Artigos

Conteúdo (article):

Estimation of survival and geographic trajectory in antepartum of premature infants with neonatal death outcome

Carolina Luiza Bezerra Silva Webster Barbosa. Universidade Federal de Pernambuco. Recife PE Brazil. carolinaluizaa@gmail.com. ORCID: 0000-0002-9080-4285
Eliane Rolim de Holanda. Universidade Federal da Paraíba. João Pessoa PB Brazil. eliane.rolim@academico.ufpb.br. ORCID: 0000-0001-6433-9271
Luciana Pedrosa Leal. Universidade Federal de Pernambuco. Recife PE Brazil. luciana.leal@ufpe.br. ORCID: 0000-0003-3776-0997
Ana Paula Esmeraldo Lima. Universidade Federal de Pernambuco. Recife PE Brazil. ana.plima@ufpe.br. ORCID: 0000-0002-8447-4072
Amanda Priscila de Santana Cabral Silva. Universidade Federal de Pernambuco. Vitória de Santo Antão PE Brazil. amanda.cabral@ufpe.br. ORCID: 0000-0003-2337-9925
Vânia Pinheiro Ramos. Universidade Federal de Pernambuco. Recife PE Brazil. vania.ramos@ufpe.br. ORCID: 0000-0002-4559-934X

Abstract
This study aimed to analyze the estimated survival of premature infants with neonatal death outcome and geographic trajectory followed during antepartum between the municipalities of the state of Pernambuco. This is a cohort study with survival and ecological analysis, through flow analysis, with 4,643 premature babies who did not survive the neonatal period between 2013-2019. Multivariate analysis was used using a generalized linear model. To analyze the geographic trajectory followed in the antepartum period, we used the spatial pattern of flows using the K-means method. Survival of premature babies was 2 days and 34% (n=1,578) lived less than 1 day. Survival time decreased by 16% when mothers’ race/color was brown. In relation to the spatial pattern of flows between residence and place of birth, routes with distances between 0 km and 61.3 km and between 61.3 km and 163.5 km accounted for 74.7% (n=2,449) and 17.5% (n=572) of deaths, respectively. There was divergence in survival times in the different mesoregions of the state. Many women still needed to travel long distances for the birth of their children, which was associated with shorter neonatal survival times, especially in areas further away from the Metropolitan Region.
Keywords: Survival Analysis, Health Services Accessibility, Infant, Premature, Early Neonatal Mortality.

INTRODUCTION
Neonatal mortality, an important component of infant mortality in Brazil and around the world, is considered a sensitive indicator of quality of life, socioeconomic development and access to health and services for a given population. Despite its significant decrease in recent decades, Brazil still presents a disparity in this indicator between different regions of the country1.
In Pernambuco, despite the expansion of important public policies to combat the high rates of neonatal mortality in the state, such as the Programa Mãe Coruja Pernambucana (PMCP), there are still obstacles to be overcome, such as inequalities related to the socioeconomic development of the regions and access to health services2.
Neonatal mortality predictors are multiple and complex, with prematurity being the preponderant factor associated with death during this period. Every year, a mean of 30,000 premature babies is born and, of these, 10% have a gestational age (GA) ≤ 32 weeks or birth weight (BW) ≤ 1,500 grams, characteristics that limit neonatal survival time3. Despite advances in care for premature babies and increased survival, mortality and morbidity rates remain high, particularly at lower GA4.
There are many factors influencing neonatal survival described in the scientific literature, and it is important to highlight the Social Determinants of Health (SDH) associated with this process. SDH are defined as the social characteristics in which life takes place, such as factors linked to income, education, living and working conditions, food availability and access to health services, such as prenatal care5.
Access to health services, considered an intermediate SDH, is an important indicator of maternal and neonatal health care and strategic organization of health services and policies, especially in developing countries6. The regionalization of birth care must be structured with the objective of reducing mortality in these groups, through decentralization of health services. By adopting regional health planning, all levels of care have a collaborative relationship to improve the quality of health services within a region7.
Despite the Brazilian legal guarantee of prior linkage to maternity and dignified and quality perinatal care, the difficulty in accessing postpartum women still persists, whether due to a lack of coherent regionalization, a reduced number of obstetric and neonatal beds or a failure in the referral and counter-referral system. Although there are regional differences, the difficulty in accessing maternity wards in Brazil was present in 16.2% of all postpartum women in the last national survey carried out in 20128.
The unfavorable trajectory, when there is difficulty in access, transportation, and when long distances need to be covered in search of effective assistance, is worse for pregnant women who do not live in capital cities. A woman’s pilgrimage in search of birth assistance can reach almost 300 kilometers, and she often visits more than one health service before admission9,10.
When there is a failure in the regionalization process, labor and birth care services become disorganized, leading to births occurring in a location other than the competent health region1. In this context, women who are surprised by an early birth are even more vulnerable to this disorder of referral to specialized health services5. In this regard this research aimed to analyze the survival estimate of premature newborns (NBs) with neonatal death outcome and geographic trajectory followed during antepartum between the municipalities of the state of Pernambuco.

