Intraurban differentials of perinatal mortality : modeling for identifying priority areas Aspectos intraurbanos de la mortalidad perinatal : modelo para identificación de áreas prioritárias

Objective: To analyze the intraurban spatial distribution of perinatal mortality, its avoidability, and relationship with socioeconomic indicators in Recife, Pernambuco, Brazil, in the period from 2013 to 2015. Method: An ecological study with data from the Information Systems on Mortality and Live Births and the Brazilian Institute of Geography and Statistics, using neighborhoods as the analysis unit. We elaborated an indicator of social deprivation formed by variables from the demographic census. We estimated the Kernel density of the deaths and calculated the Moran index of the perinatal mortality coefficients in the spatial analysis. We elaborated thematic maps of avoidable perinatal mortality and social deprivation. Results: The global statistical analysis of the mortality distribution indicated evidence of spatial aggregation. Moran's index was 0.18. We found clusters of perinatal mortality in neighborhoods of the Central, North, Northwest, and South Regions. In the North, Northwest, Southwest, and South Regions we identified neighborhoods with greater social deprivation and avoidable mortality coefficients. The primary cause of death was of fetuses and newborns affected by hypertensive maternal disorders. Conclusion: We demonstrated intraurban differentials in perinatal mortality among neighborhoods. The stratification of the urban space according to the social deprivation indicator presented a relation with the perinatal mortality and its avoidability.


INTRODUCTION
Perinatal mortality, the period comprised between the 22nd week of pregnancy and the first seven days of life, reflects the access to health care and the quality of the assistance to the prenatal, the labor, and the newborn, constituting an expressive indicator of maternal and infant health. 1 Around the world, perinatal deaths occur primarily in countries of low and medium income. 1 In Brazil, between 2001 and 2015, the perinatal mortality coefficient went from 22.3 deaths per thousand births to 17.4 (21.9% reduction), with the South Region presenting the lowest coefficient and the Northeast, the highest: 15.4 and 21.2 deaths per thousand births, respectively. 2pproximately 47.6% of perinatal deaths occur due to avoidable causes. 3eaths considered avoidable are those unnecessary or potentially preventable by the effective action of health services. 4To classify the avoidability of the deaths, the List of Causes of Death Avoidable by Interventionsof the Brazilian Unified Health System (UHS) was elaborated, 5 which contemplates the neonatal period, whose circumstances and etiologies are similar to the perinatal. 6his list classifies the causes of avoidable deaths in reducible by the following: immunoprevention actions; proper care to the pregnant women, labor, fetus, and newborn; adequate actions of diagnosis and treatment; appropriate health promotion actions linked to suitable health care actions. 5o ensure interventions directed toward perinatal deaths, there are recommendations of its inclusion in international and national pacts. 7Such deals act as instruments capable of making the magnitude of mortality explicit, in addition to favoring the planning of actions oriented towards combating these deaths. 7he global health strategy for the woman, the child, and the adolescent upon incorporating the avoidability of perinatal deaths aggregated perspectives of advances regarding the recognition of its importance. 8However, in the Brazilian inter-federative pacts, the perinatal mortality coefficient is little included, thus compromising the visibility of the problem, almost always mentioned through academic productions. 7he difficulties for reducing mortality, especially in areas with precarious living conditions, are consequences of the invisibility of perinatal deaths, which, in general, although avoidable, remain secondary in the public policies of developing countries such as Brazil. 7The unequal distribution of such deaths in the territories may demonstrate the socioeconomic segregation among populational groups 9 usually revealed by characteristics regarding income, education, occupation, race/skin color, gender, and housing or workplace conditions. 10tudies that show the magnitude of perinatal death, its avoidability and relation with social deprivation indicators in the intraurban space may unveil inequities among populational groups. 11,12The agglomerates of perinatal deaths and social deprivation may be identified through spatial analysis. 11The methods of visualization, exploratory analysis, or modeling of georeferenced data enable integrating epidemiological, socioeconomic, and environmental information without dissociating them from the geographic space, 13 allowing the detection of risk factors and the identification of priority areas for the health sector interventions. 11,13esearches that stratify the urban agglomerates by social deprivation may favor the understanding of the relationship that exists among the socioeconomic and environmental conditions and perinatal mortality.When comparing such strata with the spatial distribution of the avoidable deaths, it is possible to contribute to the planning of public policies oriented towards the areas that require a higher priority so to reduce the iniquities.The objective of this study was to analyze the intraurban spatial distribution of perinatal mortality, its avoidability, and its relationship with socioeconomic indicators in Recife, Pernambuco, Brazil, in the period from 2013 to 2015.

