Abortion in Brazil: what do the official data say? Aborto no Brasil: o que dizem os dados oficiais? Aborto en Brasil: ¿qué dicen los datos oficiales?

According to the World Health Organization, from 2010 to 2014, there were around 55 million abortions worldwide, 45% of which were unsafe. In Brazil, data on abortion and its complications are incomplete. Health care data are only available for the public sector and mortality data depend on investigations of deaths. This study sought to describe the situation of abortion in the country using public data available in the different Information Systems – SIM (mortality), SINASC (live births) and SIH (hospitalization). From 2008 to 2015, there were around 200,000 hospitalizations/year for procedures related to abortion, 1,600 of which for medical and legal reasons. From 2006 to 2015, we found 770 maternal deaths in SIM whose underlying cause was abortion. There was a discreet reduction in the number of deaths from abortion in the period, with regional variation. This number could be increased by around 29% per year if deaths with mentions of abortion and declared with a different underlying cause were considered. Among the deaths reported as resulting from abortion, 1% were abortions due to medical and legal reasons and 56.5% were non-specified abortions. The proportion of deaths from abortion identified in SIH, in relation to the total number of deaths from abortion identified in SIM, varied between 47.4% in 2008 and 72.2% in 2015. Although official health data do not allow us to estimate the number of abortions in Brazil, we were able to establish the profile of women at higher risk for death from abortion: black and indigenous women, with low educational levels, under 14 and over 40 years of age, living in the North Northeast and Central regions, without a partner.

This article seeks to describe the situation of abortion in Brazil and regions, in the periods from 2006 to 2015, based on official data from the health information systems.

Methods
This is a descriptive, time-series, population-based study that sought to describe the scenario of abortion in Brazil using public data available for download in the country's different Health Information Systems. For death information, we used SIM data from 2006 to 2015. For births, we used data from the Brazilian Information System on Live Birth (SINASC, in Portuguese), also from 2006 to 2015. We obtained the SIM and SINASC databases by downloading them from the SUS Brazilian Health Informatics Department (DATASUS) website, available at http://datasus.saude.gov.br/informacoesde-saude/servicos2/transferencia-de-arquivos.
We also obtained hospitalization data from SIH using the Tabnet tool, also from the DATASUS website, available at http://www2.datasus.gov.br/DATASUS/index.php?area=0202&id=11633 and http://www2.datasus.gov.br/DATASUS/index.php?area=0203&id=6926. We included data from 2008 to 2015, since tabulation by hospitalization year is only available through Tabnet from 2008 onward. For previous years, tabulation is only available for processing year, which may introduce distortions due to the possibility of processing hospitalizations from previous years.
In order to identify and analyze deaths from abortion, we considered the Underlying Cause of Death, as established in the International Classification of Diseases (ICD -10th review) 17 . We used the classification suggested by Guia de Vigilância Epidemiológica do Óbito Materno 11 to calculate the abortion-specific Maternal Mortality Ratio (MMR), using the following ICD-10 categories: O03 (miscarriage), O04 (abortion due to medical and legal reasons), O05 (other types of abortion), O06 (non-specified abortion) and O07 (failed abortion attempt). The calculation of the abortion-specific MRR was based on the woman's place of residence. We did not use correction factors because there is no specific standardized factor proposed for deaths from abortion. The ratio denominator was the number of live births obtained from SINASC.
We analyzed the SIH data on hospitalizations from abortion according to admission diagnosis and according to the procedure carried out. The analysis according to diagnosis included the groups "miscarriage", "abortion due to medical reasons" and "other pregnancies that resulted in abortion". The following procedures from the SUS Table of Procedures, Medications and OPM were included in the analysis: "post-abortion/puerperal curettage" (0411020013) and "post-abortion uterine evacuation via manual vacuum aspiration (MVA)" (0409060070). For the comparative analysis of the number of hospitalizations due to abortion among the different states, we used the relationship between the number of hospitalizations from abortion in SIH and the number of hospitalizations for deliveries in SIH, with the goal of reducing the influence of fertility variations in the population and the use of supplementary health services. For the analysis of hospitalizations for procedures related to delivery, we considered all procedures classified in Subgroup 10 (delivery and birth) from Group 3 (clinical procedures), except "newborn care at the time of birth" (031001002-0). We further considered the procedures classified in Subgroup 11 (obstetric surgery) from Group 4 (surgical procedures), organization Form 01 (delivery).
