Admissions due to ambulatory care-sensitive conditions (ACSC): an analysis based on socio-demographic characteristics, Brazil and regions, 2010 to 2019

: Objective : To analyze the temporal trend of hospital admissions due to Ambulatory Care-Sensitive Conditions (ACSC) in Brazil per sex, region, cause and age group, from 2010 to 2019. Methods : This is an ecological study based on the temporal trend of ACSC rates. Standardized rates were analyzed in a simple linear regression and a generalized linear model (GLM) Gamma. The percentage change was also checked over three periods: 2010–2014, 2015–2019 and 2010–2019. Results : There was a reduction in hospital admission rates between 2010 and 2019 for Brazil: 124.3/10,000 inhab. to 88.2/10,000 inhab. among women (-29.0%) and 119.0/10,000 inhab. to 88.2/10,000 inhab. (-25.9%) among men in all regions. The decrease was more prominent between 2010–2014 (-17.7%) and (-17.8%) than between 2015–2019 (-.2%) and (-5.9%) for females and males, respectively. The milder decline in the period from 2015 to 2019 was more noticeable among age groups 0–4y and 5–19y in all regions. As for the causes, most ICD-10 diseases had a downward trend, the more expressive ones being gastroenteritis (-60%); however, an increase was seen in cerebrovascular diseases among women (11.2%) and men (17.1%), and angina (15% %) and skin infections (56.1%) among men. Conclusion : There was a significant drop in ACSC rates in the period analyzed, especially for age groups 0-4y and 5-19y. The rates had a milder drop from 2015 to 2019, a period of austerity and economic crisis.


INTRODUCTION
Primary Health Care (PHC) is the core gateway to the Unified Health System (SUS) in Brazil 1 . The political decision of PHC was the result of the historical process of consolidation of SUS and of the understanding that its strengthening improves the performance of indicators, and reduces both health inequalities and costs 2 .
PHC has expanded greatly throughout the national territory since the creation of the Family Health Strategy (FHS) in 1994. Data from the National Household Sample Survey (PNAD) estimated a PHC populational coverage of 50.9% in 2008; data from the 2013 National Health Survey (PNS) estimated the coverage in 53.4% and from 2019, in 60% 3,4 . In addition to the expansion of PHC, it is necessary to assess its quality and ability to produce changes. Assessing health services is a major challenge given the diversity of methodologies available, databases and the importance of considering the perspective of different actors involved in the process: workers, managers and users 5 .
Hospital admissions for Ambulatory Care-Sensitive Conditions (ACSC) have been used internationally as an outcome indicator to evaluate PHC. In Brazil, with the creation of the national list of ACSCs in 2008 6 , this indicator started to be used to assess hospital admissions and their association with the FHS team.
ACSCs refer to hospital admissions for diseases that would be impacted by adequate care 7,8 . Different countries have listed these diseases in an attempt to develop a measurable indicator. These lists differ from country to country, which makes it hard to compare them; however, they have been widely used to assess PHC and health systems as to effectiveness in impacting hospital admissions 9,10 .  (-9,2% e -5,9%) -valores para sexo feminino e masculino, respectivamente. Esse declínio menor no período de 2015 a 2019 foi mais perceptível entre as faixas de zero a quatro anos e de cinco a 19 anos em todas as regiões. Quanto às causas, foi verificada redução para maioria dos códigos da Classificação Internacional de Doenças (CID-10), mais expressiva para gastroenterites (-60%), entretanto tiveram aumento doenças cerebrovasculares entre mulheres (11,2%)  ACSC rates in Brazil and regions tended to decline between 1999 and 2007, along with the expansion of FHS coverage across the country 11 . Several analyses of this evolution have been carried out to better understand what can impact these rates in general, looking a correlation with such hospital admission rates [12][13][14] .
ACSCs have been used as an indicator of the effectiveness of health systems 15,16 , so it is important to analyze the behavior and evolution of these rates in more recent yearsincluding periods of economic crisis 17 -as to monitor health inequalities and any possible worsening in indicators.
Further national studies stratified by causes and age group are also important to understand how ACSCs can be better used 12 .
This study aimed to analyze the temporal trend of ACSCs in Brazil according to sex, regions, causes, and age groups from 2010 to 2019.

