Fatores de risco para óbito neonatal obtidos pelo modelo de regressão multivariado de Cox

OBJETIVO: Identificar los factores asociados al obito neonatal en Sao Jose dos Campos. METODOS: Se trata de estudio epistemologico de tipo longi-tudinal para el que se construyo una base de datos con informa-ciones del recien-nacido y de la madre, obtenidos a partir de las Declaraciones de Nacido Vivo y de obito de neonatos y de madres residentes en Sao Jose dos Campos (Sao Paulo, Brasil), en los anos de 2003 y 2004. Se definio como variable dependiente el obito hasta el 28o dia despues del nacimiento. El analisis multivaria-do, utilizando el modelo de Cox, fue aplicado para verificar la asociacion entre el obito neonatal y las siguientes variables inde-pendientes: sexo, peso al nacer, duracion de la gestacion, Apgar en el 1er y 3er minuto, edad materna, nivel de instruccion de la madre, numero de obitos fetales y de hijos vivos previo, tipo de parto y gestacion unica o multiple. El nivel de significancia adoptado fue p<0,05. RESULTADOS: En el periodo analizado se identificaron 131 obitos neonatales. Edad gestacional menor que 37 semanas (RR 6,92; IC 95% 3,64-13,17), Apgar en el 5o minuto <7 (RR 3,14; IC 95% 1,95-5,04), Apgar en el 1er minuto <7 (RR 3,48; IC9%: 2,17-5,60) y bajo peso al nacimiento (RR 4,49; IC 95%: 3,36-8,53) fueron las variables asociadas significativamente con la muerte, en el modelo final. CONCLUSIONES: Las variables asociadas al obito neonatal en Sao Jose dos Campos (Sao Paulo, Brasil), estan relacionadas a la calidad de la asistencia pre-natal y perinatal.


Introduction
Infant mortality may be considered one of the best indicators of healthcare quality and of the socioeconomic level of a population (1,2) .The concept of infant mortality encompasses all deaths occurring in the first year of life, and can be broken down into two components: neonatal mortality, which includes all deaths occurring before the 28th day of extrauterine life, and postneonatal mortality, which includes all deaths occurring between the 28th day of life and the day before the child's first birthday (1) .The determinants of neonatal mortality are manifold and complex, and are associated with the interaction of biological, socioeconomic, and care-related variables (2)(3)(4) .
According to data obtained by Fundação Seade (5) in 2009, the average infant mortality rate for the state of São Paulo is one of the lowest in Brazil, at 12.5 deaths per 1000 live births (LB).Within the catchment area of the 17th Regional Health Department (DRS XVII), which includes 35 municipalities in the São Paulo portion of the Paraíba River valley, the 2009 infant mortality rate was 13.3 deaths/1000 LB.In the municipality of São José dos Campos, where the present study was conducted, the average rate was 10.8 deaths/1000 LB, lower than the averages of both the DRS XVII and the state of São Paulo.
Despite a downward trend, current rates show that neonatal deaths account for most of the infant mortality, and are still significantly high compared to the lower postneonatal mortality rates.This finding is confirmed by the 2009 Fundação Seade data, which show neonatal mortality rates of 8.7/1000 LB in the state of São Paulo, 10.0/1000 LB within the DRS XVII, and 7.9/1000 LB in São José dos Campos (5) .
Any attempt to improve these rates will require identification of risk factors for neonatal mortality in the municipality of São José dos Campos, which will provide an understanding of possible causes and enable implementation of preventive actions.This process entails investing on higher-complexity, highertechnology hospital services and on educational and public health actions, geared particularly to improvement of neonatal care (1) .Countless study designs have been used for evaluation of neonatal care (6,7) .The approach proposed herein is the Cox model, which applies multivariate regression to survival analysis in order to determine the survival time of subjects after exposure to a certain variable believed to be a risk factor.The present study thus sought to identify risk factors for neonatal death among all infants born alive to mothers living in the municipality of São  (8) , which provides access to Certificates of Live Birth and Death Certificates.
The study population was established by identification of all neonatal deaths (occurring within 28 days of birth) of infants born to mothers living in the municipality of São José dos Campos, São Paulo, between January 1, 2003 and December 31,  2004, by means of the record linkage approach.Records were paired with those of all infants born alive on the date of birth of the deceased infants, thus compiling a database of all infants who died before the 28th day of life and all those who survived past this point during the study period (4) .The final database comprised 131 decedents and 2847 neonates who survived beyond the 28th day of life, for a 21:1 ratio of survivors to decedents.Infants who were born before December 31, 2004 and died before January 27, 2005 were also included in the sample.
The outcome variable was neonatal death.The independent variables of interest included maternal and neonatal information.Due to their quantitative nature, these variables were categorized according to classification schemes recommended in the literature, thus becoming qualitative.The following variables were defined, with their respective categories: birth weight, low (<2500g) or normal (≥2500g); duration of pregnancy, preterm (gestational age <37 weeks) or full-term (gestational age ≥37 weeks); Apgar scores at 1 and 5 minutes, low (<7) or adequate (≥7); maternal age, suboptimal (<20 or >35 years) or optimal (≥20 and ≤35 years); maternal educational achievement, low (<8 years' formal education) or high (≥8 years); number of stillborn and number of live offspring, none or ≥1; mode of delivery, vaginal or Cesarean; type or pregnancy, singleton or multiple.
Univariate analysis was performed for screening, and variables with p<0.20 were included in the multivariate Cox proportional hazard regression model.Analyses were performed in the Stata 9 software environment.The Cox model yields hazard ratios (HR) with a 95% confidence interval, establishing the likelihood of neonatal death according to presence or absence of the condition of interest (9) .The final model thus enables assessment of possible associations between the study variables and neonatal death.Cox-Snell residuals can also be plotted for estimation of the fit of the model.

