Birth in Brazil survey : neonatal mortality , pregnancy and childbirth quality of care

This study examined neonatal deaths in the livebirths cohort in the Birth in Brazil survey, which interviewed and examined medical records of 23,940 mothers from February 2011 to October 2012. Potential risk factors were analyzed using hierarchical modeling. Neonatal mortality rate was 11.1/1,000, the highest rates occurring in the North and Northeast regions and in lower social classes. Low birth weight, risks during pregnancy and conditions of the newborn were the main factors associated with neonatal death. Inadequate prenatal and childbirth care point to unsatisfactory quality of health care. Difficulty in gaining hospital admission for delivery, and children with birth weight < 1,500g born at hospitals without a neonatal intensive care unit, indicate gaps in health system organization. Deaths from intra-partum asphyxia in term babies and late prematurity express preventable neonatal mortality. Better quality health care, especially hospital care during labor and birth, poses the main public policy challenge to progress in reducing mortality and inequalities in Brazil. Infant Mortality; Hospital Care; Health Services Evaluation; Parturition ARTIGO ARTICLE


Abstract
This study examined neonatal deaths in the livebirths cohort in the Birth in Brazil survey, which interviewed and examined medical records of 23,940

Introduction
Neonatal mortality has been the main component of infant mortality since the 1990s in Brazil and continues high, at a rate of 11.2 deaths per 1,000 live births in 2010 1 .In 2011, the infant mortality rate in Brazil was 15.3 per 1,000 live births, achieving the 4 th Millennium Development Goal, the commitment by United Nations member governments to improve child health and reduce child mortality by two thirds between 1990 and 2015 1,2 .Those mortality rates are considered to fall short of Brazil's potential, however, reflecting unfavorable conditions of life and health care, in addition to historical regional and socioeconomic inequalities 3,4 .
At present the main component of infant mortality is early neonatal (0-6 days of life), and infant deaths occur largely (25%) in the first 24 hours, indicating a close relation to care during labor and delivery 5 .In the literature, the main causes of death are prematurity, congenital malformation, intrapartum asphyxia, perinatal infections and maternal factors, with a considerable proportion of deaths preventable by health services 5,6 .
Although births in Brazil occur predominantly in hospital (98.4%) and delivery is attended by doctors (88.7%) 7 , the results are unsatisfactory compared with lower neonatal and child mortality coefficients achieved elsewhere 8 .This situation has been termed the 'Brazilian perinatal paradox': deliveries and births are intensely medicalized, but higher rates of maternal and perinatal morbidity and mortality persist, and are possibly related to poor quality of care and use of obsolete and iatrogenic procedures, which may affect perinatal outcomes 9,10 .One of the most prominent examples in that respect is the high rate of caesarean sections in Brazil -53.7% of births in 2011 1 .
Studies of the quality of the process of care during labor, delivery and childbirth are recent initiatives and still few in number 9,11,12 .More in-depth understanding of care processes during delivery and childbirth, and their effects on neonatal mortality, may contribute to informing actions to reduce the latter more intensively 13 .
The purpose of this study was to examine the profile of neonatal deaths identified in the national Birth in Brazil survey, and the associated factors, considering socioeconomic and demographic contextual factors, characteristics of the pregnant women and newborns, and the process of prenatal care and care during labor and delivery.