METHODS
This is an epidemiological study, carried out in two stages: the first corresponds to a retrospective cohort study with survival analysis; and the second corresponds to an ecological approach study using spatial analysis of the flow pattern. It was carried out in the state of Pernambuco, located in northeastern Brazil, and the unit of analysis was the 184 municipalities in Pernambuco. The state of Pernambuco occupies an area of 98,967,877 km² and has an estimated population of 9,058,155. It is divided into 12 Health Regions, which are organized by their socioeconomic, geographic and epidemiological characteristics and provision of health services11.
The population was made up of 4,643 cases of neonatal deaths of premature infants that occurred between 2013 and 2019. Duplicate cases of neonatal death, pregnant women who progressed to miscarriage, NBs with a gestational age greater than or equal to 37 weeks and women who did not reside in the state of Pernambuco, in order to exclusively evaluate premature NB who progressed to death and their trajectory in health services in Pernambuco, were excluded.
Data were collected in January 2021. Data from the Mortality Information System (SIM - Sistema de Informação sobre Mortalidade) and the Live Birth Information System (SINASC - Sistema de Informações sobre Nascidos Vivos) were used, provided by the Pernambuco State Health Department (SES/PE - Secretaria Estadual de Saúde de Pernambuco). The original SIM database, referring to neonatal deaths, and the SINASC database, referring to births in the period studied, comprised 5,904 and 968,762 cases, respectively. After cleaning and processing the data, the linkage technique was used, using the “Fuzzy LookUp” extension available in Microsoft® Office Excel to correlate both databases12.
The deterministic linkage was carried out by identifying the unifying variable common to both systems (SIM-SINASC), which was the certificate of live birth (CLB) number. At this stage, 5,243 cases were matched. For unpaired records at this stage - cases without CLB (600 cases) or CLB with different or incomplete numbers (61 cases) - probabilistic linkage was used, through the multiple-step strategy, using similarity with the mother’s name variable present in both databases, totaling 661 cases.
This stage was associated with the manual review of doubtful pairs, seeking to classify them as true pairs or non-pairs. For comparison, mother’s name, date of birth and NB’s sex were used. As a result, of the 661 cases in the probabilistic linkage stage, 314 were excluded in manual inspection, and the remaining 347 cases were added to the 1st matching database, totaling the final database with 5,590 neonatal deaths. After applying the exclusion criteria, the final sample consisted of 4,643 premature deaths in the period considered.
A descriptive analysis of the maternal epidemiological and obstetric profile and biological characteristics and neonatal health care was carried out with the help of the “R” program version 4.0.3. To assess categorical variables, absolute and relative frequencies were calculated, and for numerical variables, measurements of position, central tendency and variability (mean and standard deviation) were calculated.
Subsequently, survival analysis was performed from the date of birth. The event of interest was the death of premature NBs. The date that made up the failure (consequently the censorship) was the date of death or 1 month ahead of the date of birth (censorship). Survival time was calculated in minutes. For those who had a survival time of less than 1 day (up to 24 hours or 1,440 minutes), survival time was given as zero. Furthermore, the time until death was set at up to 28 days (neonatal period).