Results
In the years 2003 and 2004, a total of 18,013 infants were born alive to mothers living in the municipality of São José dos Campos, São Paulo.Of these, 131 died before the 28th day of extrauterine life, which corresponds to a neonatal mortality rate of 7.3 deaths/1000 LB.A total of 2,978 infants born on the same day of those who died survived the neonatal period; of these, 396 were randomly paired to each death record to serve as controls.
Univariate analysis of these variables yielded p-values <0.20 for mode of delivery, duration of pregnancy, birth weight, Apgar scores at 1 and 5 minutes, and gender.Kaplan-Meier curves plotted for these variables were parallel and did not cross during the study period, confirming statistical significance (unpublished data).
These variables were then analyzed by Cox multivariate regression analysis, which revealed that the variables low birth weight, 1-minute Apgar score <7, 5-minute Apgar score <7, and preterm birth were risk factors for neonatal mortality in infants born in São José dos Campos between 2003 and 2004 (Table 3).The Cox-Snell plot indicated good model fit, with residuals nearly coinciding with the ideal curve (Figure 1).

Discussion
This study, performed on data obtained for the municipality of São José dos Campos, used Cox multivariate regression analysis, a statistical approach rarely used for analysis of neonatal mortality.
The results obtained showed higher mortality rates among male infants.However, according to the Cox model, this variable did not reach statistical significance.This finding is consistent with other published studies, which always take into account higher male than female mortality rates, usually regardless of age (2,3,(10)(11)(12) .The biological mechanisms that underlie this gender divergence in mortality rates have yet to be fully elucidated, but some authors suggest that less maturity may play a role, with a higher incidence of respiratory syndromes in male neonates (11,12) .Some other variables also did not reach statistical significance as risk factors for neonatal death, but are epidemiologically important for outlining a profile of neonatal mortality in São José dos Campos, due to their care-related and socioeconomic nature: maternal educational achievement, pregnancy, and mode of delivery.
In the present study, maternal age was not significantly associated with neonatal death, but it bears noting that survival was lower in neonates whose mothers were not in the optimal age range (44% of the sample).Cesar et al (13) found this lower survival rate to be mostly due to low maternal socioeconomic status, which interferes directly with the manner in which the mother interprets and follows her pregnancy.The authors also stress the importance of cultural practices concerning care of the pregnant woman and newborn, as other studies have shown that proper prenatal care and adequate labor and delivery care have a direct influence on neonatal mortality rates, particularly where birth weight and prematurity are concerned (14) .
Educational achievement has been singled out as an important socioeconomic variable that interferes with the biological outcomes of pregnancy-such as low birth weight-and with infant mortality.Studies have shown that literacy increases awareness of and sensitivity to health education actions; rising educational achievement levels may thus act as protective factors by increasing assimilation of information on the various healthcare alternatives available to the population.This, in turn, leads to healthy personal health practices and behaviors (3) .Although this trend has been confirmed, it did not afford significant protection against neonatal mortality in the present study.
The risk of neonatal mortality in our sample was relatively higher after vaginal delivery then after Cesarean section, although this difference was again not significant.This finding is consistent with other studies that have identified Cesarean birth as a protective factor against neonatal mortality, whereas vaginal delivery may help reduce maternal morbidity and mortality (15,16) .A study conducted in Goiânia, state of Goiás, found a positive association between vaginal delivery and neonatal mortality on univariate analysis.This association was confirmed by logistic regression, but was later found to be due to selection bias-namely, a higher frequency of vaginal delivery in settings associated with high risk of neonatal death, versus near-universal performance of Cesarean section in low-risk settings (17) .Araújo et al (5) reported the opposite, with higher risk in mothers undergoing Cesarean section.
The association between neonatal mortality and prematurity was also stronger in this study: 78% of deaths occurred in preterm infants.This variable was statistically significant on both univariate and multivariate analysis, which attests to its importance as a determinant of neonatal mortality and contributing factor to low birth weight and to 5-minute Apgar scores <7.