Methods
Birth in Brazil is a nationwide, hospital-based survey of women about to give birth and their newborn babies, conducted from February 2011 to October 2012.The sample was selected in three stages.The first, comprised hospitals performing 500 or more deliveries per year, stratified by Brazil's five macro-regions and by hospital location (state capital or elsewhere) and type (private, public and mixed).The second comprised days (at least seven days at each hospital), and the third, the women about to give birth.At each of the 266 hospitals sampled, 90 women about to give birth were interviewed, totaling 23,940 subjects.More detailed information on the sample design can be found in Vasconcellos et al. 14 .In the first stage of the study, the women about to give birth were interviewed face-to-face during their hospital stay, data were drawn from the women's and newborns' patient records, and the women's prenatal record cards were photographed.Telephone interviews were then conducted at before six months and at twelve months after birth to collect data on maternal and neonatal outcomes.Detailed information on the data collection is reported in do Carmo Leal et al. 15 .
For this study, the outcome variable examined was neonatal death, defined as deaths of live-born babies, regardless of birth weight and gestational age, occurring before the 28 th day of life.In order to identify the neonatal deaths that occurred during the study period, and to obtain relevant information, probabilistic record linkage was established between the Birth in Brazil data base and the neonatal deaths that occurred in Brazil in 2011 and 2012 as recorded in the Mortality Information System (SIM) and the Information System on Live Births (SINASC).Probabilistic record linkage was performed using the software OpenRecLink 16 and the variables mother's name, and newborn's date of birth, sex, and date of death, in three stages: standardization, blocking and record pairing; the pairs were classified into true, false and doubtful 17 , and reviewed manually by the process described by Camargo Jr. & Coeli 16,18 .For variables for which there was no information in the SIM or SINASC, values were imputed according to the likelihood of each category, as estimated by regression model for each variable based on the group of infants hospitalized, because they displayed characteristics similar to those of the infants who died.
For the hierarchical modeling of determinants of neonatal death, the exposure variables were divided in to four blocks 19,20 , as follows: • Block 1 -socioeconomic and demographic: maternal domicile (region of Brazil, location of municipality, social class -ABIPEME, http:// www.abipeme.org.br),maternal characteristics (race/skin color, schooling, marital status and age in years), and sex of the newborn; • Block 2 -prior history and current pregnancy: parity; neonatal death, stillbirth, low birth weight and prior prematurity; type of gestation, adequacy of prenatal care by the Kotelchuck index 21 , and maternal complications, considering pre-existing risk conditions persisting in the current pregnancy and complications in gestation or labor; • Block 3 -care process during childbirth: referral hospital for high-risk pregnancy, hospital with neonatal intensive care unit (neonatal ICU), hospital funding type (private, public, mixed), more than one maternity facility approached before gaining admission, partograph used during labor, companion present during hospital stay for childbirth, delivery type, newborn with birth weight < 1,500g in hospital without neonatal ICU, good practices during labor and good practices during delivery.The latter two summary variables were developed as markers for care in order to evaluate the use of evidence-based procedures in care during labor, and delivery by expert consensus.The labor care practices selected -which are recommended in the literature 22,23,24,25,26 , according to systematic reviews -were the women's: having the option of a liquid or light diet during labor, in contrast with the prevailing routine prescription of fasting; being given freedom of movement, counter to the prevailing practice of bed rest; access to pain relief non-drug methods; having a companion present during labor; having a partograph used in monitoring labor; and oxytocin used only with a partograph, contrary to the prevailing practice of using oxytocin without a partograph to monitor labor.In the same way, the variable 'good childbirth practices' was constructed considering non-use of the Kristeller maneuver (uterine fundal pressure during delivery, a procedure unsupported by scientific evidence, which can produce pain and harm to parturient and newborn); and being able to choose upright delivery positions, contrasting with the predominant practice of delivery in the lithotomy position; • Block 4 -general conditions of the newborn and care for the newborn: birth weight, gestational age, congenital malformation, admission to neonatal ICU, use of mechanical ventilation, use of surfactant, presentation for birth, and Apgar < 7 at the 5 th minute of life.
Deaths were also described by the main groupings established by França & Lansky 5 for neonatal deaths, drawing on the causes recorded on the death certificate obtained by linking the survey data base and the SIM and the causes stated in the hospital patient records (only for 25 deaths not identified in the SIM).This list of causes groups the codes of the 10 th revision of the International Classification of Diseases (ICD-10) so as to give greater visibility to the main causes of death: prematurity, congenital malformation, perinatal infections, intrapartum asphyxia, maternal factors, respiratory problems, and other groupings.

