A population-based study of neonatal deaths in Indonesia based on the Indonesian demographic health survey: what determinants play an essential role?

Neonatal mortality appears to be one of the most concerning problems to fulll Sustainable Developmental Goals globally. Indonesia, as a developing country with uneven distribution of standardized health facilities over the archipelago, has been reported to be the country with the highest fatality cases of the newborn in Southeast Asia. To address this problem, we evaluate how substantive the socioeconomic spectrum and proximate determinants as a substantial predictor of the neonatal mortality rate in advance to maximize the health policies’ quality in reducing the rate of newborn death. The analysis was conducted using the data source of the 2017 Indonesia Demographic Health Survey from 11.965 live-born infants born from singleton pregnancy in the year of 2017. By using a hierarchical approach and logistic regression, the multilevel analysis was carried out to assess the possible contributing factors including socioeconomic, household, and proximate factors to neonatal mortality.


Results
At socioeconomic determinants, the odds of newborn death was signi cantly higher for those who born from mothers with poor education level (OR = 1.72, p = 0.03), insu cient antenatal visits (OR = 3.98, p = 0.01), and not being involved to postnatal care (OR = 6.60, p = 0.01). Regarding community factors, the variable of traditional birth attendants was signi cantly higher for the odds of newborn death (OR = 2.06, p = 0.01) as well as the delivery in government public hospitals (OR = 1.89, p = 0.01). In terms of proximate determinants, the odds of newborn death found to be higher for male infants (OR = 1.43, p = 0.03) and low birth (< 2.5 kg) weight infants (OR = 4.15, p = 0.01). After the adjustment of these covariates, the newborn death were associated to mothers with insu cient antenatal care (OR = 2.58, p = 0.01), not participating in postnatal care (OR = 5.66, p = 0.01), assisted by traditional birth attendants (OR = 1.46, p = 0.03), and neonatal factors such as male gender (OR = 5.66, p = 0.01 and low birth weight (OR = 4.37, p = 0.01).

Conclusion
In reducing neonatal death, public health interventions should be targeted to individual and communitylevel factors of socioeconomic determinants. Improving the quality and coverage of perinatal health services such as the utilization of either antenatal care or postnatal care, the availability of trained birth attendants, and the optimization in public hospitals' services have a signi cant meaning for better infants' lives.

Page 3/27
Background Based on data reported by the World Health Organization, approximately 2.4 million children die within 1 month after birth worldwide. Approximately, the 47 % of million children died have been contributed by around 7000 cases of neonatal deaths (1)]. Most neonatal deaths occur within the rst week of life and are linked to various aetiologies [ (1,2). Complications of preterm birth (35%) and intrapartum-related events (24%) are the leading causes of neonatal deaths, whereas infections (15%) and congenital defects (11%) play a lesser role (1,2). Although the rate of neonatal deaths has decreased by 2.6 million since 1990, the rate of decline in mortality is still slower than in children younger than 5 years. With 28 deaths per 1000 live births, countries in Central Asia and sub-Saharan Africa have the highest rates of neonatal mortality in last decade (1)(2)(3).
Sustainable developmental goals were implemented to reduce mortality rate of newborns to at least 12 deaths per 1000 live births by 2030. However, in 2016, the rate of deaths in Southeast Asia was still 14 per 1000 live births (2,3). Indonesia ranks seventh place globally in the neonatal death rate, with 64,000 fatality cases of newborns and a rate of 14 deaths per 1000 live births. Among all Southeast Asian countries, Indonesia also has the highest number of neonatal mortality cases (1). A decrease in child mortality could be realised by organising an expanded and effective intervention for preventable diseases. The quality of care around the childbirth period is associated with diseases and conditions conducive to neonatal deaths. Thus, health services and skilled personnel who perform cost-effective interventions proven to treat and prevent such diseases and illnesses must be afforded. The greatest proportion of vulnerable children comes from poorer areas and households. To attain successful consistency in pursuing sustainable developmental goal targets, international communities should take decisive roles to stop preventable newborn deaths (2, 3).
Indonesia has a large population. Before becoming a middle-income country, in the early 1970s, Indonesia was included among the poorest countries in the world because of its low life expectancy rates and literacy, as compared with current years (4). In this high-population country, health facilities are not distributed well in rural and isolated areas that contain uneducated and pauper communities (5). With regard to health insurance, Indonesia has implemented a national health insurance scheme for all citizens (6). Well educated households have better knowledge that enables them to use antenatal care services during the mothers' pregnancy and therefore secure immunisation for their children (7,8).
Indonesia is a large archipelago, comprising more than 17,000 islands, and disparities between various communities living in different areas present a challenge to the government in their goal of equalizing health service determinants over all regions, including infrastructure, service quality and skilled personnel (9)(10)(11)(12).
In this study, we use demographic health survey data to evaluate to what extent the equity of various diversifying factors can predispose the population to trends in the neonatal mortality rate. Learning from a previous study that assessed the determinants of neonatal deaths as a representative of Indonesian births from 1997 to 2002, we examine the characteristics of the socioeconomic spectrum as a substantial predictor of the neonatal mortality rate (13). This examination allows for an extended evaluation of inequality among communities in achieving the sustainable developmental goal of reducing the neonatal mortality rate.