METHODS
This is an ecological study carried out in Recife, capital of the state of Pernambuco, in the Brazilian Northeast, which had 1,599,514 inhabitants (2013) distributed among 218.5 km, 2 with a heterogeneous occupation pattern in which highly valued areas coexist with others that present relevant structural problems. 14he 94 neighborhoods are arranged in six Political-Administrative Regions (PARs): 1-Center, 2-North, 3-Northeast, 4-West, 5-Southwest, and 6-South 14 (Figure 1).
The analysis unit considered was the neighborhood.For calculating the perinatal mortality coefficients, we included all perinatal deaths (fetal and early neonatal) registered in the Mortality

Perinatal mortality and intraurban differentials
Canuto IMB, Alves FAP, Oliveira CM, Frias PG, Macêdo VC, Bonfim CV Information System (MIS) and the living births in the Living Birth Information System (LBIS) of residents of Recife, occurred from 2013 to 2015.For the avoidable perinatal mortality coefficients per neighborhood, we used all the perinatal deaths with weights over 1500g and underlying causes considered avoidable according to the criteria of the Brazilian List of Causes of Death Avoidable by Interventions of the UHS. 2 To build the social deprivation indicator (SDI), we used the data from the 2010 Census of the Brazilian Institute of Geography and Statistics (IBGE, in Portuguese, Instituto Brasileiro de Geografia e Estatística).The indicator is composed of the following variables 15 regarding the proportion of permanent private households that are: without a water supply linked to the general network; without garbage collection by a cleaning service; without an exclusive bathroom for residents or a toilet and sanitary exhaustion through the general sewage or pluvial network; without nominal monthly income; and with non-alphabetized responsible persons.We selected these indicators due to their relationship with the living condition of the population and with child and perinatal mortality. 3e calculate the SDI as the standardized average of economic and social variables.In each neighborhood, we verified the occurrence relative to the socioeconomic variables.The neighborhood with the most occurrences in a given variable received a score of one, while the one with the lowest incidence received zero.We calculated the scores of the other neighborhoods using the following formula: 15 SCRneigh, yv = (OCyv − OCmin, v) / (OCmax, v − OCmin, v), where: SCRneigh, yv = score of neighborhood "y" associated with variable "v".OCyv = ocurrence of variable "v" in neighborhood"y".
OCmax, v = maximum occurrence for variable "v", observed among all neighborhoods.
Therefore, the SDI for each neighborhood is calculated as the mean of the already calculated scores: where SDIy is the social deprivation indicator of neighborhood "y" and "n" is the number of variables used in its calculation.The neighborhoods were grouped by tercile, which resulted in the strata: low, medium, and high.
In the spatial analysis, version 2.14.3 of program QGis ® automatically located the geographic coordinates of the maternal residency points of the perinatal deaths through the georeferencing of the addresses with a search on the GoogleMaps ® database.This technique consists in the attribution of geographic coordinates to the addresses, and, for its use, we performed the adaptation of the MIS database, the georeferencing itself, and the verification of its quality.The address of each event was searched in the MIS database and compared to those contained in the address database using version 2.14.3 of program QGis ® .
We applied the kernel estimation technique to promote the statistical smoothing and verify the influence of a point's density in the existence of others in close areas. 17We manually checked the points identified as geometric centers and approximate.To estimate the intensity of the event, we used a kernel with an adaptive radius of 1500m and a quartic function.We used version 4.2.2 of program TerraView ® to construct the thematic maps of the coefficients of perinatal mortality and avoidable perinatal of the SDI, grouped by tercile and distributed by neighborhood.We applied the spatial autocorrelation of the perinatal mortality coefficient, with the respective Moran values.The calculation of Moran's index indicates clusters with analogous risks for the event of interest, with its result being from -1 to +1. 11 Results close to zero indicate null autocorrelation among the areas and their neighbors. 11Positive results signal similarities among the neighboring microregions, while negative values suggest there are no similarities. 11he cartographic base of addresses and the digital meshes used are available on the website of the city hall of Recife (http://www.recife.pe.gov.br/ESIG/documentos/Informacao/InformacaoManualArquivos.htm) and present a SIRGAS 2000/ UTM zone 25S system of coordinate references.
This research obtained the approval of the Research Ethics Committee of the Joaquim Nabuco Foundation: Protocol no.2.099.667 of June 5 th , 2017.