Statistical analysis
The descriptive statistics consisted first in presenting absolute and relative frequencies and numerical summary numbers for the chosen variables by regions of Brazil.For all the variables used in the hierarchical model, neonatal mortality rates were calculated as a measure of risk.The initial statistical analysis consisted in using the chi-square test to evaluate the hypothesis of homogeneous proportions, comparing the survivor and neonatal death groups.
Bivariate analyses were performed between the exposure variables and the occurrence of neonatal death, to obtain crude odds ratio (OR) as a measure of magnitude of association, as appropriate to the logistic regression model.Variables displaying p ≤ 0.20 were retained for multivariate analysis by multiple logistic regression, considering the hierarchization in blocks.In keeping with the conceptual model, the socioeconomic and demographic variables were analyzed as the most distal level in terms of proximity to the outcome.Variables relating to prior maternal history and current pregnancy, and to the childbirth care process, were considered intermediate.Lastly, the block of variables relating to overall conditions and neonatal care entered the model as the proximal level.Variables from each block with p ≤ 0.10 were retained in the hierarchized model to control for residual confounder effects on the variables.In analyzing and discussing the results, exposure variables with a 5% level of significance were considered to be associated with neonatal death 12,15,19,27 .Cases of collinear variables were evaluated using the variance inflation factor, and the variable with lower p-value was selected.
In the statistical analysis, the complex sample design was taken into account by way of the svy commands in the Stata software, version 12 (Stata Corp., College Station, USA) and the complex samples module of the IBM SPSS statistics package, version 18 (IBM Corp., Armonk, USA).All estimates were weighted, because the selection probabilities were unequal 15 .

Ethical considerations
The project was approved by the Ethics Research Committee of the National School of Public Health, Oswaldo Cruz Foundation (ENSP/ Fiocruz), opinion n. 92/10.The hospital managers and women selected who agreed with and signed the declaration of free and informed consent were interviewed.

