Public  policies reflected on hospitalizations for ambulatory care sensitive conditions: an ecological study in Sergipe-Brazil, 2008-2017


 Background: Hospitalization for Ambulatory Care Sensitive Conditions (HACSC) are an indicator of failure at this level of the health system. Since 2006, a series of actions have been taken in Brazil to promote Primary Health Care (PHC), a situation that has changed radically since 2014. To analyze the trend of the HACSC in Sergipe, in the period from 2008 to 2017. Methods: Ecological time-trend analysis. The data were extracted from the hospital admission authorizations on Hospital Information System in Brazilian Unified Health System. It was classified according to the Brazilian List of HACSC. The rates were standardized by the indirect method. Findings: There were 608,083 non-obstetric admissions and 125,497 HACSC (20.6%), with an overall rate of 5.8 admissions per thousand inhabitants (5.7 male and 5.8 female). The trend described a U-curve, decreasing from 2008 to 2011, with little oscillation from 2011 to 2014, rising from 2014 to 2017. The rate was considerably higher in men over the age of 70. Throughout the period, the most frequent causes were: gastroenteritis and complications (15.6%), kidney and urinary tract infection (9.4%), asthma (9.2%), heart failure (8.8%), cerebrovascular (8.1%), and diabetes mellitus (7.2%). Conclusion: The indicator improved in the initial period of the study, following the introduction of public policies that strengthened PHC, and worsened as they were withdrawn or neglected.

. ESF is a national strategy, with nancial incentives (until 2017) for its adoption by the municipality, which is voluntary and does not require complete populational coverage. Therefore, we have different coverage in the country and within municipalities.
Health management should be guided by quality information available to the manager and social control. Among the instruments for evaluating the effectiveness of PHC is the Hospitalizations for Ambulatory Care Sensitive Conditions (HACSC) indicator. Ambulatory Care Sensitive Conditions (ACSC) are health problems typically seen at the rst level of the health system and whose evolution will require hospitalization of the patient in the absence of effective and timely intervention. Examples are the prevention of the incidence of diseases, such as in the use of vaccines and syphilis treatment in pregnant women; the treatment of acute episodes, such as in the case of dehydration and bacterial pneumonia; and the control of prevalent chronic conditions such as diabetes mellitus and systemic arterial hypertension, thus avoiding or delaying hospitalizations for renal failure, peripheral arterial disease, diabetic foot, acute myocardial infarction or strokes5,6 .
The use of the indicator is based on the premise that hospitalization at the moment it occurs, is necessary for the patient. This need results from failures in the rst level of the health system5, 6 26,364 in 20068 . In Sergipe, the decrease of HACSC rates is in part due to the More Doctors Program (PMM), an emergency physician allocation program for PHC in hard-to-reach areas, supported in large part by the Cuban Government's collaboration by sending doctors9 . The Brazilian Northeast, where this study is performed, has the lowest Human Development and life expectancy rates, the worst infrastructure indicators (basic sanitation such as sewage and piped water), poor schooling, and the highest child mortality rates, compared to other Brazilian regions10 .
From 2006, a set of public policies were implemented to better organize the SUS, guided by a logic centered on universal, integral, longitudinal, and attention coordinative PHC. Since 2013, however, Brazil has been experiencing a political, legal, and economic instability scenario and it has discontinued social, cultural, and environmental policies, as measures to confront the economic crisis, the so-called "austericide"11 . In the political eld after the impeachment of President Dilma Rousseff in 2016, we culminated with the unbelievable Bolsonaro government12 . In public health, we saw the reduction of vaccination coverage by half, the increase in infant and child mortality from 2016, and maternal mortality in 2018, as well as the measles epidemic in 201913 . Therefore, we should expect that the political and institutional weaknesses, the cut in resources, and the consequent weakening of SUS management that led to the worsening of the aforementioned indicators, will also compromise the effectiveness of PHC.
No studies were found that evaluate the effectiveness of PHC in this period in Sergipe with this focus.
The objective of this study is to contribute lling this gap by analyzing the evolution of HACSC rates in Sergipe in the years 2008 to 2017, a period that encompasses investment in PHC, political instability, and nally the implementation of scal austerity policies.