METHODS
This is a temporal trend ecological study, in which the rates of hospital admission for ACSC in Brazil and regions from 2010 to 2019 were analyzed.
The data come from the Hospital Information System (SIH) available at the Information Technology Department of the Unified Health System (DATASUS, www.datasus.gov.br). The selection of groups of the International Classification of Diseases in its tenth chapter (ICD-10) was made based on the national list of ACSCs published by the Ministry of Health in 2008 6 , composed of 20 diagnostic groups divided into 120 ICD-10 groups, except the ones related to childbirth (ICD-10 O-80 to O-84), as these are the natural outcome of pregnancy, not a disease. People aged 80 years and older were also excluded because they had a higher risk of being admitted to a hospital regardless of the quality of the health service, as recommended by the ACSC-Brasil Project 11 .
The rates of hospital admissions for ACSC were calculated by dividing the number of admissions for diseases of the groups described in the Brazilian List of ACSCs 6 by the estimated population, according to the Brazilian Institute of Geography and Statistics (IBGE) in respective years; then, the result was multiplied by 10,000. This procedure was performed by sex, region, and age groups. Rates were also standardized per age distribution of the Brazilian population at the period midpoint (2014), enabling comparisons between years 11 . Rates by cause were also calculated by the number of admissions divided by the estimated population; however, standardization and age distribution were not performed.
The rates of admissions for ACSC were considered a response variable and the yearto-year distribution was considered an explanatory variable. Stratified analyses were performed by region, sex, age group, and causes of hospitalization. To verify the distribution of these rates, the Shapiro-Wilk test was applied. In the analysis groups with normal distribution over the years, a simple linear regression model was applied 18 . In the others, it was assumed that they had a gamma distribution and the Generalized Linear Model (GLM) for the Gamma distribution was applied 19 .

REV BRAS EPIDEMIOL 2022; 25: E220012
The regression coefficient β, which is the degree of association between test variables, was presented both in the linear regression and in the GLM gamma analyses: positive β means a correlation of upward rates over the years; negative β indicates a downward trend. The significance of this association (p<0.05) and the extent to which the variation in rates of hospital admissions for ACSC can be explained by the evolution of the years (R²)-that is, the accuracy of the model-were also shown. Descriptive analyzes were also carried out to verify the percentage changes in rates between the beginning (2010) and the end (2019) of the series, as well as to compare the differences in rates between two different periods: 2010 and 2014, 2015 and 2019.
Data from SIH/DATASUS were tabulated and extracted by means of the Tabwin software version 3.6. The database was exported for statistical analysis in the Statistical Package for the Social Sciences (SPSS) software version 20.0.
The study was carried out in accordance with Resolution No. 466 of the National Health Council (CNS) dated back from December 12, 2012. The SIH information is available on the DATASUS website without patients' identification, therefore, they are freely accessible.

RESULTS
In Brazil and in all regions, both the standardized rates and rates by age group and sex of hospital admission for ACSCs had a statistically significant reduction in the period studied (Table 1).
The biggest drop for both sexes was identified in the Midwest region: -40.87% among women and -35.22% among men. The smallest rate variations in the period were found in the Southeast region, with -22.37% among women and -21.7% among men, followed by the South region, with -26.54% among women and -24.29% among men. In all situations, the difference between 2010 and 2014 was greater than between 2015 and 2019, falling to half the reduction among women and a third among men (Table 1).
The highest standardized rates in 2010 were identified in the North region, with 178.97 hospital admissions per 10,000 inhabitants (178.97/10,000 inhab.) among females and 147.59/10,000 inhab. among males. In 2019, the highest rate for females was reported in the North region (117.70/10,000 inhab.) and, for males, in the South region (102.55/10,000 inhab.) ( Table 1).
The rates of hospital admissions for ACSC by age group and regions for males, observed in Table 2, decreased from 2010 to 2019. In the age groups 0-4 years, 5-19 years and 20-59 years, the rhythm of rate decline was slower when the period 2010-2014 was compared with the period 2015-2019. In some cases, there was an increase from 2015 to 2019, such as in the Midwest region, in the age groups 0-4 years and 5-19 years, and in the Southeast region, in the age group of 0-4 years ( Table 2 and Supplementary material -Table 1S).

