There has been substantial improvement in vital information systems both from the point of view of registration completeness and data quality, when it comes to access to health care and the prioritization of strategies to reach the municipalities with the lowest socioeconomic status [14].
The possibility of using continuous registration data for the elaboration of infant mortality rates in all Federative Units (FU) and Brazilian municipalities from the year 2000 onwards [16, 17] made it possible to estimate the IMR in different geographic cut-offs and to analyze the space-time inequalities.
Mortality in the first year of life, an indicator recognized by its sensitivity to living and health conditions, showed an important decrease between 1990 and 2015. The decrease also pointed out in previous studies [23, 24] undoubtedly reflects the progress achieved in terms of expanding access to medical care [25]. In this period, estimates calculated by the GBD and by the MoH showed the highest annual decreasing rate in the Northeast Region, contributing to the reduction of regional inequalities in infant mortality, which lasted for several decades [10]. The highest IMR reduction in the poorest Brazilian region undoubtedly reflects the benefits related to the expansion of primary health care, which ensured access to basic health services that are important for the health of children and women before, during and after pregnancy [6, 26, 27].
The Global Burden of Disease (GBD) 2017 study, which included estimates for the Brazilian Federal Units, represents an opportunity for studies aimed at analyzing the country's geographical inequalities, using a standardized methodology to correct problems of quality and incomplete information. Regarding the magnitude of the rates, however, the IMR GBD estimates were 20 to 30% higher than those obtained by the MoH from 1990 to 2015 and much higher than those reported by the World Bank, calculated by the United Nations interagency group, responsible for estimating infant mortality [20]. Particularly, in the year 2015, the estimated IMR by the World Bank for Brazil was 14.0 per 1000 LB, very close to that estimated by the MoH (13.5), while the GBD estimate was 16.4 per 1000 LB.
Despite the possible IMR overestimation by the GBD, when the estimates of the GBD and the MoH are compared by FU, the correlations are positive and significant, and therefore similar ratios were obtained between the FU estimates calculated by the two methods. However, in 2015, the IMR estimated by the GBD was 40% higher in the South Region, and 50% higher in the states of Paraná and Santa Catarina. This region has the best socioeconomic status in the country and is considered to have the most complete vital information. In the three states of the region, infant mortality rates are calculated directly by their definition without any correction [19]. The largest discrepancy in the South Region indicates the need to review the GBD methodology applied to Brazil at subnational levels.
The results presented in this study undoubtedly show the reduction of inequalities in infant mortality, regardless of the geographic scale and the calculation methodology. At the level of Federative Units, the ratio between the highest and lowest estimates of the IMR was reduced from 4 to 2 between 1990 and 2015, both when using the estimates produced by the MoH and by the GBD. In addition, the spatial distribution of infant mortality in 2015 in Brazilian municipalities outlines a much more homogeneous picture than in previous years.
In Brazil, the focus on inequalities at the regional level has proved especially important to promote actions and programs to reduce the socioeconomic gap. Prioritization of the poorest municipalities has shown significant impacts on reducing the historic regional gap in infant mortality rates, unnecessary hospitalizations, and under-five mortality rate due to undefined causes and unassisted deaths [28].
As a consequence of the considerable decline in infant mortality, the magnitude of the national rate in 2015 was similar to that observed in countries with the same per capita income, something that did not occur until the mid-2000s [2], when the IMR in Brazil was higher than expected according to the World Bank model that relates infant mortality to the per capita GDP of the world's countries [20]. Northeastern states, however, performed better than those in the Southeast Region, probably due to the lower rates of decrease among neonatal deaths. These findings corroborate those found in the GBD study. Based on the GBD socio-demographic index, estimates of infant mortality rates in some Northeastern states were lower than expected [25].
The comparison of IMR municipal data by income deciles also showed a significant reduction in all measures of health inequality from 2000 to 2010. The decrease in all health inequality measures that go beyond the reduction of income inequality is likely to be reflecting the effects of the income transfer program “Bolsa Família”, which, coupled with the expansion of primary care, has reinforced this effect on child health [29].
While progress in reducing child mortality is evident, there are still challenges to overcome. The current pattern of mortality in the first year of life, which is concentrated in the early neonatal period and presents a lower rate of decrease the closer to delivery, shows the importance of factors related to pregnancy, delivery and postpartum, generally preventable through quality health care [30, 31].
Despite the marked reduction of inequalities in various indicators of maternal and child health and the remarkable expansion of coverage of prenatal care and hospital delivery after the implementation of the Unified Health System [32], additional efforts aimed at improving access to good quality childbirth care are crucial. It is necessary to integrate actions developed in primary care to the services of childbirth care, to potentiate the municipal capacities in providing adequate assistance to maternal and delivery care [33].
In this context, we highlight the importance of monitoring the perinatal mortality rate, considered as a key health outcome for interpreting the impact of maternal-child health actions [34, 35]. In Brazil, as in other countries, the analysis of perinatal mortality statistics faces additional difficulties, since definitions of stillbirth are not always obeyed [36], and thousands of newborns are not registered as having been born [37]. In 2015, the reported number of stillbirths weighing more than 2500 g was higher than 8000, with almost 70% of those carried for 37 or more weeks of gestation, indicating that many of the deaths occurring shortly after delivery may have been misclassified as fetal deaths.
Other difficulties faced are the recent cuts in social investments as well as in the Unified Health System [38]. The recent economic crisis led to a reduction in private health plan users, and, consequently, to an increase in demand for public services [10]. Despite the progressive decrease in IMR between 2010 and 2015, analyses of infant mortality in 2016 showed a 3% increase in relation to the 2015 estimate, which has raised concerns on the IMR trends after budgetary constraints.
The increase in IMR in 2016 was not only due to the decrease in the number of live births in 2016, probably attributed to the proportion of women who avoided pregnancy soon after the Zika virus epidemic, but also to the increase in the number of infant deaths in the post-neonatal period, in the number of infant deaths due to diarrhea, and in the proportion of ill-defined deaths, while deaths due to congenital anomalies continued to decrease. The increases, although punctual, occurred for specific categories, which were all associated to the worsening of living conditions and lack of access to medical care. Although post-2016 mortality data are not available to assert that there is a reversal in IMT trends, the increases in specific causes seem to reflect the cuts in social policies in recent years.
The limitations of this study refer to the estimates of infant mortality. Although underreporting of deaths and live births has decreased considerably throughout the country, it is not yet possible to estimate the IMR directly, by the ratio of the number of child deaths per 1000 live births. Given the restrictions on the use of mortality estimates based on demographic methods [12], efforts have been made in Brazil to improve the vital information systems and use more appropriate correction factors based on empirical results of active search studies of deaths and births conducted in 2010 and 2014. However, the factors used to adjust the number of deaths and live births may not be adequate for some municipalities. Another limitation is the probable misclassification of newborns who die shortly after birth as fetal deaths, underestimating the infant mortality rate.