RESULTS
In the period studied, the MIS registered 1,112 perinatal deaths of residents of Recife, with the perinatal mortality coefficient being of 15.80 deaths per thousand births.Figure 2A shows spatial agglomerates of perinatal deaths with a high-intensity kernel in two neighborhoods of each of the following PARs: North (Água Fria and Alto Santa Terezinha), Northwest (Alto José do Pinho and Morro da Conceição), and South (Brasília Teimosa and Pina).
In the thematic map (Figure 2B), we found spatial agglomerates with higher values in three neighborhoods of the Center PAR, five of the North, eight of the Northwest, six of the Southwest, and two of the South.The spatial autocorrelation of the perinatal mortality coefficients was positive with priority areas (high-high clusters) made evident by the Box Map in the Center, North, Northwest, and South PARs (Figure 2C).In the Moran Map (Figure 2D), areas with statistically significant autocorrelation (I=0.18;p-value=0.02)and high attention priority stood out, shown in darker shades, in two neighborhoods of the Center PAR (Cabanga and São José), one in the North (Fundão), three in the Northwest (Alto José Bonifácio, Brejo da Guabiraba, and Vasco da Gama), and two in the South (Pina and Brasília Teimosa).The perinatal deaths classifiable by the Brazilian List of Causes of Death Avoidable by Interventions of the UHS, after the separation by weight, presented avoidable causes (n=333; 66.20%), ill-defined causes (n=64; 12.73%), and not clearly avoidable causes (n=106; 21.07%).
In the thematic map of the social deprivation indicator (Figure 3A), one may visualize neighborhoods with high social deprivation in all the PARs.In the thematic map of the avoidable perinatal mortality coefficient (Figure 3B), we noticed spatial agglomerates with higher values in neighborhoods of all the PARs.For presenting high SDI and avoidable perinatal death coefficients, the neighborhoods that deserve emphasis are two in the North PAR (Cajueiro and Torreão), one in the Northwest (Morro da Conceição), and two in the Southwest (Jardim São Paulo and Totó) (Figures 3A and 3B).Table 1 shows that the highest avoidable mortality coefficients are in the strata with the worst social conditions.
The perinatal deaths could have been avoided especially  with the adequate care to women during pregnancy (n=164; 49.25%).The primary causes of death were the following: P00.0 -Fetus and newborn affected by hypertensive maternal disorders, which presented a higher proportion in the North PAR neighborhoods (n=13; 28.89%); and P20.9 -Intrauterine hypoxia, with a higher value in the West PAR (n=12; 19.35%).It is worth emphasizing the A50.0 -Congenital syphilis, which presented the highest proportion in the Northwest PAR (n=7; 11.48%) (Table 2).