Bivariate analysis and modeling were carried out between study variables and the time of death of premature NBs. At this stage of the study, it was decided to exclude all lines that contained missing data, reducing the sample size to 2,504 neonatal deaths.
In bivariate statistical analysis, the Kruskal-Wallis test was applied. To carry out modeling, the generalized linear model was used based on the inverse normal probability distribution adjusted for a high volume of zero, known as Zero Adjusted Inverse Gaussian (ZAIG)13.
The dependent variable was neonatal death of premature babies. Independent variables were those contained in CLB and death certificate (DC) forms. Explanatory variables to model the mean time were inserted if their p-value in bivariate analysis was less than 0.2. With the insertion of such variables, the model went through the variable selection phase using the backward and forward stepwise algorithm, using Generalized Akaike Information Criteria (GAIC) as a metric.
Variable selection was made for each model parameter, in the following order: variance parameter around the mean lifespan; parameter that estimates the probability of living less than a day and finally; parameter that estimates the mean lifespan. Statistical analyzes were calculated using R software version 4.0.3. For modeling, the GAMLSS package version 5.3-1 was used.
The spatial component was included in the analysis with the aim of expanding the understanding of the dynamics of birth and death of premature babies, which is an essential measure to highlight inequalities in access to health services and their interface with the birth process and territorial heterogeneity of service networks. Taking into account the need to incorporate attributes capable of identifying variations in neonatal mortality among preterm infants, it was decided to estimate intercity displacement during antepartum as a key element for characterizing access to health services and the presence of attention centers.
Thus, to investigate the displacement traveled by cases of neonatal death of premature babies, spatial analysis of the flow patterns between the place of residence and the place of death, and the place of residence and the place of birth, was used. To this end, geographic coordinates of places of residence (in the format neighborhood, city, state), place of birth and places of occurrence (in the format street, neighborhood, zip code, city, state) were collected using the Google Maps API.
In spatial analysis, it was decided to exclude cases that had birth and death addresses not filled in or with formatting errors. Thus, the number of cases analyzed was reduced to flow from residence to place of occurrence (N = 3,892) and flow between place of residence and place of birth (N = 3,279).
The flow connections that make up the study were hierarchized according to their volume up to the point of arrival, thus identifying the hospitals with the highest frequency of births/deaths. Cluster analysis was used using the Euclidean distance (straight line) between the locations that make up the analyzed flows. To this end, the K-means method was applied, which generated 4 distinct groups. The number of groups was decided by the Elbow method. The calculations and graphs were made using “R” version 4.0.3. The ggmap version 3 package made it possible to use the Google Maps API.
This study was approved by the Universidade Federal de Pernambuco Research Ethics Committee, under CAAE (Certificado de Apresentação para Apreciação Ética - Certificate of Presentation for Ethical Consideration) 34372620.8.0000.5208 and Opinion 4.497.260.

