Determinants of early neonatal mortality in Afghanistan: an analysis of the Demographic and Health Survey 2015

We used data from the Afghanistan Demographic and Health Survey (AfDHS 2015) to investigate determinants of early neonatal mortality. The AfDHS 2015 was the first DHS implemented in Afghanistan. It was a part of the global DHS Program. The survey was conducted from June 2015 to February 2016, and primarily provided estimates of basic demographic and health indicators of the country. It used an updated version of the ‘Household Listing Frame’ as the sampling frame. Details of this population-based survey including survey design, methodologies, sample size calculation and questionnaires have been described elsewhere [4].

A stratified two-stage sample design was followed in AfDHS 2015 to allow estimates of key indicators at the national level, in urban and rural areas, and for each of the 34 provinces of the country. During the first stage, the sample points (i.e., clusters) consisting of EAs were selected. A total of 950 clusters were selected in this stage; 260 and 690 separately for urban and rural areas, respectively. Some of the regions (10%) were difficult to reach due to security reasons; 101 reserve clusters were selected in all of the provinces to replace the inaccessible clusters [4].

The AfDHS used three questionnaires: women’s, men’s, and household’s questionnaires. All ever-married women with 15–49 years of age living in the households were eligible to be interviewed. They were interviewed using the women’s questionnaire. During the interview, several types of information were collected. These included: background characteristics (including age, education, and media exposure), birth history and child mortality, knowledge and use of family planning methods, fertility preferences, antenatal, delivery, and postnatal care, breastfeeding and infant feeding practices, vaccinations and childhood illnesses, marriage, women’s work and husbands’ background characteristics, awareness and behavior regarding HIV/AIDS and other sexually transmitted infections (STIs), adult and maternal mortality, knowledge, attitudes, and behavior related to other health issues (e.g., tuberculosis, hepatitis, fistula) and domestic violence (questions asked of one woman per household) [4].

In total, 25,741 households were selected for the AfDHS, of which 24,941 were occupied during the survey fieldwork. Of the occupied households, 24,395 were successfully interviewed with 98% response rate. In the interviewed households, 30,434 ever-married women age 15–49 were identified for individual interviews; interviews were completed with 29,461 of these women, yielding a response rate of 97% [4]. Overall, the survey was successfully carried out in 956 clusters [4].

Among the 29,461 interviewed women, we examined the determinants of early neonatal mortality among the live-births of a cohort of women who had at least one live-birth within four years preceding the survey. A total of 19,636 women delivered at least one child within this period. We included their most recent live-births.

Figure 1 shows the modified conceptual framework that was adapted from Mosley and Chen’s framework for survival of children in developing countries [14]. This framework was modified based on published reports, limitations and the structure of the AfDHS dataset. Variables of the first two levels were considered individual level, which were nested within the community (i.e., cluster) where an infant lived.

The outcome variable for this study was early neonatal death, defined by a binary variable as the death of a live-born baby within the first week of life. The survived mothers were interviewed about the day of deaths of their children. The aggregated individual level characteristics of the community level were used to construct community level variables. The high and low categorization of the aggregate variables was done based on the distribution of the proportion values calculated for each community. Table 1 shows the list of the included explanatory variables (i.e., risk factors for neonatal deaths) along with their definitions and categories.

Stata 14.0 (Stata Corporation, College Stations, TX, USA) was used to analyze data of this study [15]. Weighted frequency was calculated for all study variables after adjusting for two-stage cluster sampling design of the survey. This analysis included one live-birth per woman as a single unit.

First, a frequency table was used to describe the basic characteristics (Table 2). After reporting the frequency distribution, logistic regression analysis was conducted to investigate the association between the potential determinants and early neonatal deaths. Before conducting the logistic regression analysis, continuous (e.g., maternal age) and discrete variables (e.g., parity) were converted into categorical variables. Crude odds ratios (CORs) were calculated by entering all potential predictors into the baseline equation (i.e., one variable at a time) with early neonatal mortality as the outcome variable (Table 3).

The hierarchical nature of the AfDHS data was considered for the multivariable analysis. Thus, a two-level multivariable analysis was applied. Proximate and socioeconomic factors were considered to be nested within the community (i.e., cluster) level. Covariates with a predetermined significance level (p < 0.20) in bivariate analyses were included in the adjustment for multivariable analysis (Table 3) and adjusted odds ratios (AORs) were calculated. To prevent residual confounding in multivariate analysis, the significance level of 0.20 is considered sufficient [16]. Odds ratios (ORs) were reported with 95% confidence intervals (CIs) and significance levels (Table 3). Variance inflation factors were assessed to examine collinearity between variables before entering them into the multivariable models. The AfDHS employed the principal component analysis of basic housing construction materials (i.e., materials used to construct the walls, roofs, and floors), sources of water, sanitation facilities, electricity, and household belongings to construct wealth index score for the households' wealth status. Then the wealth status was stratified into quintiles: poorer, poorest, middle, richer and richest [17, 18].