Results
In this study, 24,061 live births and 268 neonatal deaths were identified, resulting in a weighted neonatal mortality rate of 11.1 deaths per 1,000 live births.Table 1 shows some characteristics of the deaths, by location of the maternity facility where the birth occurred, by region of Brazil.The deaths were concentrated in Brazil's Northeast (38.3%) and Southeast (30.5%) regions and among premature and low birth-weight newborns (81.7% and 82%).Southeast, Central and South regions of Brazil had the highest preterm.Extreme prematurity (< 32 weeks) and very low birth weight (< 1,500g) represented 60.2% and 59.6% of deaths, respectively, with highest rates in the Central and Southeast regions.The highest rate of full-term newborn deaths was in the Northeast (21.3%).
Of the groups by cause of neonatal death, the prematurity group predominated, responding for about 1/3 of cases, followed by congenital malformation (22.8%), infections (18.5%), maternal factors (10.4%) and asphyxia/hypoxia (7%).The Northeast and North regions showed the highest rates of death recorded as from infection (26.9% and 20.7%), compared with 10.5% and 7.7% in the South and Central regions, respectively.The rates of deaths recorded as from congenital malformation were higher in the South and Southeast (42.1 and 35.9%), while the North and South returned the highest rates of death from asphyxia.
Tables 2, 3 and 4 show the distributions of births and deaths, neonatal mortality rates and bivariate analyses of the blocks of variables proposed for the hierarchical model.The women who participated in the study were predominantly 20 to 34 years old (70.8%), had brown skin (54.6%), and belonged to social class C (49.1%), had begun or completed middle school, and lived with a partner.Most of the births and deaths occurred in municipalities other than the state capital, and 56.7% of the children were delivered by caesarian section.As regards the neonatal deaths, 21.2% of the mothers were adolescents, 33.5% did not live with their partner, and one third had fewer than eight years' schooling (Table 2).
The lowest neonatal mortality rates (per 1,000 live births) were found in the South (6.2), Southeast (8.0) and Central (8.4) and the highest, in the North (22.3) and Northeast (14.5).Notable among the other socioeconomic and demographic variables studied (Table 2) was that  the maternal mortality rate was higher for male newborns, for mothers of social classes D+E, living in state capitals, adolescents and those 35 or more years old, and was four times higher for mothers with little schooling (Table 2).No association was found between neonatal death and race/skin color.
The highest neonatal mortality rates occurred among children weighing less than 1,500g born in hospitals without neonatal ICU, those with very low birth weight (< 1,500g), extreme prematures (< 32 weeks), those with Apgar < 7 at the 5 th minute of life, those who used mechanical ventilation or surfactant, those with con- genital malformation, those in breech presentation, twins, those whose mothers reported approaching more than one hospital before being admitted, who had unfavorable prior maternal and obstetric histories, for whom no partograph was charted during labor, babies born in public hospitals, in referral hospitals for high-risk pregnancies, in hospitals with neonatal ICU, mothers who had no companion during their hospital stay for childbirth, and babies born by vaginal delivery (Tables 3 and 4).
Rates were also high among mothers who received inadequate prenatal and labor care.Care practices considered adequate were used during labor in only 1% and 1.2% of births, i.e., in the vast majority of cases they were not used or were used only partially.There was no statistical difference between deaths and live births in that good practices were used during delivery in similar proportions in both groups, only 2.1% and 2.3%, respectively.Meanwhile, non-recommended practices were frequent: the Kristeller maneuver, for instance, was used in 36.5% of all vaginal deliveries and in 21.5% of neonatal deaths (data not presented).
The proximal variables, those relating to the newborn and the current pregnancy, displayed stronger associations with neonatal death in the Note: the total n may vary by the presence of disregarded variables.
bivariate analysis (Table 4).Extreme prematures and babies with low birth weight were 200 to 300 times more likely to die in the first 28 days of life than full-term newborns with birth weight ≥ 2,500g.Likelihood of the same outcome was also very high for newborns with use of mechanical ventilation, those with < 1500g born in hospitals without neonatal ICU, those with intrapartum asphyxia, those with use of surfactant, and those with congenital malformation.Newborns with breech presentation at birth and twins displayed a strong association with neonatal death (OR 5 to 7).Other factors strongly associated with neonatal death were: risks in the current and prior pregnancies (prior stillbirth, premature and low birth weight babies), mothers with little schooling, multiple hospitals approached before gaining admission, residence in the North region of Brazil, good practices not used during labor and delivery, partograph not used during labor, inadequate prenatal care, public hospital (Brazilian Unified National Health System, SUS), referral hospital for high-risk pregnancy and hospital with neonatal ICU, and not having a companion at some or any time during hospital stay for childbirth.
In relation to the markers for care and for health system organization, most of the deaths occurred in hospitals of the SUS, 50% of babies weighing < 1,500g were born in hospitals without neonatal ICU, 23.3% of mothers received inad-equate prenatal care entailing four times higher risk of neonatal death, while about 40% were not admitted to the first hospital approached and had no companion during their hospital stay for childbirth.Little use was made of partographs to monitor labor, either among newborns who survived (35.7%) or among those who died (36.5%) (Tables 3 and 4).
Table 5 shows the final model, which does not include the variables mother's race/skin color and parity (p > 0.20), delivery type (already considered in the variable partograph), hospital with ICU, and < 1,500g birth weight in hospital without neonatal ICU (collinearity).The variables removed from the final model (p > 0.10) were social class and municipality type (Block 1), prior stillbirth and low birth weight (Block 2), use of surfactant and gestational age (Block 4).
The variables that continued associated with neonatal death were, in decreasing order of magnitude of association: very low birth weight; mechanical ventilation; congenital malformation; intrapartum asphyxia; maternal complications during pregnancy; breech presentation; twins; mothers with little schooling; North and Northeast regions; prior stillbirth; multiple hospitals approached for admission; partograph not used during labor; inadequate prenatal care; mother without companion; prior premature; high-risk pregnancy referral hospital; and male sex.