Data sources
This population-based study aims to examine the individual and community determinants that affect neonatal mortality. We identi ed the 2017 data set of the Indonesia Demographic Health Survey (IDHS) retrieved from all of Indonesia, including 26 provinces, and recorded all cases of neonatal deaths that occurred in Indonesia between July and September 2017. We classi ed the IDHS samples received from every province into either urban or rural areas.

Data collection
A multistage strati ed random sampling technique was performed to clarify the population census over 3 consecutive months to determine the census blocks, the primary sampling unit expressed in each level.
Beginning from the strati cation, we divided households into those residing on Javanese Island and those living on the other islands or parts of Indonesia. We obtained information from the IDHS in 2017 using a standard designed commonly for countries with a high prevalence of contraception, the Demographic and Health Survey Model Questionnaires (14), to frame three questionnaires, including the Household Questionnaire, the Women's Questionnaire for married women, and the Men's Questionnaire for married men. The questionnaire covered personal information on age, education, occupational status, mother's demographic characteristics (such as full obstetric histories and antenatal care-receiving history) and neonatal characteristics (including age, sex, birth weight and postnatal care history). We used two-stage probability sampling to collect samples from the Demographic Health Survey and obtained the data randomly. To classify samples into subgroups or homogeneous strata to minimise sampling error, we performed strati cation of samples based on geographical site and urban/rural areas.

Conceptual framework
We conducted our analysis based on the study by Mosley and Chen, which applied a conceptual framework to evaluate child survival in developing countries (15). Using this framework, we subdivided the information from the IDHS data sets containing the determinants into socioeconomic and proximate determinants ( Table 1). We de ned all variables that can have a direct impact on neonatal death, including maternal-and neonatal-derived factors, as proximate determinants.

Study variables
In this study, the death of neonates (i.e. newborns who lived for 1 month after delivery) is the primary outcome. The neonatal mortality rate is de ned as the number of newborn deaths per 1000 live births. Using a model of descriptive analysis, we used a binary variable with a coding system to describe the outcome. We recorded information on socioeconomic determinants, including individual-, household-and community-level variables and proximate determinants linked to maternal and neonatal factors contributing to neonatal death. Table 1 presents further descriptions of the variables evaluated in this study.

Study design
The design of this study was cross-sectional. We performed a univariate analysis using a table of distribution frequency to measure the distribution of data and thereafter carried out bivariate and multivariate approaches to determine the variables' signi cance to the outcome.