RESULTS
A total of 4,643 neonatal deaths of premature infants were identified during the study period, eligible for analysis. The minimum age of death was up to 24 hours (lifetime equal to zero) and the maximum was 27 days. The mean was 2 days. Table 1 presents the maternal epidemiological and obstetric characteristics of cases investigated.
Table 1

Table 2 presents the biological and health care characteristics of premature NBs, with the majority of hospital births being in the morning, with male prevailing. The babies had Apgar in the 1st and 5th minutes below 7, cephalic presentation, extremely low birth weight and conditions originating in the perinatal period as the basic cause of death.
Table 2

Regarding premature babies’ survival analysis, time ranged from 0 to 27 days, with a mean of 4.41 days and a median of 2 days. It was found that 34% (n = 1,578) of the deaths investigated occurred within 24 hours (therefore, life span equal to zero). The 1st quartile was zero and the 3rd quartile was 6 days, which represented a prevalence of 76.1% (n = 3,533) of deaths within 6 days.
In bivariate analysis for explanatory variables, the following findings were significant and related to longer survival time: birth in the early morning period (survival time - ST=5.8 days; p<0.001); maternal age >35 years (ST =5.5 days; p=0.02); double or multiple pregnancy (ST=6.2 days; p<0.001); non-induced birth (ST=5.1 days; p<0.001); cesarean section (ST= 5.7 days; p<0.001); absence of congenital anomaly (ST=5.4 days; p<0.001); and Apgar in the 1st and 5th minutes greater than or equal to 7 (ST=7.8 and 7.2 days, respectively; p<0.001).
Survival time also showed an increasing trend; the greater the length of maternal education (p<0.001), the greater the number of prenatal consultations (p<0.001) and GA (p<0.001). Brown maternal race/color was associated with shorter survival time, with an average of 4.8 compared to the others, which were longer than 6 days on average (p = 0.01).
When mothers live in the Região Metropolitana de Recife (RMR), the mean lifespan is longer (p<0.001). Being born outside the hospital increases the chance of death by 8.85 times compared to being born in hospitals (p<0.001). When the underlying cause of death was external causes of morbidity and mortality, diseases of the circulatory system, diseases of the respiratory system and diseases of the nervous system, the mean survival time was approximately four times greater in relation to other underlying causes (p< 0.001).
As for the statistical model, double or triple pregnancy and Apgar scores in the 1st minute and 5th minute greater than or equal to seven accounted for increasing the mean life span by 18%, 18% and 27%, respectively, in relation to the reference class. When the mother’s race/color was brown, the mean survival time decreased by 16% in relation to the reference class (Table 3).
When death occurred in the morning, afternoon or night, the chances of NBs surviving less than 1 day (0 days) increased by 48% (p = 0.03), 167% (p<0.001) and 248% (p<0.001), respectively, in relation to the early morning period. Birth in a non-hospital establishment increased the chance of death in less than 24 hours by 8.85 times (p<0.001) (Table 3).
Table 3

Figure 1 illustrates the spatial pattern of flows between place of residence and occurrence of deaths. Flows with a distance between 0 km and 62.6 km correspond to 60.9% of cases (n = 2,946), whose flows present a concentration of one to 25 cases throughout the study period. It was estimated that there are 248 health services in this distance segment, of which ten establishments account for 72.3% of cases (n = 2,135).
As for the segment with distances between 62.6 km and 164.3 km, there is a total of 15.4% (n = 743) of cases, in which the flows are repeated from one to 10 times. For this segment, 105 different establishments were found, where five of them constitute 69.5% (n = 517) of cases.
In the case of the segment with distances between 164.3 km and 319.8 km, there is a total of 9.9% (n = 478) of cases, in which the most frequent flows are repeated six times. For this segment, 69 establishments were found, four of which account for 74.7% (n = 321) of cases. For the segment with distances above 319.8 km, the total number of cases is 2.23% (n = 108), in which the highest frequency flow is four times. For this segment, 26 establishments were found, two of which account for 63.9% (n = 69) of cases, both located in the state capital.
Figure 1

Figure 2 shows the flows between the place of residence and birth of NBs. For flows with a distance between 0 km and 61.3 km, there are flows with a concentration of one to 17 births throughout the study period, where the total number of births is 74.7% (n = 2,449). A total of 110 establishments were listed for this distance segment, where eight establishments stand out, together accounting for 71.6% (n = 1,764) of cases.
As for the segment with distances between 61.3 km and 163.5 km, there is a total of 17.5% (n = 572) of cases, in which the flows are repeated one to five times. For this segment, 27 different establishments were found, three of which constitute 75.1% (n = 430) of cases, all located in the capital.
Between 163.5 km and 329.3 km, there is a total of 5.9% (n = 194) of cases, in which the most frequent flows are repeated four times. For the segment with distances above 329.3 km, the total number of cases is 1.95% (n = 64), in which the highest frequency flow is four times. For both distances, three establishments stand out, with 79.9% (n = 155) and 82.8% (n = 53) of cases, respectively, all from the capital of Pernambuco.
Figure 2