We assessed data on mortality from abortion and on hospitalizations for diagnoses or procedures related to abortion according to the following variables (when available): woman's place of residence, sociodemographic characteristics (age, educational level, marital status, race/color), abortion classification (legal/miscarriage/other types/non-specified), year of occurrence and costs of hospitalizations from abortion. We analyzed the variable marital status for the period from 2011 to 2015 because of changes in the DC and Live Birth Certificate forms, with the inclusion of the field stable union.
This study was not submitted to an Ethics Review Board because it used public Ministry of Health data. Results are shown in aggregate, with no possibility of identification, in accordance with Resolution 466/2012 of the Brazilian National Health Council.

Results
The SIM analysis showed that, between 2006 and 2015, 770 deaths with abortion as the underlying cause were recorded in Brazil. Despite the variation over the years due to the low magnitude of the absolute numbers, we observed a trend of reduction among the deaths from abortion in Brazil, with regional differences. The regions North, Southeast and Central had, at different times, the highest abortion-specific MMR over the period and the South Region had the lowest abortion-specific MMR in the whole period. The Central Region is the only Brazilian region that shows a clear rising trend in mortality from abortion based on data recorded in SIM. Abortion-specific MMRs for Brazil and each macro-region are presented in Figure 1.
There was a reduction in the proportion of deaths from abortion among maternal deaths recorded in SIM from 2006 to 2015. Deaths from abortion were 5.7% of maternal deaths in Brazil in 2006 and 4.1% in 2015, and were only 3.2% in 2014. The Southeast region had the highest reduction in the proportion and the Central Region saw an important increase, going from 3.9% in 2006 to 6.1% in 2015.
Non-specified abortion (category O06 of ICD-10) was the most frequent underlying cause among deaths from abortion in the period we analyzed, with an average of 56.5% of cases. Among the 770 deaths whose declared underlying cause was abortion, only 7 (0.9%) were from abortions performed for medical and legal reasons (O04), 115 (14.9%) were declared as miscarriages (O03), 117 (15.2%) as other types of abortion (O05) and 96 (12.5%) as failed abortion attempts (O07).
Between 2006 and 2015, deaths from abortion were most frequent in the 20-29 years aged group. Only in the year 2007 did the age group 30-39 years register a higher number of deaths. All regions also had a greater volume of deaths in the 20-29 year group in the set of analyzed data. The abortionspecific MMR was higher among women over 40 years of age and, in 2010 and 2014, among adolescents aged 10-14 years. In the Northeast and Southeast regions, which concentrate the highest number of deaths from abortion in SIM, we observed the highest abortion-specific MMR at both extremes of the age groups. In the other regions, despite the small proportion of deaths with no information (0.004% of births and no deaths), the small number of deaths from abortion leads to great variation to the indicator, which makes the analysis difficult.
The distribution of the total number of deaths from abortion between 2006 and 2015 according to skin color maintained a relatively stable distribution, with approximately half of deaths among brown women. The abortion-specific MMR according to skin color shows that, in Brazil, according to data recorded in SIM and SINASC, black women had the highest values from 2006 to 2012. In 2013 and 2014, indigenous women had the highest MMR. In 2015, with only one death from abortion and with 0.36% of the country's births, Asian women had the highest MMR in Brazil. In the 10-year average, black women maintained the highest MMR, however, there was an important reduction in this group's MMR in the past five years ( Figure 2). Only in the Southeast region do black women still have the highest MMR. This variable also has a low proportion of missing data (around 5%) regarding deaths from abortion and births in the period.
The proportions of deaths from abortion according to marital status remained stable from 2011 to 2015. The analysis of the average of the years, considering the categories "living in a union" (married and stable union) and "not living in a union" (single, separated, widow) shows that proportions, excluding records with no information (average of 8.2%), have remained at approximately 30% and 70%, respectively. The higher proportion of deaths from abortion among women not in a conjugal relationship is present in all Brazilian regions, varying in magnitude. This proportion is of 60.6% in the North, 63.5% in the Northeast, 73.8% in the Southeast, 90.5% in the South and 66.8% in the Central.
With regard to educational levels, we found a relative improvement in the quality of the data on deaths from abortion, with a 49% reduction, approximately, in the percentage of missing data, however, 16.7% were still missing this information in 2015. The lower educational categories ("no schooling" and "1 to 3 years of schooling") had a percentage and absolute increase in the number of of abortions. The remaining categories had a reduction in the number of deaths. The comparative analysis of abortion-specific MMR according to educational groups is hindered by the small number of deaths in each category and the percentage of missing information for this variable.