Statistical analyses
To present the true distribution of the country, the distribution of the sample rst needed to be weighted. The weighting process for the sample was based on the statistical guidelines of the DHS. We had to perform the weighting process rst because some states or regions were over-represented and some were under-represented. This unweighted distribution does not represent the population accurately. The frequency tabulation was carried out to describe the data distribution among the variables analysed and therefore used to measure the prevalence of neonatal mortality. This was followed by contingency table analysis, to elucidate how those variables could affect the predicted outcome, without adjusting them with the other covariates. We measured the association between the determinants and the outcome using an odds ratio (OR), 95% con dence interval (CI) and p-value.
To investigate the spread of data in our analysis, we carried out frequency tabulations. We further analysed the signi cance of the effects of possible determinants on neonatal mortality using a crosstable in the Statistical Package for the Social Sciences (SPSS) software. According to the conceptual framework that elucidated a systematic relationship among the tested variables, we conducted multilevel logistic regression using such approaches. First, we analysed the impacts on neonatal mortality of the variables from the socioeconomic and proximate determinants entered in the rst model. Using SPSS version 16, we analysed the data set with a chi-square model. The variables analysed were then de ned as signi cant if p > 0.05. In the rst model, the correlation between the determinants and outcome was presented by ORs and 95% CIs and therefore weighted to illustrate sampling probability.
Variables that demonstrated a signi cant association with the outcome remained eligible for the second model. Through the second model, we computed the signi cant variables from both socioeconomic and proximate determinants, to examine the signi cant effects of certain variables on the outcome, against other variables that may act as an adjustment. We used the SPSS program to perform statistical analyses using binomial logistic regression. From the second model, the ndings were recorded by measuring the p-value and adjusted OR (95% CI) that was obtained by estimating how strong the variable analysed was, as compared with other covariates in deciding in the study outcome.

Results
We recruited 11,965 newborns from singleton pregnancies within 3 years of the survey. With regard to geography, 6076 (50.8%) were from rural areas and 5889 (49.2%) were from urban areas. West Java, which accounted for a total of 1187 cases (9.9%), was the area in Indonesia with the highest number of neonatal deaths over the 3 years. Based on the 3-year analysis shown in Table 2, there was 148 cases (1.2%) of neonatal deaths. Nevertheless, as shown in Table 3, and as compared with Javanese regions with lower percentages of neonatal deaths, West Papua had the highest number of neonatal deaths, 7 of 165 cases (4.2%). Not far behind, North Sulawesi reported 5 of 126 (1.70%) newborn deaths, which was the second highest throughout the years. Most cases of foetal deaths were among mothers with a lower educational status, including 3067 uneducated women (25.6%) and 6722 undergraduates (56.2%), as compared with 2176 graduates (18.2%). In terms of attitude towards pregnancy, 10,571 women (88.3%) were aware of their pregnancy and attended at least four routine antenatal visits. Two months after delivery, 8068 women (67.4%) had received su cient postnatal care. From the total family populations, up to 11,666 couples (97.5%) were employed, whether as either a single-earner or dual earners, whereas the other 299 households (2.5%) did not have a regular job.
With regard to childbirth status, 9960 parents (83.0%) reported that their pregnancy was planned, and they embraced their child, whereas 1079 (9.0%) of the total population reported the childbirth as an unwanted event. In addition, 926 parents (7.7%) did not show affection to their child and were not content with their infant's birth. Furthermore, in terms of the categorisation of the families' wealth index, a large number of families did not make an adequate living, with 3217 (26.9%) families being destitute and 2839 (20.0%) families being poor. Following this, 2224 families (18.6%) were classi ed in the middle wealth category, with the remainder being above average in wealth. Table 3, which presents community-level factors including perinatal care during the pregnancy period, 357 women had undergone prenatal care not given by healthcare professionals, midwives or physicians. During labour, most women (8540; 72.1%) were assisted by midwives, whereas 2197 (18.6%) were aided by physicians. However, the other 832 women (7.1%) did not receive proper medical assistance and thus sought help from traditional birth attendants (TBAs). Surprisingly, 228 (2.2%) deliveries had not been assisted by anyone. After the delivery process, 554 mothers (4.6%)

As shown in
reported not receiving appropriate medical attention for postnatal care. Delivery events occurred mostly at public hospitals (4042; 33.8%), followed by private hospitals (4822; 40.2%). Approximately 259 women (2.2%) could deliver their babies at unit kesehatan berbasis masyarakat (UKBM), a health facility pioneered by communities, and 2842 women (23.8%) reported labouring in their homes.
Further analyses to compare neonatal deaths and neonatal live births revealed remarkable ndings that clari ed the assessed determinants ( were not signi cantly increasing the OR for neonatal deaths after adjustment. As shown in Table 5, the regions of Papua and West Papua had the lowest percentages of antenatal care and postnatal care utilisation. Maluku Island had the highest probability of delivery being assisted by a TBA.