Methods
The state of Sergipe, the smallest in territorial extension in Brazil, with 21,926,908 km2, is part of the Brazilian semiarid region, which has, among other characteristics, the daily percentage of water de cit equal to or greater than 60%. The Human Development Index was 0.665 in 2010. The population of Sergipe, in its 75 municipalities, in 2010 census had 2,068,017 inhabitants, of which 51.4% were women and 73.5% lived in an urban area, according to Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geogra a e Estatística -IBGE) in 2010. PACS and PSF in Sergipe were created in 1998 and ESF began in 2007, a year in which 127 family health teams were established, covering 87% of the population13 , and remained at these proportions until the end of the study in 2017.
An ecological study was conducted using the Databases from the SUS Hospital Information System (BD-HIS/SUS). The so called "reduced les" of Hospitalization Admission Authorization (AIH) were used, from which the main diagnosis of the hospitalization, date of admission, gender, age, and city of residence were obtained. These "AIH les" are made available in separate les by State and month of admission billing (called "month of competence"), which is the same month of hospitalization in most cases. But the AIH can be billed later, which can cause both inclusion of cases occurring a few months before the period of interest and loss of cases in the last months of the selected period that were billed later. To avoid these problems, the data les of the State of Sergipe were read from January 2008 to July 2018, afterward, only those cases whose hospitalization occurred from 01/01/2008 to 31/12/2017 were selected. Only cases of citizens residing in Sergipe were also selected. Finally, hospitalizations for obstetric procedures to treat non-morbid conditions such as childbirth and abortion were excluded. Hospitalizations were classi ed as ACSC according to the Brazilian List of Ambulatory Care Sensitive Conditions, composed of a wide range of 19 groups of causes.
The crude rates of HACSC per year were calculated, then standardized by sex and age by the indirect method (calculation of the Standardized Hospitalization Ratio -SHR), taking 2014 as the reference year, since it is the lowest crude rate. Thus, the meaning of every presented SHR is how much the rate that year was higher than in 2014, adjusted for gender and age -leaving 1 to the value presented and multiplying by 100, obtaining the value in proportion, i.e., an SHR = 1.241 means a HACSC rate 24.1% higher than that observed in 2014, the lowest rate of the period.
The standardized rates trend in the period was analyzed by simple linear regression, despite a positive autocorrelation (Durbin-Watson test = 0.6; p < 0.001). Given the observed distribution, two explanatory models of the distribution of SHR as a function of time were tested: one considering the year plus a quadratic term of the year (SHR = intercept + year + year²) and another dividing the period into three (2008-2010, 2011-2014, 2015-2017), analyzed as a categorical variable. Considering the few observations and the percentile distribution of the residues, the models were considered adequate. Data capture on the internet, reading of les, classi cation of hospitalizations, analysis, and graphical presentation were performed using the R statistical software. For data capture and reading, the microdatasus14 package was used and the classi cation was made by the csapAIH15 package. Compressed les (DATASUS .dbc format) are read by the read.dbc16 package. Since AIH les are publicly accessible without locks (with access restrictions from outside Brazil since 2017), the process we have performed can be reproduced in R by running the command script le, which may be requested from authors.

Results
During this period of ten years analyzed, there were 608,353 hospitalizations, excluding obstetric hospitalizations, well distributed between genders (49% males, 51% females). In the period, there were 125,497 cases of HACSC, representing a global rate of 5.8 HACSC per thousand inhabitants ( Table 1).
The rate distribution shows a curve in U, falling until 2011, stabilizing with some oscillation until 2014, when it reaches the lowest rate, and then rising consistently until the end of the period of study, with a The crude rate of HACSC in the period was 5.7 cases per thousand inhabitants in males and 5.8 in females, although the standard rate by age group is slightly higher in men. The proportion of hospitalizations is similar between genders: in males, 20% of hospitalizations were due to ACSC, and 21% in females. The women presented ages from 0 to 109 years, mean of 40.4, a standard deviation of 30.0, and a median of 42 years. Men were aged from 0 to 117 years, mean 39.4, standard deviation 30.8, and median 44 years. Just under a quarter (23%) of hospitalizations for ACSC are under ve years old, and 15% of children under two years old. The median age was 42 years and the 75th percentile was 67 years old. The median and mean age is higher in 2017, while decreasing the coe cient of variation, re ecting population aging. From 70 years of age, the rate is considerably higher in men (Figure 2). The elderly are the age group that most bene ted between 2008 and 2014, but in 2017 they already had rates close to those at the beginning of the period. Children had little improvement between 2008 and 2014 and the rates in 2017 were higher than at the beginning of the period (Figure 3).