REV BRAS EPIDEMIOL 2022; 25: E220012
The biggest drop in rates was reported in the North region for the age groups 0-4 years (-34.34%) and 5-19 years (-37.33%), and in the Midwest region for the age groups 20-59 year (-37.29%) and 60-79 years (-39.08%). In the age group 0-4 years, however, this decrease was noticed only in the first period (2010 to 2014), with little variation in the second period, and with an increase by 9.57% in the Midwest region between 2015 and 2019 (Table 2 and  Supplementary material -Table 1S).
The highest hospital admission rates in the age group 0-4 years were found in the North region in 2010 (465.69/10,000 inhab.) and in 2019 (305.79/10,000 inhab.). In the age group 5-19 years, the highest hospitalization rates were reported in the North region in 2010 (73.02/10,000 inhab.) and in the Northeast region in 2019 (48.53/10,000 inhab.). The rhythm of decline in rates also decelerated in 2015 to 2019 and, even though it grew by 3.0% in the Midwest region (Table 2 and Supplementary material - Table 1S).
In the age group 20-59 years, the highest rates were found in the North region in 2010 (88.14/10,000 inhab.) and in the South region in 2019 (64/10,000 inhab.), also with a reduction in the percentage of decline from 2015 to 2019. For the age group 60-79 years, the highest rates were seen in the South region, with approximately 560.8/10,000 inhab. in 2010 and almost 409.2/10,000 inhab. in 2019, with a reported reduction in the downward trend from 2015 to 2019. The correlation of hospital admissions with the years showed an   Table 1S).
In Table 2, the rates of hospital admission for ACSC by age group and regions among females also decreased from 2010 to 2014. Between 2015 and 2019, the drop in rates was less marked, while the increase by 6.4% in hospitalization rates in the age group 0-4 years was identified in the Midwest region (Table 3 and Supplementary material - Table 2S).
The highest hospitalization rates were found in extreme age groups: 0-4 and 60-79 years. Among children in the first age range, the highest rates were observed in the North region in 2010 (433.8/10,000 inhab.) and in 2019 (279.8/10,000 inhab.). As for the elderly of the second age range, the highest rates were found in the Midwest in 2010 (488.9/10,000 inhab.) and in the South in 2019 (303.7/10,000 inhab.). The correlation of hospitalization rates with the years showed an association and decrease also among females. The β coefficients ranged from -0.03 to -25.57 with p-value <0.001 in all years (Table 3). R² suggests that the variable hospitalization rate can be explained over the years (0, 63 to 0.96) (Supplementary material - Table 2S).

DISCUSSION
The study of temporal trends in hospital admissions from 2010 to 2019, using the ACSCs list, found a reduction in hospitalization rates in Brazil and across regions; however, the drop reported from 2010 to 2014 was greater than the one reported from 2015 to 2019. Among children aged zero to four years and five to 19 years old there was an increase or reduction in rhythm of decrease in rates from 2015 to 2019. Most causes showed a downward trend, while cerebrovascular diseases increased among women and men, as well as angina and skin infections among men.
The decrease observed in this study is consistent with the reduction reported in studies carried out by the ACSC-Brasil Project from 1999 to 2007 in Brazil and regions 11 . A literature review with the theme of ACSC among children under five years of age found a reduction 13 . When data were analyzed by age group, higher rates were found in the elderly aged 60 to 79 years and in children aged zero to four years of both sexes, which can be explained by the greater vulnerability in the extremes of life, which also have higher mortality rates 20 . For both sexes, the greatest reduction in the age group 0-4 years was found in the North region and, in the age group 60-79 years old, in the Midwest region. In a study carried out by Amorim et al. in Brazil and regions, in which the period from 2003 to 2012 was analyzed, better results were also found for the Midwest region, especially in the age range 60 years and over. One hypothesis would be the great economic development of this region of the country in the period, driven by agribusiness, which had a positive impact on indicators in several areas, including the health of the population 21 .
The proportion of drops was similar in all age groups and in the standardized rates for the entire population, varying from 20 to 40% in the period analyzed, as also reported by studies on previous periods, carried out in several cities and states 15,22,23 . This decline may reflect the continued expansion of PHC in the period, the investments in improvement programs and in PHC, namely the Program for Improving Access and Quality of Primary Care (PMAQ-AB) 24,25 , the National Policy of Primary Care of 2012 26 and the program "more physicians" 27 . For the age groups 0-4 and 5-19 years, this drop occurred mostly from 2010 to 2014. Between 2015 and 2019, there was a decrease in the rhythm of decline, which has also been described in studies on infant mortality in Brazil, which increased after 2015 28,29 .
Some authors have related the increase in infant mortality between 2015 and 2016 with the Zika virus 28 , others with austerity policies such as Constitutional Amendment 95 (EC95), which froze the increase in public spendings for 20 years 29 . The findings in the age group from zero to four years old are consistent with research that even project an increase in mortality in children under five years of age by 2030 in Brazil if the crisis and austerity policies are maintained 30 .
The relation between crisis, reduced spending on social policies, and the consequent worsening of health indicators has been reported in studies from European countries, especially countries with less robust health systems and higher proportions of vulnerable populations 17 .