Discussion
In this analysis of the IDHS data in 2014-2017, fewer than four antenatal visits throughout pregnancy and the absence of postnatal care services had the most signi cant effects on increasing the odds of neonatal mortality. This result is in line with a study carried out by Laksono, which demonstrated that women with su cient antenatal care (more than four visits) were at a lower odds of witnessing their newborn's death (16). Our ndings also support the results of the previous study using 2017 IDHS data, in which the authors found that attending more than four antenatal care visits during the pregnancy period had a protective effect on neonatal death. However, the study conducted by Masruroh et al. did not detect other signi cant factors that prevent a rise in the odds of neonatal mortality incidence (17). The use of antenatal care varies among regions in Indonesia, with a low coverage of antenatal care (fewer than four visits) observed in the Maluku and Papua regions (18). Antenatal care services were reported to be underserviced in rural areas outside the Java-Bali region, which consists of families with a low household wealth index and low maternal education level. The other associated factors reported were the distance between the mother's residence and the health facility and the lack of obstetric complications before delivery (8). Based on a report by Suharmiati et al., this low utilisation of antenatal care visits in the remote and border islands with extreme topography was affected by the low availability of healthcare services and facilities, non-optimised health infrastructure and inappropriate reward for healthcare providers (19). Furthermore, in urban areas of Indonesia, the constraints of actualising e cient antenatal care services are made worse by several factors, including poor laboratory and management records during antenatal care services, inadequate monitoring documentation and the fact that less than 5% of women receive all the routine measurements, consisting of urine tests, blood screening tests, iron supplementation, tetanus vaccination and information on pregnancy complications (20). Meta-analyses conducted in African countries have also implied that the bene ts of antenatal care visits facilitated by skilled providers can reduce the incidence of neonatal deaths (21,22). Another meta-analysis that investigated the effect of antenatal care visits in various developing countries revealed that the risk of neonatal death could be reduced by adequate antenatal care visits (23). With regard to postnatal care, delivery facilities that supply postnatal care might attenuate the risks of neonatal mortality and therefore improve the lives of newborns (24,25). A study conducted in Garut, Sukabumi and Ciamis found that women who attended antenatal and postnatal care services experienced nancial di culties related to the cost of health services or transportation, the distance to the nearest facilities and the lack of awareness, as indicated by the need to use the services only if obstetric complications had already occurred (26). Based on a report from Titaley, the non-participant group of mothers in postnatal care services was related mostly to the low household wealth index, lack of understanding of pregnancy complications and low education levels (27). Furthermore, similar to our nding that postnatal care plays a signi cant role in reducing the risk of neonatal deaths, a nationwide case-control study in India reported an increase in the odds of neonatal death among women undergoing childbirth in delivery facilities who did not attend a postnatal check-up (25). In general, this study supports the previous studies of the IDHS 2002-2003 that revealed the association between inadequate utilisation of perinatal health services, either antenatal care or postnatal care, and the increased odds of neonatal mortality. Clusters of deliveries that received assistance by TBAs had a signi cant reduction in the odds of neonatal death (13)].
Nevertheless, this study also revealed varying factors that induce neonatal deaths as perinatal outcomes. The delivery process facilitated by a TBA caused the odds to increase. The absence of education in mothers was found to increase the OR of neonatal deaths signi cantly. This nding suggests, in line with another report in Brazil, that a low level of maternal education has a signi cant effect on the increase in newborn death. In other words, low levels of education can induce a lack of discipline and adherence of women to a bene cial health curriculum (28). Analyses of the Indian National Family Health Survey in 1994 found that low education levels, regardless of age (i.e. adolescents or older women), increased the odds of neonatal death (29). In addition to the effect of low education levels of mothers, a study in Gaza, Palestine, reported that fathers who had an educational certi cate at the highest level of secondary school was associated signi cantly with neonatal mortality. The difference between the groups with low and high educational levels (diploma or a university degree) was associated with the mothers' adherence to antenatal visits and to breastfeeding their infants, whereby most parents with a lower educational level did not meet the required criteria (more than four antenatal care visits and exclusive breastfeeding) (30).
Andriano et al. conducted a demographic study in Malawi to evaluate the effects of education system reform (i.e. free primary education for individuals during the 1990s) and found that education quality resulted in a lower probability of death in infants and children younger than 5 years (31). In general, these studies suggest that knowledge barriers between healthcare professionals and families, because of low educational levels, re ect the mothers' attitudes to modern health services (28)(29)(30)(31). However, after adjustment, the OR was not signi cant as compared with other covariates.