DISCUSSION
Cases of neonatal death of premature NBs had a mean survival of two days, highlighting the predominance of early neonatal mortality. A study that analyzed the profile of neonatal mortality in the state of Pernambuco from 2013 to 2017 also observed 76.19% of neonatal deaths in the early neonatal period14. It is estimated that the risk of death during this period is 30 times higher in low-income countries compared to high-income countries15.
Although Apgar scores in the 1st and 5th minutes greater than seven are associated with an increase in mean life span of premature babies, it is noteworthy that almost half of newborns who did not survive had Apgar scores ≥ 7 in the fifth minute, which may be related to the quality of neonatal care and difficulties in accessing specialized health units16.
In relation to race/color, it is observed that brown mothers are associated with a higher risk of mortality, with shorter neonatal survival times compared to mothers of white race/color. Ethnic-racial inequalities reflect important points that still exist in society, which highlights the need to adapt health policies aimed at the black, brown and indigenous population to local delivery and birth care settings17. Similar findings were also observed in other studies17-18.
In relation to the causes of neonatal death, the majority are considered preventable, mainly due to actions of health services that provide maternal and neonatal care, such as identification of risk groups for occurrence of death and premature birth, women in situations of social vulnerability, adequate prenatal care, identification of gaps in the state’s obstetric care networks and training of professionals to care for high-risk pregnant women and NBs who need intensive care19.
Birth in the early morning period has a strong association due to longer survival time in the study. On the other hand, NBs born in the afternoon or at night have shorter survival times. This finding may perhaps reflect urban traffic conditions and their influence on the time taken to access specialized health services, as during the early hours there is less urban circulation.
The relationship between the period of birth and occurrence of neonatal death is still little evidenced in the literature, requiring new assessment studies, mainly at local and regional levels. Knowledge of the period of birth with the highest incidence of neonatal deaths is important data for health management, where strategies to minimize risks can be employed in order to avoid injuries.
GA is a determining factor in the risk of dying, as observed in the inversely proportional relationship between pregnancy weeks and chances of death. The higher the GA, the lower the risk of neonatal mortality, with emphasis on the group of extremely premature infants (<28 weeks of gestation), evidenced in the literature as having a greater association with early neonatal death20-21.
Care for pregnant women during the pregnancy-puerperal cycle, providing prenatal care and assistance during birth and the postpartum period, are initiatives that favor reduction of maternal and neonatal mortality rates. The number of insufficient prenatal consultations constitutes a serious public health concern and a factor associated with unfavorable neonatal health outcomes, as it reflects the deficiency in coverage of obstetric care in health services22. In this study, an association was found between the number of prenatal consultations and survival time of NBs.
The place of birth is also a crucial factor for better obstetric and neonatal indicators, as evidenced by the greater chance of death when the place of occurrence corresponds to an environment outside the hospital. This fact is directly related to birth conditions and availability of resources22.
As for the flow pattern spatial analysis, it is observed that in most cases the spatial patterns observed in the movement of pregnant women show that the place of birth was close to or around the mother’s residence. However, a significant percentage of pregnant women still travel long distances from their place of residence to their place of birth, with the Metropolitan Region being the most sought after, possibly because it is where the large reference centers for high-risk NBs are located.
The geographic location of the health establishments with the greatest demand for birth assistance indicates a pattern of inequality in the distribution of obstetric beds specialized in high-risk pregnancies, alternating between a shortage of flow in certain more inland regions and an excess in others (RMR). This fact may reflect failures in access to health services specialized in high-risk pregnancies and in the organization of obstetric care networks in the state, contributing to the pilgrimage of women in search of specialized assistance and higher rates of neonatal deaths in reference centers23. Unforeseen long distances traveled by postpartum women highlight the need for interventions in the planning and regulation of these networks and in neonatal mortality monitoring.
Although the majority of the study sample were residents of the RMR, it was observed that cases residing in the São Francisco mesoregion had a much shorter survival time than the others (3.4 days). The difficulty in accessing reference health services contributes to the pilgrimage of women in search of specialized assistance7. Case control study that assessed the pilgrimage of women during prepartum,identified an association between pilgrimage and neonatal mortality1.
The organization of labor and birth care networks, following the guidelines of regionalization and decentralization and the strengthening of qualified obstetric and neonatal care based on scientific evidence by health professionals, can favor the adaptation of health services in the state to maternal and child care, enabling the reduction of health problems and neonatal mortality rates.
There is a spectrum of disparities that influence a pregnancy. Although biological aspects are important, income, housing and family arrangements can influence greater risks to the dyad’s health. Health policies must be comprehensive and structured, and enable managers to raise awareness and train professionals who work at the forefront with the aim of reducing social indicators that impact preventable deaths19.
The incomplete data filling related to birth and death provided by SIM and SINASC, which collect CLB and DO data, was an important limitation of this study. Furthermore, it is inferred that using the linear distance between municipalities may not correspond to the real conditions of travel during the prepartum period of NBs, which may make it difficult to compare geographic location and associated factors.
The different realities of the municipalities in the countryside, the political and cultural influence and the ways of life in these locations instead of the RMR are also limiting factors for divergences in survival times between the regions. The associated factors do not mean a causal relationship between deaths.

FINAL CONSIDERATIONS
The results presented showed neonatal deaths, mostly from preventable causes, mainly in the early neonatal period and with a large percentage of deaths on the first and second day. Furthermore, survival times differed in the different mesoregions of the state, with emphasis on the Metropolitan Region and São Francisco, accounting for the longest and shortest neonatal survival times, respectively.
Many women still needed to travel long distances for the birth of their children, which was associated with shorter neonatal survival times, especially in areas further away from the Metropolitan Region. The results highlight the need to improve service flows aimed at maternal and child care, in accordance with the Brazilian Health System principles of regionalization and decentralization, with specialized care.
Knowledge of the variables associated with neonatal death and its trajectory enables professionals and managers to plan actions aimed at reducing mortality and speeding up care for risk groups. Furthermore, training strategies for health professionals who work at different levels of complexity involving maternal and child care, in welcoming actions and improving specialized care, can favor neonatal mortality reduction.

ACKNOWLEDGMENTS
This work was supported by the Coordination for the Improvement of Higher Education Personnel – Brazil (CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) – Financing Code 001, and the Dean of Graduate Studies at Universidade Federal de Pernambuco (PROPG-UFPE).
References
1. Batista CB, Carvalho ML, Vasconcelos AGG. Access to and use of health services as factors associated with neonatal mortality in the North, Northeast, and Vale do Jequitinhonha regions, Brazil. Jornal Pediatr 2018; 94(3):293-9.
2. Lima RG; Vieira VC.; Medeiros DS. Determinants of preterm infants’ deaths at the Neonatal Intensive Care Units in the Northeast Countryside in Brazil. Revista Brasileira de Saúde Materno Infantil 2020; 20(20):535-544.
3. Del Río R, Thió M, Bosio M, Figueras J, Iriondo M. Prediction of mortality in premature neonates. An updated systematic review. AnPediatr 2020; 93(1):24-33.
4. Yakoob MY. Predictors of very preterm births (born between 23 and 29 weeks’ gestation) at a tertiary care center in Karachi, Pakistan: additional multivariate analyses on data from primary cohort. BMC Res Notes 2019; 12(1):1-4.
5. Leal MDC, Bittencourt SDDA, Torres RMC, Niquini RP, Souza PRBD. Determinants of infant mortality in the Jequitinhonha Valley and in the North and Northeast regions of Brazil. Rev Saude Publica 2017; 51:12.
6. Faria RM. A territorialização da atenção básica à saúde do Sistema Único de Saúde do Brasil. Cien Saude Colet 2020; 25(11):4521-30.
7. Daniali ZM, Sepehri MM, Sobhani FM, Heidarzadeh M. A regionalization model to increase equity of access to maternal and neonatal care services in Iran. J Prev Med Public Health 2022; 55(1):49-9.
8. Lansky S, Friche AAL, Silva AAM, Campos D, Bittencourt SDA, Carvalho ML, Frias PG, Cavalcante RS, Cunha AJLA. Pesquisa Nascer no Brasil: perfil da mortalidade neonatal e avaliação da assistência à gestante e ao recém-nascido. Cad Saude Publica 2014; 30 (Supl. 1):192-207.
9. Melo JKC, Santos RB, Barros SCP, Pinheiro HDM. Itinerário da assistência ao parto de puérperas assistidas em uma maternidade de referência. Braz J Health Review 2021; 4(1):1929-43.
10. Amorim T, Felisbino-Mendes MS, Schreck RSC, Ribeiro SP, Rezende EM, Martins EF. Born in Belo Horizonte: the trajectory of parturient women and their reproductive outcomes. Rev Esc Enferm USP 2019; 53:e03441.
11. Instituto Brasileiro de Geografia e Estatística (IBGE). Cidades e Estados – Pernambuco [Internet]. 2022 [cited 2023 Aug 08]. Available from: https://www.ibge.gov.br/cidades-e-estados/pe.html
12. Maia LTS, Souza WV, Mendes ACG, Silva AGS. Use of linkage to improve the completeness of the SIM and SINASC in the Brazilian capitals. Rev Saude Publica 2017; 51:112.
13. Heller G, Stasinopoulos M, Rigby B. The zero-adjusted Inverse Gaussian distribution as a model for insurance claims. in Proceedings of the 21th International Workshop on StatistialModellingll [Internet]. 2006 [cited Jan 18, 2022]. Available from: https://www.researchgate.net/profile/Gillian-Heller/publication/228620979 _The_zero-adjusted_inverse_Gaussian_distribution_as_a_model_for_insurance_ claims/links/00b7d525dba22b5d9e000000/The-zero-adjusted-inverse-Gaussian-distribution-as-a-model-for-insurance-claims.pdf
14. Santos MGS, Queiroz MP, Oliveira NB, Oliveira JSB, Sousa NFC, Leal LP, Lima APE. Profile of early neonatal mortality in Pernambuco and potential for avoidability. Res Soc Develop 2020; 9(11):e4819119953.
15. Teixeira JAM, Araujo WRM, Maranhão AGK, Cortez-Escalante JJ, Rezende LFM, Matijasevich A. Mortality on the first day of life: trends, causes of death and avoidability in eight Brazilian Federative Units, between 2010 and 2015. Epidemiol Serv saude 2019; 28(1):e2018132.
16. Veloso FCS, Kassar LML, Oliveira MJC, Lima THB, Bueno NB, Gurgel RQ, Kassar SB. Analysis of neonatal mortality risk factors in Brazil: a systematic review and meta-analysis of observational studies. J Pediatr 2019; 95(5):519-30.
17. Souza BFN, Sousa NFC, Sette GCS, Lima APE, Leal LP, Holanda ER. Determinants of neonatal mortality in a municipality of the Zona da Mata in Pernambuco. Rev Esc Enferm USP 2021; 55:e03726.
18. Lima JC, Mingarelli AM, Segri NJ, Zavala AA, Takano OA. Population-based study on infant mortality. Cienc Saude Colet 2017; 22(3):931-9.
19. Prezotto KH, Oliveira RRD, Pelloso SM, Fernandes CAM. Trend of preventable neonatal mortality in the States of Brazil. Rev Bras Saude Mater Infant 2021; 21(1):291-299.
20. Gaíva MAM, Lopes FSP, Ferreira SMB, Mufato LF. Neonatal deaths of low birth weight newborns. Rev Eletr Enf 2018; 20(20):1-10.
21. Pereira MUL, Lamy Filho F, Anunciação OS, Lamy ZC, Gonçalves LLM, Madeira HGR. Neonatal death in the city of São Luís: causes and factors associated with early neonatal death. Rev Pesq Saude 2017; 18(1):18-23.
22. Garcia LP, Fernandes CM, Traebert J. Risk factors for neonatal death in the capital city with the lowest infant mortality rate in Brazil. J Pediatr 2019; 95(2):194-200.
23. Dutra HRR, Trombela LKB, Lima RS, Alberto NSMC, Ramos CV, Pereira TG. Factores relacionados al bajo peso al nacer em maternidad referente en Piauí. Rev Interd 2021; 14(1):9.