According to our literature review, this study is the most comprehensive investigation regarding factors associated with neonatal mortality in an international border region. Previous studies in the Brazilian border region focused on the causes of infant mortality, including only Brazilian resident mothers  . We opted to include all deaths, without distinction of maternal domicile, because the assistance to foreigners can affect the provision of services to the local population and also impacts the health indicators in the municipality. Additionally, the public health services in Foz do Iguassu impose several barriers to foreigners (e.g. preventing foreign patients from accessing the health services ), which is illegal and instigate the foreigners to falsely affirm that they live in Brazil; thus, the identification of the pregnant who live in Brazil would be imprecise. Moreover, our study focused on neonatal mortality because it is the main component of IMR and is the current challenge to reduce infant mortality .
In our investigation, the occurrence of congenital fetal anomaly presented the strongest statistical association with neonatal death, followed by low birth weight, low first minute Apgar score, zero to 3 prenatal appointments and preterm birth. As expected, our results indicated that the newborn’s characteristics are the most proximately related to neonatal mortality. Of note, all the statistically associated risk factors presented a moderate/strong association with neonatal death, with three of the five risk factors showing an OR above 15 .
There is consensus on the importance of prenatal care to minimize the risk of adverse outcomes for the mother and newborn   . The Parana’s Mother Network (Rede Mãe Paranaense) Program, a strategy implemented in the state of Parana, recommends seven prenatal appointments, preferably starting in the first trimester . In the present study, only 39.5% of the cases performed the minimum number of recommended prenatal appointments, against 69.5% of the controls; however, caution is suggested in interpreting this result, because the low number of prenatal appointments may be associated with pregnancies of less than 30 weeks that would prevent the occurrence of 7 or more appointments. Births before the completion of 30 gestational weeks were more frequent among cases (42%) than among controls (0.3%).
The Apgar score shows the physiological conditions and responsiveness of the newborn to extrauterine life and has been used for more than 50 years as an useful and low-cost tool that remains relevant to contemporary practice in predicting the risk of newborn death and, consequently, to identify those who need additional care in the neonatal and post-neonatal period  . The first minute Apgar score below 7 was statistically associated with neonatal mortality in our study and half of the deaths also presented low Apgar score, corroborating other studies   . Despite being considered an endpoint, the low Apgar score precedes death and can be prevented, especially with adequate attention to pregnant women. We believe that Intrapartum asphyxia demands specific attention during delivery and the prevention of problems related to intrauterine hypoxia and timely interventions may decrease the risk of death due to intrapartum asphyxia in Brazil, considering that almost 100% of births occur in health institutions  .
Prematurity is one of the leading causes of child death in the world  . The gestational age, which was statistically associated with neonatal mortality in this study, is a relevant factor because it is associated with low birth weight, incomplete fetal development, and low Apgar score. The supply of oxygen through the placenta increases with the course of gestation according to the fetal needs  ; a premature birth interrupts this process, exposing the newborn to oxygen deprivation, which can be harmful to the tissues, especially the nervous system , increasing the risk of cerebral palsy, visual disturbances and chronic disease in adulthood . Consequently, a premature newborn increases intensive care expenses, and also affects the family's social, financial, and emotional status. Therefore, investing in prenatal care, focusing on the identification of pregnancies at risk of prematurity may reduce both the financial burden and the social impact of this event.
For this study, all congenital anomalies recorded in SINASC were included according to their code of Chapter 17 of the 10th International Classification of Diseases. The strong association between congenital fetal anomaly and neonatal death observed in this study corroborates the evidence found in other studies  . Congenital anomalies contribute significantly to premature birth, morbidity, and neonatal mortality. Since the improvement of healthful conditions and the reduction of infant deaths due to infectious and parasitic diseases, the anomalies became an important risk factor associated to IMR, evidencing the process of epidemiological transition that most countries experience. According to an a posteriori analysis, most newborns with congenital anomalies were born at normal gestational age (76%) and normal birth weight (78%). Therefore, the only risk factor considered to be unavoidable seems not to be related to prematurity and low birth weight, indicating that it may be possible to reduce neonatal mortality by acting on avoidable risk factors .
Maternal sociodemographic factors can impact childcare, not only about access to health services but also living conditions, basic sanitation, and nutrition . Although previous studies have shown negative effects of maternal age and education on perinatal outcomes  , in our study, those variables were not associated with neonatal mortality. The lack of association between maternal sociodemographic characteristics and neonatal death may be related to the fact that two-thirds of the deaths occurred in the early neonatal period when most mothers and newborns are being assisted in hospitals, suggesting poor healthcare provided in the early postpartum period, regardless of sociodemographic characteristics. At least partially, this poor healthcare provided in the early neonatal period may be related to hospital overload. In Foz do Iguassu, the only hospital with a Neonatal Intensive Care Unit (NICU) is a reference for maternity to 9 municipalities within western Parana; additionally, it receives pregnant women and newborns from neighboring countries. Nevertheless, at the end of this research, the hospital had only 18 intensive or intermediate care unit beds, which is 25% below the established by the Brazilian Ministry of Health.
In an a posteriori investigation, we found that the rates of prematurity, low birth-weight, low first minute Apgar score, and congenital anomaly incidence are not worse than those found in Parana state and Brazil. In contrast, Foz do Iguassu presents twice the rate of unsatisfactory prenatal appointments in comparison to Parana state and Brazil. Thus, the high rates of neonatal deaths in this region possibly translate the poor quality of the local health services.
Although some variables of the proximal level, such as Apgar score, birth-weight, and prematurity, are considered endpoints, we understand that they precede death and reflect distal and intermediate level variables, being relevant in the model constructed for this study. Besides, they are risk factors for neonatal death already established in the literature. With this in mind, we proposed the analysis of these risk factors through the conceptual framework, which analyzes the effect of each variable on neonatal death hierarchically mediated by the other variables.
We acknowledge some limitations in this research. Firstly, as every study that relies on secondary data analysis, missing or incorrect information is an inherent risk. Also, the official databases used in our research do not provide information on maternal diseases, which may also be associated with neonatal mortality. Regarding the strengths of the study, to improve the reliability of our data, we performed an a priori data completeness analysis for the period from 1996 to 2016. Considering the high frequency of missing data before 2012 and the modifications in the data gathered through SINASC in 2011, we analyzed only births occurred from 2012 to 2016. Besides, the choice of analysis through the conceptual framework allowed us to model different factors according to their precedence over time and their relevance to the outcome determination; moreover, this analysis model represents a strategy for dealing with a large number of conceptually related variables present in epidemiological studies.