The analysis of the percentage of deaths of woman of reproductive age that were investigated reveals differences between the Brazilian regions. In 2015, this percentage varied between 92.5% in Cad. Saúde Pública 2020; 36 Sup 1:e00188718  the Central and 86.7% in the North. Despite the large number of investigated deaths, the percentage of deaths of woman of reproductive age, excluding maternal deaths, with no information on the moment of death with regard to the pregnancy-puerperal cycle, varied from 15.4% in the South to around 36% in the North in 2015. In Brazil, this percentage varied from 54% to 29.2% between 2006 and 2015.
The cause of death considered in mortality analyses is the underlying cause of death. There are, however, other diagnoses listed in death certificates, not selected as underlying causes, but that are associated with the cause of death. The multiple causes analysis shows that, between 2006 and 2015, in addition to the already-mentioned 770 deaths, there were 220 deaths that mentioned abortion, however, with a different underlying cause. Of these, 58.2% had an underlying cause classified in Chapter XV of ICD-10, that is, they were maternal deaths declared for other causes, but with mentions of abortion. The most frequent underlying causes from Chapter XV among the 220 deaths with mention of Cad. Saúde Pública 2020; 36 Sup 1:e00188718  abortion are those classified in category O99 of ICD-10 (Other maternal diseases, classified in another part, but that complicate pregnancy, delivery and puerperium), followed by O85 (Puerperal infection), O23 (Genitourinary tract infection in pregnancy) and O96 ("death, by any obstetric cause, occurring more than 42 days, but less than one year, after delivery").
Further, among the 220 deaths that mention abortion, there are deaths whose underlying cause was classified in 13 other chapters of ICD-10. On average, in the years we have analyzed, in addition to those classified with an underlying cause from the abortion-specific ICD-10 categories, there were a further 28.8% of deaths with mentions of abortion. The proportion of the mention of abortion was different in each region, with considerable variation over the years. The average annual percentage was of 41.5% in the North, 34% in the Northeast, 26.9% in the Southeast, 51.6% in the South and 25.7% in the Central. Figure 3 shows the total of deaths with abortion as the underlying cause and the total of deaths from other underlying causes with mentions of abortion.
SIH recorded an average of approximately 200,000 hospitalizations/year for procedures related to abortion from 2008 to 2015. These hospitalizations represented a total value of BRL 40,000,000.00 per year, divided between professional services (35% on average) and hospital services (65% on average). These percentages varied little over time. The procedure informed for approximately 95% of cases was "Post-abortion/puerperal curettage". "Post-abortion uterine evacuation via manual vacuum aspiration (MVA)" corresponded to only 5% of procedures, on average, in the period we analyzed. There was a progressive reduction in the number of hospitalizations in the period, which was present in all regions, albeit unequally. The highest percentage reduction occurred in the Northeast (12.1%), followed by the Central (11.4%), Southeast (10.7%), North (4.5%) and South (0.9%). At the national scale, this reduction was of 9.5%.
In the period we analyzed, SIH also recorded a reduction of hospitalizations for procedures related to deliveries. This reduction, which was of 5.9% for Brazil as a whole, is also present in all regions, but unequally. The Northeastern region also had the greatest reduction (12.2%), followed by the North (7.6%), Central (4.9%), Southeast (0.8%) and South (0,2%). The ratio of the number hospitalizations from abortion and the number of hospitalizations for deliveries experienced a discreet reduction in the country (3.8%), going from 105 hospitalizations for abortions for each 1,000 hospitalizations for delivery in 2008 to 102 hospitalizations for abortion for every 1,000 hospitalizations for delivery in Cad. Saúde Pública 2020; 36 Sup 1:e00188718   Table 1).
The analysis according to hospitalization diagnosis, including the groups "miscarriage", "abortion due to medical reasons" and "other pregnancies that end in abortion", from 2008 to 2015, shows that an average of approximately 212,000 hospitalizations/year were recorded in SIH. The total average value was of around BRL 47,000,000.00 per year, divided between professional services (34% on average) and hospital services (66% on average), with no important variations in the period.
As in the analysis of hospitalizations according to procedure, there was also an 8% reduction in the number of hospitalizations between 2008 and 2015, however, with no equivalent reduction in the total value informed. There was a reduction in the number of hospitalizations in all Brazilian regions except the South, where there was a very discreet increase.