REV BRAS EPIDEMIOL 2022; 25: E220012
Another finding of the study was that, in general, higher rates of hospitalization were found in the North region among females. In the Southeast region, lower rates were found for all age groups, in all years. Among males, the highest rates were also found in the North region, except for the 60-79 age group, whose higher rates were identified in the South region. The lowest rates, for all age groups, were also found in the Southeast region. Regional differences have also been described in some editions of PNS, showing worse indicators in the northern region of the country 3 . This can be explained by the lower socioeconomic development in this region and the care gaps, with a lower supply of health services 31 .
The main causes of hospitalization found in Brazil among females were infectious gastroenteritis and complications, kidney and urinary tract infections, heart failure, cerebrovascular diseases, and asthma; among males, data were similar, but less expressive when it came to urinary tract and skin infections. Similar results were obtained in a study carried out in Pernambuco between 2008 and 2012 23 , with significant drops in hospitalizations for infectious gastroenteritis (60%). This was also described by Macinko et al., who observed the most marked declines in rates of hospital admissions for gastroenteritis and asthma in the age groups up to 20 years old from 1999 to 2007 in Brazil 22 . These results dialogue with those of studies carried out in the North region, in a temporal trend from 2008 to 2017 32 , and others conducted with children under one year of age, showing a reduction in hospital admissions for gastroenteritis 33 . These significant drops can be explained by general improvements in living conditions, better sanitation, improvements in water supply, and in access to health services, especially PHC 32 .
Pinto Junior et al., in 2020 33 , pointed out that the ACSC list has room for some adjustments considering different types of disease by age, such as the Brazilian List of Preventable Causes of Deaths from Interventions of the Unified Health System. This list considers a group of causes for people aged zero to four years and other ICD-10 causes specific for the age group 5-74 years, but not the same ICD-10 in all age groups.
The Unified Health System Performance Index (IDSUS), created in 2012, uses 24 indicators to provide an overview of SUS coverage in the national territory. Hospitalization for causes sensitive to primary care, an IDSUS indicator, do not take into account in their calculation hospitalizations for cerebrovascular diseases, diseases related to intestinal ulcers and prenatal care/delivery, as these are conditions in which PHC would not have an important impact on hospital admission, since they would depend on responses at other levels of the system 31 . In our study, significant increases in hospitalizations were observed from 2010 to 2019 for these three causes in both sexes in Brazil. We highlight the importance of further studies on causes defined as preventable in the ACSC list.
The limitations of this study were related to reliability and accuracy of diagnoses recorded in hospital admission authorizations (HAA) due to inadequate filling. Another limitation is data scope, which was restricted to SUS and partnered network. Caution is recommended when making comparisons between rates stratified by causes, as they have not been standardized. Even so, it is noteworthy that HAAs allow important epidemiological analyses on hospital morbidity 11 and, despite potential information biases, the study brings advances in identifying the evolution of rates of hospital admissions due to ACSC by age group and causes.

REFERENCES
This study described a significant reduction in hospitalization rates for ACSC in the period analyzed for both sexes, all age groups and regions. Among causes, gastroenteritis showed the most prominent decline. These findings reflect improvements in PHC, but also in general living conditions. However, a reduction in the rhythm of decline in rates from 2015 to 2019 was identified and, in some regions, even an increase, especially in the age group of 0-4 years. These findings may echo the economic crisis after 2015, as well as the austerity policies implemented in that period. PHC has been experiencing setbacks since 2016, including the dismantling of PMAQ-AB, cuts in specific fundings for the Family Health Support Center (NASF), and the weakening of the "more physicians" Program, added to the resource freezing resulting from EC95, which increases health inequities.
Further analyses of health outcomes and indicators in the coming years are needed so we can understand how these policies will behave. Knowing the trends in hospitalization rates can guide public policies aimed for investments in PHC, which is essential for improving the quality of life and health of the country population.