Apgar scores were also associated with neonatal mortality in the present study.The accuracy of 1-minute Apgar as a diagnostic test or marker for asphyxia is questionable, as fewer than half of neonates with now Apgar scores at 1 minute are actually hypoxic on arterial blood gas studies.The sensitivity and specificity of 1-minute Apgar for detection of asphyxia range from 11-78% and 75-99% respectively, whereas the 5-minute Apgar score is more closely associated with the prognosis of infant survival and neurological function (18) .Pereira et al (18) note that roughly 840,000 of the 4 million neonates who develop moderate-to-severe asphyxia each year will die, with the same number surviving, but with sequelae.Overall asphyxia-related neonatal mortality may thus be estimated to range between 15% and 40%.
One-minute Apgar scores are rarely used as a study variable; in the present study, however, they affected the fit of the whole model, particularly on residuals analysis.Neonatal mortality studies have found that, the lower the 5-minute Apgar score, the lower the odds of survival (5) .However, it has been established that Apgar scores near 6 may occur in otherwise healthy preterm infants with poor muscle tone due to high-risk pregnancy,  Cesarean section or complications during delivery (18) .This finding was confirmed in the present study; survival was found to be lower in neonates with low Apgar scores, with the mortality rate reaching 40% when 5-minute Apgar scores were <7.Low birth weight and prematurity are universally recognized as the single most important risk factors for neonatal death (15) .As was to be expected, this study found a significant association between neonatal mortality and birth weight, with 30% of low birth weight infants dying before the 28th day of life.The prevalence of low birth weight (<2500g) has been associated with unfavorable socioeconomic conditions, inadequate prenatal care, suboptimal maternal age, high parity, reduced birth interval, maternal malnutrition, and maternal smoking (5) .Araujo et al (10) believe low birth weight is associated with disadvantaged socioeconomic status, which renders neonates more vulnerable to unfavorable conditions such as preterm birth.Moreover, the frequency of low birth weight is clearly higher in multiparous women, usually due to shorter birth interval.This association is one possible explanation for the findings of this study, such as increased risk of mortality among infants born to multiple pregnancies.
In the present study, neonatal death was strongly associated with low birth weight, preterm birth, and 5-minute Apgar scores <7.These findings are consistent with those reported by other authors who have conducted studies focusing on risk factors for neonatal mortality (19)(20)(21) , although the association between 1-minute Apgar scores and mortality has only rarely been researched.Our results thus confirm those of other authors who have pinpointed prematurity and low birth weight as universal risk factors for neonatal mortality.Furthermore, Apgar scores are closely related to other variables and can define prognosis in the first minutes of extrauterine life.It bears stressing that the above-listed factors should not be viewed in isolation, but rather as determinants of neonatal mortality.
The risk factors for neonatal death identified in the present study can be minimized, or even altogether avoided, if adequate measures are implemented to encourage prenatal care, improve labor and delivery, and ensure immediate postpartum care of the newborn.These actions would depend on human resources and equipment, but the key goal should be allowing pregnancy to occur and progress in the best possible biological, physical, and psychological setting.Furthermore, measures should be implemented to reduce social inequality, as the most significant risk factors for neonatal death are closely tied to low maternal socioeconomic status.
José dos Campos, state of São Paulo, between 2003 and 2004, by means of Cox survival analysis.This longitudinal epidemiological study was approved by the relevant Research Ethics Committee.The data on which the study is based (2003-2004 data for the municipality of São José dos Campos) were obtained in late 2006 from the Health Information System (Sistema de Informações à Saúde) available on the São Paulo State Department of Health website methods

table 1 -
Maternal characteristics of liveborn infants who did or did not survive the neonatal period in São José dos Campos, SãoPaulo, 2003-2004.

table 2 -
Neonatal characteristics of liveborn infants who did or did not survive the neonatal period in São José dos Campos, SãoPaulo, 2003-2004.

table 3 -
Cox multivariate regression analysis of factors associated with neonatal death, expressed as relative risk (RR), 95% confidence intervals (95%CI) and p-values