The age group 20 to 29 years was the most frequent among women hospitalized with an abortion diagnosis in Brazil for the entire period of analysis, however, it had a reduction form 49.8% in 2008 to 44.5% in 2015. There was an increase in the proportion of the 30 to 39 year group, going from 25.9% in 2008 to 31.8% in 2015. This pattern in common to all regions. There are no hospitalizations recorded with no age information.
The high percentage of records with missing information (average of 38.2%) makes it impossible to analyze hospitalizations with an abortion diagnosis according to the race/color variable. The data on hospitalization according to SIH diagnosis available in the Tabnet tool do not include the variables educational level and marital status.
It is impossible to identify, based on the procedure, if the abortion was induced or if it was a miscarriage and there is no specific ICD-10 category for induced or illegal abortion. These should be classified in the category O05 (other types of abortion) or O07 (failed abortion attempt). Abortion due to medical and legal reasons, in turn, have a specific ICD-10 code. The analysis of SUS hospitalizations with diagnoses in the group "abortion due to medical reasons" is one of the possible ways of assessing access to legal abortion in Brazil. The total of these hospitalizations in the country had an initial reduction between 2008 and 2009 and was stable from 2009 to 2015. The average number of Cad. Saúde Pública 2020; 36 Sup 1:e00188718 hospitalizations in this period was of approximately 1,600/year. The same stability in the total hospitalizations was verified in the Southeast. In the Northeast, there was an important increase and, in the remaining regions, there was a reduction in the number of hospitalizations in the period, especially in the South. The average number of hospitalizations according to diagnosis in the other groups was of 109,000 hospitalizations/year in the group "miscarriage" and of 100,000 hospitalizations/year in the group "other pregnancies that end in abortion". There was an inversion in the proportion of hospitalizations between the groups "Miscarriages" and "other pregnancies ending in abortion". "Miscarriages" stopped being the most frequent category, corresponding to 47% of hospitalizations in 2015.
The SIH records of hospitalizations related to abortion was considerably inconsistent.

Discussion
In 2015, Brazil investigated 91.5% of deaths of woman of reproductive age. Despite the high percentage of investigations, there are clear problems. The investigation proportion is different when calculated based on the SIM investigation module or using the specific variable in the database, which suggests that the investigation information may not be fully incorporated into the database. The percentage of missing information regarding the moment of death, in terms of the pregnancy-puerperal cycle (29.2% for Brazil in 2015) reinforces this idea. It is possible that the lack of correction, by the investigation, of the underlying cause compromises the identification of maternal deaths, contributing to under-reporting. The same determinants that condition the under-reporting of maternal deaths in general apply to deaths from abortion recorded in SIM, especially those related to induced abortion. Other factors may play a role, such as legal, religious, ethical and family factors, among others. These factors may interfere with the determination and identification of the cause of death from abortion.
We identified 770 maternal deaths with abortion as the underlying cause in SIM from 2006 to 2015 which, if added to the 220 deaths that have abortion listed as a cause, would represented an increase of approximately 29% in the total of deaths associated with abortion in the period. Among the 220 deaths with causes associated with abortion, there are some with underlying causes that are frequently associated with abortion, such as puerperal infection. In some of these cases, the incorporation of the investigation results could identify abortion as the cause. There are others in which the ICD-10 rules determine that they should receive a specific code, such as late maternal deaths (more than 42 days of puerperium), which must always appear as the underlying cause in the category O96 of ICD-10 ("death, by any obstetric cause, occurring more than 42 days, but less than one year, after delivery"), even if abortion is the underlying cause. Among deaths with mentions of abortion with underlying causes out of ICD-10 Chapter XV, some have well-specified underlying causes, such as neoplasms and external causes, in which abortion may have been a consequence of other conditions. There are, however, deaths with mention of abortion and unspecified causes, such as sepsis and peritonitis, which may be consequences of abortion. A study carried out in Minas Gerais, from 2000 to  Cad. Saúde Pública 2020; 36 Sup 1:e00188718 2011, on abortion-related maternal mortality in light of multiple causes concluded that abortion was identified as the underlying cause in around 11% of deaths and a further 4% had causes associated with abortion. In the deaths whose associated cause was related to abortion, underlying causes were classified as maternal in 56% of deaths, while 44% were classified as non-maternal causes 18 . This study found similar results.
