively. The program had a great importance in reducing pneumonia in children.⁴⁵ In the Pan American Health Organization (PAHO), an arm of the WHO, the Brazilian NIP is cited as a world reference. What has been achieved by Brazil, in immunizations, is far beyond what has been achieved by any other country of continental dimensions and of such great socioeconomic diversity.⁴⁷

The Ministry of Health of Brazil points out a potential for protection against invasive pneumococcal disease caused by the serotypes contained in the PCV 7 vaccine of 70%, as well as the inclusion of the PCV 10 vaccine as an important advance for Brazilian Public Health, because it protects children against invasive and noninvasive pneumococcal diseases.⁴⁸ The Ministry of Health of Brazil recommends an ideal rate of vaccination coverage of 95% capable of making the 10-valent pneumococcal vaccine efficient in reducing cases of PAC and other invasive pneumococcal diseases.^48,49

It is verified that in cities where the vaccination coverage was greater than 95%, there was a reduction in cases of pneumonia, in contrast to cities with less coverage, such as São Paulo with 75% and Porto Alegre with 85%, where such reduction was not observed.⁵⁰

A study assessed the impact of the implementation of PCV 10 in the state of Santa Catarina, southern region of Brazil, in children under 1 year, comparing pneumonia mortality in the period 2006–2009 in relation to the period 2010–2013. The study identified a reduction of 11% in mortality,⁵¹ and a similar result was observed in the study conducted by Silva et al.,⁵² who also observed a 19% reduction in cases of pneumonia in children under 1 year after the implementation of PCV 10.

Reduction in hospitalization rates for pneumonia may also be associated with the creation and implementation of the cash transfer program called Bolsa Família Program (PBF) in Brazil. Implemented in 2003, it is considered one of the largest programs that deal with conditional cash transfer in the world, emphasizing that in recent decades, this program has managed to achieve high coverage as a social and popular safety-net program.⁵³

The PBF contributed to reducing the level of poverty throughout the country, mainly in the northeast region, where a large portion of the population escaped extreme poverty.⁵⁴ Therefore, the PBF was important, especially in the less developed regions of Brazil, as an important strategy that may have contributed to the reduction of hospitalization rates for pneumonia.

Although there was a decreasing trend in hospital admissions for pneumonia in children, there was a clear disparity between regions, showing deep regional inequalities that result from the country’s historical inheritance as well as the country’s political and economic conformation.⁵⁵

Regarding spatial analysis, the present study identified a weak spatial autocorrelation, according to the global Moran value. The maps showed formations of high-high IRUA clusters and high RR of occurrence of high rates of hospitalization in the south, whereas the northeast and southeast presented clusters of low rates and RR values below 1. In addition to regional and economic differences, these data may be influenced by factors such as climate, air pollution, and access to private health services, among others.

Surveys indicate the continued concentration of medium and high complexity equipment in a small number of cities and the need for large displacements between macro-regions and states in the country for access to certain health services.^45,56 On the basis of this premise, it is possible that the low rates of hospitalization for pneumonia presented by the northeast region are due to the difficulty of accessing services of medium and high complexity because there is a need for large displacements, which hinders full access to all levels of healthcare. On the other hand, the higher rates observed in the south of Brazil may be related to the greater availability of hospitals of medium and high complexity and concentrations of specialized equipment, which provides a more qualified and more resolute service.⁵

A study conducted in England also found that the greatest variation in the risk of hospitalization for pneumonia in children aged 0 to 14 years was related to economic deprivation, lower scores in child well-being index, number of vulnerable children and great distances from hospitals and health services to the place of residence of the people.⁵⁷

Other points that may contribute to the lowest rates found in the northeast include the lack of information about signs of the disease among families, leaving clinical symptoms to go unnoticed by families, who do not take the child to health services for care. Another possibility is diagnostic errors that may confound the signs of pneumonia with those of other diseases, thus causing underreporting of cases.⁵

Atmospheric pollution and climate in the regions have also been associated with the risk of developing respiratory diseases. The continuous exposure to pollutants impairs the development of the lungs, which develop progressively until age 10 years; therefore, younger children are the most affected.⁵⁸ A study carried out in the city of Cuiabá-MT, in the midwest region, detected an association between exposure to pollutants and hospitalizations for pneumonia in children.⁵⁹ Other studies have observed the same association.^60,61 With regard to climatic conditions, studies by Natalli et al.⁶² and Santos et al.⁶³ identified a higher proportion of hospitalization for pneumonia in periods with the first cold fronts that cause sudden changes in temperatures, a fact that corroborates our study, in which one of the most affected regions was the south, which is considered the coldest in Brazil and also has one of the highest concentrations of industry, which may be related to the highest concentration of air pollutants.