DISCUSSION
The perinatal mortality coefficient of Recife from 2013 to 2015 was of 15.80 per thousand births, 10.1% below the national average (17.4). 2 The spatial analysis of the perinatal mortality made explicit the differentials among the neighborhoods of the city, visualized in the kernel density map and the clusters with areas of priority attention without agglomerates in the Center, North, Northwest, and South PARs, with such areas harboring pockets of concentrations of the municipality's poorer populational groups and robust social inequalities. 18esearch carried out in the United States which thoroughly categorized the territory revealed segregations of health conditions on the geographic space, to aid the improvement of maternal and infant health conditions. 19Similarly, it is possible to signal areas with higher concentrations of deaths and indicate the presence of risk factors in specific localities, similar to the described in the literature. 20n Recife, the quality of the data on vital information considered reliable accompanied the improvement of the information systems of continuous record in Brazil, enabling to the public management a broad view of the locations with higher risks for mortality and collaborating for its combating and the reduction of inequalities. 21National and international research that value the use of spatial analysis techniques, signaling urban agglomerates without maps of points or of rates with a need of priority interventions are widely used due to their importance in subsidizing the management of health actions. 11,22,23his study, as another carried out in municipalities of the state of São Paulo, analyzed the spatial distribution of underlying causes of deaths. 24Regarding the classification of the avoidability of perinatal deaths, the highest proportion was considered avoidable, reducible especially with the proper care to women during pregnancy, a finding similar to previous ones regarding child mortality in Recife 25,26 and other municipalities of the country. 27The predominant cause of death in all the PARs except the Perinatal mortality and intraurban differentials Canuto IMB, Alves FAP, Oliveira CM, Frias PG, Macêdo VC, Bonfim CV West was the fetus and newborn being affected by hypertensive maternal disorders, as described by other authors. 25,26The situation of congenital syphilis remains unacceptable, especially in the Northwest PAR, a result of the low effectivity in the care of pregnant women, particularly those more vulnerable. 28n the literature, most of the problems that result in perinatal deaths originate during pregnancy and are treatable when identified early. 29According to studies that report the Brazilian reality, when adequate, prenatal care enables the immunization of pregnant women, the early detection of morbidities and prematurity risk factors, 30 the reduction of the vertical transmission of diseases, and the treatment of abnormalities, 29 in addition to directing high-risk pregnant women to hospital units capable of providing intensive care to the potentially serious newborns increasing their chances of survival. 31Guidance is also provided about the care with the newborn in the post-partum. 30n this research, the avoidable perinatal mortality coefficient is higher in the strata with significant social deprivation indicators.Therefore, the social disparities favor the exposure of the more impoverished population to health risks. 32Furthermore, it makes opportune the access to the services and quality for privileged populational groups, characterizing the iniquities, as a Mexican study demonstrated. 33The spatial distribution of social deprivation may indicate areas with greater potentials of occurrence of adverse health outcomes, 15 and the agglomerates of avoidable deaths are capable of revealing iniquities. 24

CONCLUSION
Upon detecting equivalences among spatial agglomerates of perinatal mortality, its avoidability, and indicators of social deprivation, this study identified inequalities and signaled areas that require priority attention to reduce deaths.The analysis unit employed imposes limits that may mask inequalities within neighborhoods considered of low priority.Likewise, it is not possible to infer that the indicators analyzed are homogeneously distributed in the areas deemed of priority.However, when incorporated to the actions of the health sector, the spatial analysis will be able to subsidize the decision making of the managers by indicating locations in the territory that present more significant health risks and need immediate allocation of resources.

FINANCIAL SUPPORT
Brazilian National Council of Scientific and Technological Development (CNPq) for funding the research with a scholarship of scientific initiation to the first author, from July of 2016 to June of 2017.

Figure 1 .
Figure 1.Geographic localization of the municipality of Recife and its division by Political-Administrative Regions.Recife, 2013 to 2015.Source: Municipal Government of Recife.Recife (PE), 2014.

Table 1 .
Avoidable perinatal mortality coefficient according to the strata of social deprivation indicator.Recife, 2013-2015.