SIH recorded approximately 200,000 hospitalizations/year for procedures related to abortion from 2008 to 2015. "Post-abortion/puerperal curettage" was the procedure described in 95% of cases, while "Post-abortion uterine evacuation via manual vacuum aspiration (MVA)" corresponded to only 5%. SIH records procedures that require hospitalization. Procedures carried out without hospitalization are not recorded. The difference in the need for hospitalization for both procedures may account for part of the difference in the proportion of hospitalization for both procedures.
The number of hospitalizations from abortion identified in this study is considerably inferior to the estimates of the number of abortions in Brazil. A study carried out in Brazil, Chile, Mexico, Peru, Colombia and the Dominican Republic based on records of curettages performed in public services and on interviews with health professionals estimated, in 1991, the occurrence of 1,443,350 abortions in Brazil 7 . More recent estimates pointed to the occurrence of between 687,347 and 865,160 induced abortions in Brazil in 2013 19 . The number of hospitalizations from abortion that can be identified is influenced by the segmentation of the information for the public sector and by the already-mentioned quality of the information regarding the abortion diagnosis. It further refers to the proportion of terminations, whether miscarriages or induced abortions, which require hospital admission. The number of hospitalizations from abortion is only one source of data for estimating the total number of abortions that take place in Brazil.
The trend of reduction in the number of hospitalizations from abortions, of 9.5% in the analysis according to procedures, is also identified by other studies. A study that used SIH data from 1995 to 2013 estimated the number of induced abortions in the country and concluded that there was a 27% decrease in the hospitalizations of woman of reproductive age due to complications from abortion and the estimate of the annual number of induced abortions was reduced by 26% 19 .
The increase in the proportion of deaths recorded in SIH in relation to deaths from abortion in SIM, from 47.4% in 2008 to 72.2% in 2015, could have two explanations. The first is the increase in the proportion of deaths within hospitalizations in the public system, which are recorded in SIH, in comparison with the private sector. The second is the better recording of deaths in hospitalizations in SIH, despite the already-discussed fragility of the data, which mean this analysis must be considered with caution.
The profile of women who died from abortion that we identified coincides with findings from other studies. One study carried out in Minas Gerais 18 presented the characteristics of women who died in relation to abortion as: women aged 20 to 34 years, single (68%) and black (70.5%), most with less than 7 years of schooling. The authors further noted that around 40% of data referring to the variables educational level and the moment of death in terms of pregnancy or puerperium were blank. Death from abortion characterizes a situation of health inequity due to the higher number of deaths among more vulnerable groups (low educational level and black race/color) 18 .
One limitation of this study was that we were unable to include data on the occurrence of abortions that did not require hospitalization and those that resulted in hospitalizations in the private sector. On the other hand, it has shown that more vulnerable women are more likely to die after an abortion, even observing only those who turn to the public system, who are known to be more susceptible to an unsafe abortion 1 .

Conclusion
There are no data on unsafe abortion in the Brazilian information health systems.
The official databases do not enable estimates of the number of abortions that occur in Brazil. The available data are restricted to deaths from abortion and hospitalizations due to complications from abortion in the public health system.
Cad. Saúde Pública 2020; 36 Sup 1:e00188718 The attention to adequately entering information into DC is a crucial factor for gaining greater precision in the mortality information. The quality of the information regarding the cause of death, the under-reporting of maternal deaths and those related to abortion, in addition to the completeness and correction of other variables in SIM could be improved with the systematic incorporation of information from investigations of deaths into the SIM database.
These could have an indirect repercussion on the better identification of deaths from abortion. The identification of deaths with mention of abortion in the DC but with underlying causes classified in other ICD-10 categories suggests the need to include a multiple causes analysis to better determine the scale of mortality involving abortion. The improvement of SIH data quality, especially with regard to the regularity of records, diagnoses and deaths, in addition to systematic, regular access to information from the private health system are actions that would contribute to a better understanding of abortion in Brazil.
Despite the under-reporting of deaths, official data on births and deaths enable us to establish the profile of the women who are at highest risk of dying from abortion in Brazil by calculating specific MMRs. These are black and indigenous women, with low educational levels, over the age of 40 or under the age of 14, in the regions North, Northeast and Central, and not in a conjugal relationship. This profile can be used as a risk marker in post-abortion hospitalizations, so as to increase clinical surveillance of these cases, resulting in a more favorable evolution. These groups mainly need access and qualification of reproductive planning and prenatal care actions so as to reduce the risk of abortion, whether induced or miscarriages.