Our demographic study found that childbirths assisted by a TBA increased the risk of neonatal death signi cantly. With regard to the person assisting the childbirth, Titaley et al. analysed the effects of trained delivery attendance and place of delivery on early neonatal deaths. The results of that study further support our analyses by elucidating that women living in rural areas and experiencing home delivery with appropriate assistance from skilled healthcare professionals had a signi cantly reduced chance of having their newborns die. In addition, delivering in a healthcare facility or in the home without the assistance of trained delivery attendants could increase the risk of neonatal death (32). In another study on the IDHS data sets conducted by Badriah et al., labour facilitated by a skilled birth attendant was correlated with a higher risk of labour complications, namely, prolonged labour and postpartum fever. Besides this, there were no differences in stillbirths or neonatal deaths between women assisted by trained birth attendants and those assisted by untrained attendants (33). The use of services provided by TBAs among pregnant women in Indonesia is still inevitable, because of the traditional beliefs and recommendations from families that convince women in achieving a satisfying pregnancy and perinatal care (34,35). Proximity to institutional healthcare, transportation problems and low nancing within the family units are major constraints to accessing su cient healthcare services, including delivery assistance from TBAs (34,36). Ironically, based on reports in Bangladesh, some communities did not access a health facility because no proper facility was available in their area or the care available was of poor quality. Thus, women who do not want to attend a healthcare facility to manage their pregnancy and who prefer untrained birth attendants may fail to recognise early the alarming signs of dangers in pregnancy (36). In Indonesia, traditional beliefs affect women's preference between midwives and TBAs and therefore in uence their adherence to antenatal care visits in several rural areas. Based on trends among communities in Indonesia, women believe that pregnancy is a part of their ordinary life cycle and is not a complexity that can appear as a risk factor for comorbidities and complications. This trend then in uences women to be more active in seeking appropriate health facilities, including su cient antenatal care services and delivery assistance from TBAs (35).
Our study found that delivering in public hospitals could increase the odds of neonatal death, although it was not signi cant. A meta-analysis assessing the effect of delivery at healthcare facilities on neonatal deaths showed that delivering in a health facility was effective in reducing the odds of neonatal death, especially in low-to middle-income countries (37). Our nding is in line with Titaley's report, which elucidated higher odds of women experiencing the death of their neonates in public hospitals than in private hospitals (32). Based on a report in Ethiopia, the rate of neonatal deaths in public hospitals was in uenced mostly by preterm birth, birth asphyxia and neonatal infection (38). In addition, Adams et al. suggested that a disparity exists between the ability to alleviate the burden of the neonatal mortality rate in public hospitals and private hospitals, in which public hospitals demonstrated a higher rate of neonatal deaths. This disparity was probably affected by different clinical practices, obstetrician-led care models and congenital anomalies experienced by babies born in public hospitals (39). Occasionally, the number of neonates died outside of hospitals could be higher that than in hospitals because communities cannot recognize the various life-threatening conditions in young babies, namely, birth asphyxia, prematurity and birth injuries. However, this phenomenon led to the late detection and treatment of babies whose condition was commonly complicated. For instance, a few reports from countries in the Middle East and South Asia found that prematurity with complications and sepsis were the most frequent contributing factors causing neonatal deaths in the hospital setting (40)(41)(42).
Proximate determinants including male gender and birthweight < 2.5 kg also increased the odds of neonatal deaths. This nding is in line with previous demographic studies, which indicated that the risk of neonatal death was higher in male infants and newborns who were smaller in size (32,43,44). Mostly, the disadvantage of the rising number of male infant deaths was attributed to perinatal conditions in the decades before improvements in obstetric practices. However, the innate biological differences between both gender groups turned out to be more signi cant in developing the risk of male newborn deaths (45). Reports from other countries indicated various trends. Based on multilevel analyses carried out in Brazil to determine the factors associated with neonatal mortality, maternal characteristics were not associated with neonatal deaths. Instead, neonatal characteristics, including foetal congenital anomaly, low Apgar score at 1 minute and low birthweight, presented a signi cant association (46). In Ethiopia, a few regions including Tigray, Omhara and Benishangul Gumz showed a signi cant association with neonatal deaths as the study outcome. Maternal age < 18 years was signi cant to the study outcome, whereas in terms of neonatal aspects, male gender and being born in the winter or rainy season were correlated signi cantly with the outcome (47).