Discussion
The ESF has been proposed as a priority model for organizing PHC in several aspects, especially in improving accessibility and reducing inequities in less empowered groups of the population17,18 . Several studies have found an inverse association between HACSC and ESF coverage19,20 . However, in the period studied, ESF coverage in Sergipe remained above 80%, with little variation, suggesting that in this period the effectiveness of PHC can be better assessed by other aspects than ESF coverage. In other words, the changes observed in the period are not due to a change in the supply of the service in terms of population coverage.
In the initial phase of the study, from 2008 to 2014, we observed a decrease in HCASC in Sergipe, which a lower value. From then on, the indicator worsened, with a sharp increase in rates, which, at the end of the period, are similar to the beginning, showing a worsening of the advances achieved. The worsening occurs in all age groups, particularly among children, who have the HCASC rates in 2017 higher than in 2008. Statistical analysis showed that the values found at the end of the period are signi cantly different from the beginning, indicating that the behavior observed is not the result of random variation. The analysis also shows the aptness of the indicator to describe the outcome of health policies, which is consistent with the conceptual methodological framework that presents it with an indicator of the rst level of the health system in all its scope, and not only of the care provided by the health team5,6 .
The public policies that preceded and contributed to the performance of the indicator throughout the study period began with the publication of the 2006 PNAB, which marks the transition from PSF to ESF status. Integrative and Complementary Health Practices (also called "Alternatives") and the National Health Promotion Policy were also important milestones that year. In 2007, the Health at School Programme was created, and in 2008 the Family Health Support Centre (Núcleo de Apoio à Saúde da This entire set of federal government policies has invested in improving the structure and health process of care, and expanding the accessibility, resulting in better effectiveness of PHC. However, as of 2014, the Brazilian economic crisis had repercussions in the rate of unemployment and increased social inequality, nancial cuts, and social instability24 , and the indicator began to assume the upward slope of the curve. In this regard, the approval of the Constitutional Amendment 95/2016 that establishes a spending cap in the next 20 consecutive years, along with environmental and educational policies will compromise the sustainability of the SUS and its constitutional premise of universal coverage and other attributes of the PHC24,25 . In 2017, there was the third edition of the PNAB, with major setbacks and reformulation in its political orientation, and withdrawal of nancial incentives to the ESF, which is why it suffered severe criticism from the Brazilian scienti c community25,26 .
The most frequent groups of causes in children up to 14 years old were gastroenteritis, asthma, bacterial pneumonia, and ear, nose, and throat infection. There was a signi cant decrease in gastroenteritis and complications, ranging from 50% of the HCASC in 2008 to 22% in 2017; asthma and pneumonia remained stable in the period. These causes were the same as those found in a study in Minas Gerais, Southeastern Brazil, for children and teenagers27 .
A study conducted in Sergipe from 2002 to 2012 on hospital admissions in teenagers showed a 143.1% reduction in HCASC28 . In the population of 50 or more years, the most frequent causes were heart failure, cerebrovascular diseases, diabetes mellitus, systemic arterial hypertension, nutritional de ciencies, and angina. A study conducted in Paraná, Southern Brazil, in the range of 60-74 years, presented similar results20 . The improvement observed in the indicator before the most pronounced political changes from 2014 onwards was mainly due to more effective attention to the population over 50 years of age. It is notable that even among the very elderly, rates improved a lot, and then worsened. These results show that the exclusion of the elderly in studies that address the HCASC can be a mistake, since it was shown that it responds to the political actions driving the PHC. Thus, we should continue analyzing the indicator with all age groups -consistent with the principles of universality, equity, and integrality of PHC.
This study has its limitations, most dependent on the quality of information, as it uses secondary data, and the fact that SUS, despite being an universal system, has not total coverage of the hospital admissions. So, it is important to emphasize that about 88% of hospital admissions in Brazil are through SUS29 . Nevertheless, the SIH-SUS has shown high reliability for the study of HACSC30.

Conclusion
The investment in PHC from 2006 was followed by the improvement of its effectiveness, measured by the HACSC rates. The changes in the country's political direction from 2014 onwards were accompanied by the worsening of the effectiveness of PHC in the state of Sergipe, in both genders and all age groups, returning to the levels observed at the beginning of the decade. The improvement in the indicator following the implementation of policies to strengthen PHC and its worsening in response to the country's political and economic crisis demonstrates the damage to Sergipe people's health caused by policies to withdraw social rights and diminish the state, concerning the most common health problems. This study reports us to the need for a national study to evaluate the effects of national political changes on the whole Brazil, as well as its Regions and States.