Fires are considered another factor that can influence the increase in hospitalization rates for pneumonia. The midwest region has had the largest number of fires per year and the largest deforested area in the Legal Amazon since the early 1990s. The study by Nascimento et al.⁶⁴ detected more than 160,000 outbreaks of fires from 2008 to 2009 with an increase in hospitalization rates for pneumonia in children. Fires are more frequent in the dry months (May–October), corresponding to 90% of the occurrences.⁶⁵

The northern region had an increase in RR over the 5-year periods, along with an increase in hospitalization rates from 2016 onward. Although this increase was not significant, it is noteworthy because the northern region is unique for several reasons, such as having the highest concentration of indigenous people in relation to other regions, and there is evidence of an increase in cases of pneumonia in this population. The study by Caldart et al.⁶⁶ found that hospitalization rates for ACSC such as pneumonia were significantly higher among indigenous Yanomami children compared with nonindigenous children in the state of Roraima in the northern region of Brazil.

Also noteworthy is the impact that the H1N1 influenza pandemic in 2009 may have had on the increase in hospitalization rates for pneumonia in Brazil. This fact may be related to the increase observed in the temporal trend of the hospitalization rate from 2008 to 2009 in our findings. Hospitalization rates in countries in the Southern Hemisphere showed values from 23.6% to 30.6% for individuals with influenza A H1N1 in 2009.⁶⁷

With regard to socioeconomic factors and the provision of health services, our findings indicate a paradox because regions with the MHDI > 0.7 and low supply of services and professionals, such as the IRUA in the state of São Paulo, had the lowest rates of hospitalization for pneumonia. This paradox may be related to the greater use of private health care services.

Data obtained through the National Health Survey (NHS), a population-based household survey with the main objective of uncovering the determinants, conditions, and health needs of the Brazilian population to provide a representative database of the country, conducted in 2013, corroborate our results: people living in the south and southeast regions had greater access to health services compared with residents of other regions.⁶⁸ Viacava et al. sought to analyze the evolution of the supply of health facilities and resources in Brazil from 1988 to 2018 and identified that in 2013, the SUS paid 67% of the costs related to hospitalizations in Brazil.⁶⁹

Other research also found important differences in the provision of health services between regions in Brazil, with higher proportions of medical consultations in the south and southeast, which have the best living conditions and the highest HDI values.^70,71

Another important point to consider is the access to private health services because the data presented here only took into account hospitalizations in SUS, not those covered by health plans. In this context, the NHS⁶⁸ findings identified that there is a large coverage by health plans in the southeast and south regions of Brazil and low coverage in the northeast.

In view of the data presented, it appears that despite the reduction of pneumonia in children under 5 years of age, the disease is still a challenge in Brazil, due to both the magnitude of the problem and the unequal distribution between regions, which potentiate the vulnerability of certain populations. The findings presented here are important to guide health actions aimed at the reduction of morbidity in children, mainly due to preventable causes such as pneumonia, in all regions of the country through strategies that guarantee equal and comprehensive access to health services.

The limitations of the present study include the use of secondary data obtained from HIS/SUS, which, in addition to being subject to errors in recording and processing, take into account only hospitalizations paid by SUS. Furthermore, rehospitalizations are not distinguished. Also, there were few studies addressing this theme at the national and international levels.

The study identified a decreasing trend in the rates of hospitalization for pneumonia in children under 5 years of age in Brazil in the past 20 years; however, high rates are still seen in some regions. Disparities between regions were also observed.

The present findings may foster the creation of public health policy strategies to minimize inequalities between regions and thereby contribute to reducing pneumonia rates in all regions of the country, in addition to strengthening policies within PHC. Ecological studies addressing this theme are fundamental to subsidize these actions and contribute effectively to reform or creating public policies, especially aimed at child healthcare, as well as to reduce costs to SUS.