Discussion
The neonatal mortality found in this study approximates to that observed for Brazil in 2010, of 11.2 per 1,000 live births.Prematurity and low birth weight (especially extremely low birth weight) were the main factors associated with neonatal death.The predominant characteristics of neonatal deaths (very low birth weight and prematurity, followed by congenital malformation) approximate Brazil to more developed countries, where the vast majority of infant deaths occur among newborns with lesser likelihood of survival 4,28,29 .Variables indicating greater severity in newborns were maintained associated to neonatal death.
Late prematurity was associated with 17.1% of neonatal deaths, nine times the likelihood of neonatal death for term newborns.This is related to the growing concern in Brazil over the tendency to increasing prematurity, which affected 11.5% of live births in 2011 30 , a level far higher than in developed countries, where rates are around 7% 1,28,31 .In that scenario, it is fundamental to invest in measures to prevent avoidable prematurity, in addition to improving care for these more vulnerable newborns.Two goals should be kept in focus: prevention in prenatal care, including control of infections and risks in pregnancy, and prevention of iatrogenic prematurity 3,29 relating to undue interruption of pregnancy, as in caesarian sections without technical indication, which are a serious problem in Brazil and contribute to the high total prevalence of caesarian sections 3,31,32 .
Another cause of avoidable and neonatal death that demands specific action is intrapartum asphyxia: 18% of the newborns who died were term babies and 40.9% scored Apgar < 7 at the 5 th minute of life.In Brazil reducing this cause of death is closely related to hospital care during labor and delivery, because the vast majority of deliveries and deaths occur in hospitals and are attended by trained professionals 6 .The causes of intrapartum asphyxia need to be recognized if it is to be prevented.In addition to prenatal care measures to prevent problems of intrauterine hypoxia, according to Lawn et al. 33 , delay in adequate interventions by health services could avert some 36% of intrapartum related deaths in countries where access has been provided to care during childbirth, as in Brazil.
This study pointed to problems in the quality of prenatal care and care during childbirth.Good practices during labor were not used just as frequently among live births (96.6%) as among deaths (99%) and, as a result, there was no significant difference between the groups.It was practically an exception for good practices to be used during childbirth.The Kristeller maneuver, which is not recommended by scientific evidence, was often performed, both among surviving live births and among newborns who died 23,34 .Meanwhile, little use was made of recommended practices, such as the partograph to monitor labor, upright delivery positions, and so on 29 .
The association between perinatal outcomes and the care process has not been as widely recognized Brazil as it should.Implementation of good practices during labor and childbirth is a powerful measure to prevent avoidable neonatal deaths and, as a result, to reduce infant mortality.Persistent use of procedures not recommended by scientific evidence -such as excessive use of oxytocin, immobilization in bed and delivery in the lithotomy position, in which the major vessels are compressed, for instance -compromises intrauterine oxygenation, prolongs labor and delivery, and may cause adverse effects on perinatal outcomes 35 .Coupled to these poor practices, the situations of stress that mothers about to give birth are subject to, such as fasting, solitude, insecurity and disrespectful treatment, also influence perinatal outcomes 25 .In order to change the present situation, it is fundamental to change the model of care, especially in labor and birth, by improving the quality of care provided once access is assured; it is no longer enough to go on offering more of the same 36 .The Ministry of Health initiative, Rede Cegonha 37 , proposes to alter the model of labor and delivery care by using multidisciplinary teams that include obstetric nurses/obstetricians, using protocols and monitoring service indicators with target-coupled funding.That model is being encouraged in other countries, including New Zealand, Canada, United Kingdom, Holland, Japan and Australia 38 , and is being used experimentally at the local level in Brazil with significant perinatal outcomes, such as reduced mortality from intrapartum asphyxia 39 .
Brazil's successful efforts to improve the quality of care for newborns requiring neonatal resuscitation needs to be expanded to primary prevention of intrapartum asphyxia.It is recommended that perinatal care, from the moment the expectant mother is admitted and throughout labor, should indeed be the work of a team including those professions that traditionally take responsibility for the child only after birth (pediatricians, nurses and other personnel).The multidisciplinary team should be co-responsible from the moment the expectant mother is admitted, in order to ensure that appropriate technologies are used (immediate reception, free choice of companion, doula, methods to afford comfort from pain, free choice of birthing position, and so on), and use of protocols to foster physiological progression of labor without unwarranted interventions.
Other care markers reflected problems in the organization of the perinatal care system, such as women about to give birth having to approach more than one hospital before gaining admission, and at-risk newborns being delivered in inappropriate places.Women in labor need urgent care and should be attended to immediately at a health service.Extreme prematures should be born in a higher-complexity hospital; that may be decisive to their survival and obviates the need for subsequent transfer to such a facility after delivery, which entails added risks.In this study, approximately 50% of babies weighing < 1,500g born in hospitals without neonatal ICU died.