Study limitations
Despite the strength of this study, which is related to its capability of forming a nationwide representation based on a survey and multilevel analyses to clarify the association between tested determinants and the outcomes speci cally, this survey did not cover data on maternal deaths or all variables speci c to neonatal aspects, such as genetic and environmental factors. In terms of maternal aspects, other complications, including vaginal bleeding, fever and convulsions, were not speci ed to minimise data distribution.

Recommendations
As a result of the disparities in perinatal care services, including antenatal care and postnatal care utilisation over various regions in Indonesia, healthcare providers and policymakers must refer to the ndings of the IDHS to make a decision to maximise the utilisation of such services. Healthcare providers should implement a few interventions, such as education targeting mothers regarding the importance of antenatal care programmes for their condition, composed of screening, detection, management and monitoring, because these interventions can be effective in increasing mothers' awareness and can provide integrated care starting from the pregnancy process to childbirth and the postnatal period. In the other words, mothers' awareness should not be limited to worrying about obstetric complications but should be focused on every second of the pregnancy period. To equalise the quality of antenatal care services in Indonesia, policymakers must establish a policy of implementing funding reallocation to improve the health infrastructure in regions with a low coverage of perinatal care services; otherwise, the standardised reward for healthcare workers should be guaranteed.
In addition, facilities and delivery attendants in public hospitals should be trained effectively to ensure that the delivery practice is safe. Traditional delivery attendants deemed to be unskilled must be encouraged to undergo training in perinatal self-care skills. In other words, to improve maternal and newborn care, public hospitals must be used e ciently by every pregnant woman who needs perinatal care and childbirth services, by establishing training for midwives and physicians working at such health facilities. In addition, TBAs must be empowered to work collaboratively with healthcare providers. An integrated system is needed that provides training for unskilled birth attendants, including midwives and TBAs, in every region of Indonesia that is still isolated from healthcare centres.

Conclusions
With regard to proximate determinants, notable factors associated with neonatal factors including male sex and low birth weight is likely to prompt neonatal deaths. Furthermore, a few socioeconomic determinants, namely, perinatal care services, either antenatal care or postnatal care, and TBA as birth attendants turned out to be the signi cant factors causing neonatal deaths. To reduce neonatal deaths, it is recommended that all pregnant women should be advised to attend the services offered by healthcare facilities, in the form of su cient antenatal visits (more than four times throughout the pregnancy) and postnatal care. A better integrated system with contributions by healthcare providers and policymakers should ensure that every mother can increase their awareness and be facilitated during their attempts to obtain routine perinatal care services, including antenatal visits and postnatal care. In addition, unskilled birth attendants should receive adequate training to ensure that delivering babies will be safer. We downloaded data from the DHS after presenting our project proposal and receiving permission from the DHS Program (132989.0.000). The original data sets of the IDHS were collected in accordance with international and national ethical guidelines. The datasets from 2017 IDHS have already obtained ethical clearance approved by the National Institute for Health Research and Development of the Indonesian Ministry of Health to be generally used for demographic study purposes.

Abbreviations
The National Institute for Health Research and Development of the Indonesian Ministry of Health also gives the approval for informed consent stated by study participants to be waived for this study purpose. The utilization of 2017 IDHS data in this demographic study has already con rmed by ICF International through the website as follows: https://dhsprogram.com/data/new-user-registration.cfm after reporting and communicating the study purposes and the analysis method.

Consent for publication
Not applicable

Availability of data and materials
The dataset used during the current study are in the public domain and can be obtained from the DHS Program (http://dhsprogram.com/) or from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests