Age-specific neonatal mortality and associated factors in the 2021 state of Rio de Janeiro (

Objective: To analyze the causes of age-specific neonatal deaths and death-associated factors in the 2021 state of Rio de Janeiro birth cohort. Methods: Retrospective cohort of live births (LB) followed up to 27 days of delivery (<24hs, 1–6 and 7–27 days). Data obtained from the Information Systems on Live Births (2021) and Mortality (2021/2022). We described the distributions of maternal and newborn characteristics and causes of death. We used multinomial regression models with hierarchical levels of determination of neonatal death. Results: Of the 179,837 LB, 274 died within 24 hours, 447 within 1-6 days and 324 within 7-27 days. The neonatal mortality rate was 5.8‰ LB (CI 95%: 5.5–6.2). Neonatal survivors and deaths were heterogeneous according to the analyzed characteristics, except for the reproductive history (p<0,05). 78% of causes of death were avoidable. Causes reducible by adequate care for pregnant women (<24 hours and 1-6 days) and newborns (7-27 days) predominated. Low schooling showed a significant association for deaths between 7-27 days (OR ajusted =1.3); mixed race, for deaths between 1-6 days (OR ajusted =1.3), and black color for both age groups (1-6 days: OR ajusted =1.5 and 7-27 days: OR ajusted =1.8). Health care and biological factors of LB (intermediate and proximal levels) remained strongly associated with neonatal death, regardless of age. Conclusion: Causes of death, factors associated with neonatal death, and strength of association differed according to death-specific age. Preventive actions for neonatal death should consider sociodemographic vulnerabilities and intensify adequate prenatal and perinatal care.


INTRODUCTION
The first month of life is the most vulnerable period for a child. In 2020, in the world, almost half of the deaths of children occurred in the neonatal period -the first 27 full days 1 . In Brazil, the neonatal mortality rate (NMR) dropped from 25.0 (1990) to 9.0 per thousand live births (LB) in 2020, an annual reduction of 3.6% 1 -less intense than for total mortality in childhood, 4.9% 1 . Of the total neonatal deaths (ND) in the country, 75% are early ones 2 .
Regarding the trend of neonatal mortality in Brazil, the NMR was 7.9/1,000 LB in the Southeast Region in 2017, and the highest NMR in the region was in the state of Rio de Janeiro (RJ), 8.6/1,000 LB. NMR's trend in the state was downward, with an annual reduction rate of 1.6% 3 .
Among infant deaths, the highest risk of death occurs on the first day of life 4 . About 2/3 of ND occur on the first day, and about 60% occur with up to three days of life in developing countries 5 . In Brazilian states, including Rio de Janeiro, mortality within 24 hours of delivery concentrated 1/5 of infant deaths from 2010 to 2015 4 .
The causes of neonatal mortality in the country, especially in the first week of life, are related to prenatal care, childbirth, and newborn care, and are considered, for the most part, to be avoidable 4,6,7 . Causes not clearly avoidable, such as congenital anomalies, showed growth in both the proportion of total deaths and the cause-specific mortality rate from 2007 to 2017 3 .
As for the factors associated with ND, a meta-analysis of Brazilian studies identified variables of sociodemographic aspects (age ≥35 years, absence of a partner), assistance (inadequate prenatal care, maternal morbidity, cesarean delivery), and related to the newborn (low weight, prematurity, asphyxia, and congenital malformation) 8 . Other studies corroborate these factors [9][10][11][12][13][14] and appended maternal education 9,14 and fetal presentation at delivery 12 .
Most studies analyze the factors associated with total ND 9,10,12,14 either by early and late components 13 , on the first day 4,15 or up to the first day, and from one to 27 days 16 , without considering all these moments of occurrence and the potential age-specific differentiated behavior.
This study analyzed the magnitude of ND, causes of death and associated factors, by age-specific, in the 2021 state of RJ birth cohort.

METHODS
It is a retrospective 2021 state of Rio de Janeiro (Brazil) birth cohort. The follow-up period is from birth to 27 completed days of life. The outcome is age-specific neonatal death (<24 hours, one to six days, seven to 27 days) occurring in 2021 and 2022.
The data sources were the Information Systems on Live Births (Sistema de Informações sobre Nascidos Vivos -SINASC) and Mortality (Sistema de Informações sobre Mor-talidade -SIM). Data were obtained from the Health Department of the state of Rio de Janeiro without nominal and residential identification.
Deaths were described according to investigation and the underlying causes of death (10 th Revision of the International Classification of Diseases -ICD-10 18 ) and preventability (Brazilian List of Avoidable Causes of Death/Lista Brasileira de Causas de Morte Evitáveis -LBE <5 years 19 ).
To describe the mention of COVID-19 (any line of the Death Certificate -DC, Part I or Part II), the recommendation of the Ministry of Health 20 was followed: ICD B34.2, infections caused by a coronavirus, along with one of the COVID-19 markers (U07.1 -laboratory-confirmed case) or U07.2 (probable or suspected case -clinical, epidemiological history with inconclusive or unavailable tests) on the same line of the DC.

Statistical analysis
Losses due to the non-linkage of the bases were compared with the NDs related to SINASC. The data source to test the homogeneity of the non-linkage losses was the SIM www.scielo.br/rbepid Age-specific neonatal mortality and associated factors. Rev Bras Epidemiol. 2022; 25:e220038 3 https://doi.org/10.1590/1980-549720220038 (maternal education and age, gender, weight, gestational age, and age of death). The NMR (quotient between the number of neonatal deaths by the number of LBs in 2021) per thousand LBs and the confidence interval (95%CI) were calculated. The absolute and percentage distributions of the variables studied were described for neonatal survivors and age at death. As gestational age information didn't compose the prenatal inadequacy variable, this variable was additionally analyzed according to prematurity (<37 s: yes/no). To test the homogeneity of the loss versus linked NDs and the survival versus the death of the study population, the Pearson's χ² and Fisher's exact tests (statistical significance level of 5%) were used.
To investigate the factors associated with ND, simple nominal (significance level <0.2) and multiple (significance level <0.5) logistic regression models were used. The strategy of the theoretical model of hierarchical determination of the ND 21 was applied, adapted according to available data from the information systems. The distal hierarchical level was represented by maternal education and color; intermediate level I, by maternal age group and parity; intermediate II, due to fetal presentation and adequacy of prenatal care; and, at the proximal level, by characteristics of the newborn (weight, gestational age, and Apgar at the 5 th minute) except for the gender variable (not belonging to the hierarchical levels), approached independently, and considered without adjustments (Supplementary Figure 1). Covariates from the same and previous hierarchical levels were considered as potential confounders. The computer program Stata SE (version 12) was used.

RESULTS
SIM records not linked to SINASC corresponded to 12% of deaths (194 losses out of 1,618 ND). When applying the eligibility criteria, this percentage reduced to 9.9% (109 losses out of 1,099 ND). The losses were not selective regarding age at death and maternal and neonatal characteristics. In the cohort of 179,837 LB, 274 ND occurred within 24 hours, 447 from one to six days and 324 from seven to 27 days (Supplementary Figure 2). Neonatal mortality rate was 5.8‰ LB (95%CI 5.5-6.2).
Survivals and deaths in the neonatal period were heterogeneous, except for reproductive history ( Table 1). The proportions of teenage mothers (<20 years) and those aged ≥35 years were more frequent among deaths than survivors. Among the survivors, mothers with high education, white, with prenatal care, and a fetus in the cephalic presentation were more prevalent. The neonates who died were predominantly male, with a higher frequency of low birth weight, prematurity, and asphyxia in the 5 th minute (Table 1).
In the analysis of deaths by age ( Table 1), those that occurred within 24 hours had the highest proportion of inadequate prenatal care and non-cephalic presentation and the lowest proportion of vaginal deliveries. The inadequacy of prenatal care was highlighted for deaths within 24 hours in premature neonates (35.4%), 40% higher than in non-premature neonates (25.3%). Regarding weight, gestational age, and Apgar at the 5 th minute, a more disadvantageous situation was observed than in the other deaths (Table 1). Deaths from one to six days, the most numerous, showed a high proportion of cesarean sections and a higher frequency of males. Late ND showed a higher frequency of mothers in the extremes of age, with low education and black. A decreasing gradient in the proportion of mothers with 12 years and more of education and an increasing gradient in the proportion of maternal black color were observed between survivors and deaths and, among deaths, with increasing age.
Of the deaths, 70.4% were investigated (66% of ND<24 hours and 72% of the others ND), and 74% were classified as avoidable. The causes reducible by adequate care for the pregnant woman (first position) and adequate care for the newborn (second position) were highlighted ( Table 2). Some differences were observed in the analysis of avoidability by age: among the ND<24 hours, the group of other causes (not clearly avoidable, which include most congenital malformations) occupies the second position, followed by causes reducible by adequate attention to the childbirth. The 1-6 days ND follow the pattern of the total: higher percentage for causes reducible by adequate care for pregnant women. There is an inversion of this pattern to the late deaths: the first cause is adequate attention to the newborn. Deaths classified in the groups of causes reducible by adequate diagnostic and treatment and health promotion actions are much less frequent and restricted to ages 1-6 and 7-27 days. Ill-defined causes are rare at all ages, and there were no deaths from vaccine-avoidable causes ( Table 2).
The most frequent grouping of cause of death was specific infections in the perinatal period (ICD P35 to P39.9), more pronounced for the 1-6 and 7-27 days groups (Table 2). In this group, the specific cause of unspecified bacterial septicemia of the newborn (ICD P36.9) stood out with 121 cases, first among the specific causes, except for ND<24 hours, among which other congenital malformations prevailed (ICD Q89) (Supplementary Table 1). In second place came maternal conditions (CID P00) ( Table 2). For ND<24 hours, the first cause was intrauterine hypoxia and birth asphyxia (ICD P20 and P21). Three late neonatal deaths had a mention of COVID-19 (CID B34.2 and U07.1 or U07.2 in the same line of the DC), with the underlying causes being coronavirus infection of unspecified location (CID B34.2) and asphyxia at birth, unspecified (ICD P21.9).

DISCUSSION
In the 2021 birth cohort in Rio de Janeiro, for every thousand LB, about six died in the neonatal period, or a little more than ¼ of deaths in less than one day. Causes of death, associated factors, and strength of association differed according to age of death, confirming the importance  Maternal vulnerabilities, such as low education and black and mixed race, were relevant risk factors for deaths from the first day of life. Care variables, such as inadequate prenatal care and non-cephalic presentation, as well as newborn characteristics, such as weight, gestational age, and Apgar at the 5 th minute, were strongly associated with ND, regardless of age. Neonatal mortality rates in cohort studies with longitudinal data, are close to the values obtained with cross-sectional data (deaths and LB in the same year) due to the short follow-up period (up to 27 days). The NMR of the present study and that calculated with the cross-sectional data available on the website of the State Health Department of RJ were very close. NMR estimates were not found in national cohort studies for the current decade. In the previous decade, lower NMR values were estimated in cohorts of LB from Florianópolis city, state of Santa Catarina, from 2012 to 2014 (5.4‰ LB) 10 and from the Rio de Janeiro city, state of Rio de Janeiro, in 2015 (5.9‰ LB) 12 . On the other hand, in Goiânia city, state of Goiás, in 2012 (9.4‰ LB) 13 , in the state of São Paulo, from 2004 to 2013 (from 9.1 to 7.4‰ LB) 14 , and in Pelotas city, state of Rio Grande do Sul, in 2015 (8.7 ‰ LB) 11 , the values were higher than those of the present study (5.8 ‰ LB), even considering the deaths not related in the present study (n=1,090), which would increase NMR to 6.4 per thousand LB. It should be highlighted that there has been a decline in neonatal mortality in the country, and probably the values of the current decade for the mentioned places have reduced, mainly in the southern region of the country, whose NMR in 2018 was estimated at 7.2‰ LB 2 .
The group of causes reducible by adequate pregnancy care was the most frequent, corroborating the national study by Prezotto et al. 22 . The order of magnitude of reducible causes of death in Brazil was the same as the one found in the current investigation for the state of RJ. The observed lower adequacy of prenatal care among pregnant women whose newborns died supported this finding. It is noteworthy that, in the mentioned research, the avoidable NMR of the state of Rio de Janeiro was 6.5 per thousand LB, the highest in the Southeast 22 .
There was a difference in the pattern of causes and avoidability according to the neonatal age range. The closer to birth, in addition to maternal causes, the more lethal congenital malformations and asphyxia had an impact. The later the death, causes reducible by attention to the newborn, especially the specific perinatal infection, had considerable importance. However, the high frequency of diagnosis of unspecified bacterial septicemia in the newborn deserves reflection (ICD P36.9), which does not con-tribute to the knowledge of the causes of death. This condition was responsible for about 80% of infant deaths from sepsis in Brazil 23 , of which 80% were preterm and/or had low birth weight. Therefore, unspecified sepsis is hardly the underlying cause and can be considered a garbage code 24 . Other debatable diagnoses as the underlying cause are those related to intrauterine hypoxia (ICD P20) and birth asphyxia (ICD P21), attributed in the LBE classification to childbirth care but which may result from maternal causes not detected during prenatal care 25 .
Despite not being among the most frequent causes of neonatal deaths in the studied cohort, two infectious diseases deserve to be highlighted as a public health problem: congenital syphilis (ICD A50), neglected and endemic disease (0.7; 2.9 and 1.9%, respectively, of deaths <24 hours, 1-6 days, and 7-27 days) (Supplementary Table 1), and the before mentioned cause COVID-19, emerging and pandemic disease (data not shown in table). An increasing trend of congenital syphilis from 2001 to 2017 in the state of Rio de Janeiro, the highest in the Southeast Region in 2017, was observed even considering prenatal care, which therefore reflects failures in maternal and child care 26 .