The traditional association of neonatal death with factors such as mother's race/skin color and social class did not continue to hold, but this inequality was demonstrated in the association with mothers' lack of schooling.Some authors have pointed to the decrease in neonatal mortality inequality as a result of the economic betterment of the lower-income population and improved access to health services, growth in the numbers of private health plan users in the major metropolises and so on 40,41,42 .On the other hand, the results may have been influenced by the fact that the poorer population (resident in smaller municipalities with facilities handling < 500 births/year and home births) did not take part in the study.Future studies could consider more sensitive indicators and analyses to capture probable intra-urban and intra-regional socioeconomic differences in neonatal mortality.Another aspect that must be considered a limitation of this study was the utilization of hospitals record data of births and deaths, which hinges on the quality of information (for instance, recording the health care process performed, causes of neonatal death and socioeconomic and demographic particulars, such as race/skin color), which can obscure possible inequities.
The North and Northeast regions continued to show association with neonatal death and showed the highest rates of death from perinatal infection, reflecting the need for greater local investment in organizing and improving the quality of care.If effective, timely, quality care actions are made to reach the most excluded population groups, that will also cause more rapid decline in mortality and in the still existing inequalities 44,45 .
Evaluation of the quality of care offered during labor and childbirth in hospitals deserves more in-depth examination, because these are the predominant place of birth in Brazil, and the outcomes of the care processes and markers analyzed in this study were not satisfactory.Prior studies have pointed to important differences in perinatal outcomes associated with hospital performance (material and human resources and care practices), independently of the client characteristics 12,46 .Other studies could explore these aspects in greater depth, detailing care indicators designed to evaluate labor and childbirth care, particularly those relating to the main causes of avoidable death, such as intrapartum asphyxia, iatrogenic prematurity, and infections that are preventable during prenatal care and hospital care for the newborn.Other important aspects to be considered are health professionals' training and placement in childbirth care, evaluation of models of care provided by multidisciplinary teams, sociocultural considerations, such as women's playing a leading role and continuous support for women during childbirth, so as to inform public policies designed to achieve greater reductions in infant mortality.
Intrapartum asphyxia and late prematurity account for approximately 23% of newborn deaths, expressing the avoidability of such deaths and the possibility that implementing recommended practices during labor and delivery will produce greater impact in less time, given that the health services are available.To maximize the reduction in neonatal mortality in Brazil it will be necessary to reinforce public policies with measures more directly related to improving the quality of health care.First, a regionalized perinatal care network must be implemented.Second, investment must be made in implementing scientific evidence-based practices and improving the quality of care processes in prenatal services, and particularly those providing care during labor and childbirth.On the one hand, the perinatal care network must assure pregnant women and their newborns timely access to services at an appropriate level of complexity.On the other, they must be guaranteed access to the best care practices available in state-of-the-art knowledge, and the severe, generalized gap between childbirth care practices in Brazil and scientific evidencebased recommendations must be closed.One example would be the presence of a companion during labor, which is still only incipient in Brazil, although the right has been guaranteed by federal law since 2005.
This study profiled neonatal deaths in Brazil and the main related problems.It indicated that further progress in reducing neonatal mortality and, as a result, infant mortality -as well as maternal deaths and avoidable fetal deaths, given that the related problems in care are similar -will depend on establishing a regionalized, hierarchical, integrated network and on improving the quality of care processes, especially during labor and childbirth.
Mortalidad Infantil; Atención Hospitalaria; Evaluación de Servicios de Salud; Parto Contributors S. Lansky took part in the conception, design, and data analysis and interpretation; drafting of the article, final approval of the version for publication, and was responsible for guaranteeing the accuracy and completeness of the article on all aspects of the study.A. A. L. Friche, A. A. M. Silva and M. L. Carvalho participated in the data analysis and interpretation, drafting the article and critical review of the intellectual content.D. Campos and S. D. A. Bittencourt collaborated on data acquisition, analysis and interpretation, and critical review of the intellectual content.P. G. Frias, R. S. Cavalcante and A. J. L. A. Cunha contributed to critical review of the intellectual content.

Table 1
Distribution of neonatal deaths by selected variables and regions of Brazil, 2011-2012 * .

Table 3
Distribution of live births, neonatal deaths, neonatal mortality rate, crude odds ratio (OR) and chi-square: prior history and current pregnancy -Block 2. Brazil, 2011-2012.
** Weighted rates.Note: the total n may vary by the presence of missing values.95%CI: 95% confidence interval.

Table 4
Distribution of live births, neonatal mortality rate, crude odds ratio (OR) and chi-square: care during childbirth and overall conditions of newborn -Blocks 3 and 4. Brazil, 2011-2012.
*** Excludes women who went into labor.

Table 5
Final model of factors associated with neonatal mortality.Brazil, 2011-2012.