In a meta-analysis on the impact of COVID-19 on maternal and fetus health, an increase in prematurity and low birth weight was observed, but the risk of neonatal death was considered low 27 . In the 2021 state of RJ birth cohort of 324 late NDs, two cases of COVID-19 were considered suspect and one, confirmed.
As for the associated factors, the differences according to the specific age of death occurred only in the distal and intermediate I hierarchical levels and with the gender variable. None of these factors was associated with death <24 hours, and maternal education and age were associated only with the late component of ND. Even using the hierarchical modelling analysis, which favors the maintenance of distal determinants along the causal chain, the distal factors associated with neonatal death in other studies are not consensual. In national LB cohorts, factors such as schooling, color, and marital status in Goiânia (GO) 13 , color and schooling in Pelotas (RS) 11 , and schooling and marital status in Florianópolis (SC) 10 were not associated with neonatal mortality due to early and late 13 and total 10,11 components. In the LB cohort from the city of Rio de Janeiro (RJ), education and color (distal level) were associated with total neonatal death 12 . In a recent meta-analysis, the authors concluded that for each additional year of maternal education, there is a 1.5% reduction in total neonatal mortality 28 . Care and newborn factors, closer to the outcome, regardless of age at death, were strongly associated. Four or more prenatal visits reduce the chance of dying on the first day of life by 30% in countries in Latin America, the Caribbean, Africa, and Asia 29 . In 2015, the state of Rio de Janeiro had the highest proportion of inadequate prenatal care (29.5%) when compared to other states in the https://doi.org/10.1590/1980-549720220038 Southeast 2 . In the present study, using the same indicator of adequacy of access to prenatal care, the proportion of inadequacy was high but lower, around 25%; however, when comparing deaths with survivors, the adjusted odds ratios (OR) showed a decreasing gradient with increasing age (6.3, 4.5, and 4.0, respectively, for ND<24 hours, 1-6 and 7-27 days). The same gradient of the strength of association with age occurred with the non-cephalic presentation, which makes up the same intermediate level of prenatal care, in agreement with the results of two national studies 12,30 . The type of delivery variable was not incorporated in our model because of the ambiguity of the risk of death or iatrogenic cesarean sections 14 . Strong associations of proximal factors determining neonatal death, weight, gestational age, and Apgar at the 5 th min are results widely documented [10][11][12][13]15,31,32 . Some of these studies incorporated the gender of the LB at the proximal level of the model 11 . We found it inappropriate because the variables of the previous hierarchical levels don't determine the gender of the newborn.
As positive points of the study, we highlight some methodological aspects. The quality of the SINASC and SIM, as a function of the completeness of the SLB number field in the DC, the linkage key (loss of 9.9%), allowed to reconstitute the 2021 state of RJ birth cohort and as well as the completeness of most of the other variables analyzed, to investigate neonatal mortality and associated factors. We found appropriate the use of the number of vaginal deliveries and the number of previous cesarean deliveries variables to represent parity. Despite the alternative to the definition of parity, based on the information in the database, which would consider both the variable alive children and deceased children/miscarriages, this second variable had a high incompleteness (22%) and included miscarriage in its measurement. Although the reliability of the information was not evaluated in the study, there are publications on the improvement of national 33 and state 34 information contained in the SLB and DC. As for the underlying cause of death, according to the LBE, the frequency of ill-defined causes was low, probably due to the investigation of infant death. In the analyzed cohort, about 70% of the ND were investigated, with the lowest percentage of death investigation for the ND<24 hours (66.1%), which concentrates newborns below 2,500 g, not a priority in the epidemiological surveillance of infant death. However, we questioned some causes of garbage codes. Thus, efforts are imperative to improve the filling out of the SLB and DC.
Another positive aspect of the study was the methods used. The strategy of hierarchical analysis of the determination of the ND allows for preserving the strength of the associations of social determinants (distal level), even after incorporating variables at the more proximal levels of the model 35 . In the multinomial regression model, the effects of the regressors (OR) are adjusted by the same covariates for each specific age of neonatal death (outcome) analyzed, allowing a direct comparison of the results 11,31 .
Neonatal mortality in the state of RJ in 2021 was 5.8‰ LB, concentrated in the first 24 hours and 1-6 days. NMR was lower than in previous years, but mostly from avoidable causes, resulting mainly from adequate care for pregnant women. The analysis of associated factors showed inequalities regarding maternal age, schooling, and color, in addition to reinforcing the biological determinants of the newborn. Investments in attention to the mother-child binomial are necessary, especially in prenatal care and in the most vulnerable women.