Table 2 presents distribution of selected risk factors for early neonatal mortality. Numbers and percentages of the analysis were weighted by the individual sampling of the sampling weight from the AfDHS data. Among the included 19,636 weighted live-births that took place within four years preceding the survey, 1.4% (n = 266) children died within the first week of their birth. Approximately 52.2% of the study sample was males. Percentage of neonatal mortality was higher among males (1.6%) compared to females (1.1%). Nearly 61.1% (n = 11,727) of the infants were average size at birth and had lower mortality rate than smaller or larger infants. Most of the newborns were singletons, 98.9%. About one-sixth of the mothers were primi (i.e., became pregnant for the first-time in life). A vast majority of the mothers were in the middle age group (19–34 years, 77.3%), followed by older (35–49 years, 17.7%) and younger age groups (≤18 years, 5.0%). A majority of the mothers received antenatal care, 61.9%. Nearly half of the deliveries were conducted by skilled attendants that had a lower percentage of early neonatal mortality in comparison to the deliveries conducted by unskilled attendants, 1.0% and 1.7%, respectively. About 55.0% of the deliveries were conducted at home. Most of the households’ head were males. A vast majority of the mothers (82.9%) had no formal education. Approximately an equal number of respondents were obtained from each household wealth quintile group or each region of the country. More than three-fourths of the participants were from rural areas. About one-fourth of the participants were from a community where at least 25% of the women received four or more antenatal visits. Most of the infants were from a community with ‘low’ education level (87.9%), which in turn had a higher percentage of neonatal deaths than the communities with higher education level, 0.4% and 1.4%, respectively.

Table 3 shows results of logistic regression analyses. In unadjusted level, multiple gestations had the highest odds of early neonatal mortality (COR: 8.6; 95% CI: 4.7–15.5). Male children had a higher likelihood of dying than their female counterparts (COR: 1.6; 95% CI: 1.1–2.1). Both smaller and larger babies had higher odds of mortality than average sized babies. Birth rank was found to be associated with a higher likelihood of mortality among a birth order of five or more (COR: 1.9; 95% CI: 1.3–2.9). Children delivered by mothers with a birth interval of less than two years had higher odds of dying than a birth interval of two or more years (COR: 1.5; 95% CI: 1.0–2.3). Maternal age also had a significant association with the outcome variable. Children delivered by mothers who received antenatal care (COR: 0.5; 95% CI: 0.4–0.7), delivered by skilled health personnel (COR: 0.6; 95% CI: 0.4–0.8) and delivered at heath facilities (COR: 0.6; 95% CI: 0.4–0.8) were less likely to die than the children delivered by mothers who did not utilize these services. Parental education levels also had a significant association with the outcome variable. Rural infants were nearly two times more likely to die than urban infants (COR: 1.9; 95% CI: 1.1–3.2). Odds of deaths were significantly lower in north-western (COR: 0.3; 95% CI: 0.2–0.6) and central-western (COR: 0.6; 95% CI: 0.4–0.9) regions than the north-eastern part.

We found two pairs of collinear variables: parity and delivery attendance with maternal age and place of delivery, respectively; we kept maternal age and place of delivery for adjustment. After adjusting the variables, ordered from the most significant odds, the following categories were significantly associated with an increased likelihood of early neonatal mortality: multiple gestations (AOR: 9.3; 95% CI: 5.7–15.0), larger (AOR: 2.9; 95% CI: 2.2–3.8) than average birth size, birth interval of < 2 years (AOR: 2.6; 95% CI: 1.4–4.9), smaller (AOR: 1.8; 95% CI: 1.2–2.6) than average birth size, maternal age ≤ 18 years (AOR: 1.8; 95% CI: 1.1–3.2) and ≥ 35 years (AOR: 1.7; 95% CI: 1.3–2.3). The variables associated with lower odds of dying were: antenatal care by a skilled provider (AOR: 0.7; 95% CI: 0.5–0.9), institutional delivery (AOR: 0.7; 955 CI: 0.5–0.9), paternal higher education level (AOR: 0.7; 95% CI: 0.5–1.0), living in north-western (AOR: 0.3; 95% CI: 0.1–0.6), central-western regions (AOR: 0.5; 95% CI: 0.3–0.9) or in a community with higher maternal education level (AOR: 0.4; 95% CI: 0.2–0.9).

An integrated or multifaceted approach is required to address all potential factors of various levels associated with early neonatal deaths in Afghanistan. From a program-planning perspective, it is essential to take modifiable risk factors into account to design newborns’ survival program. Implementation of the continuum of care approach is required to further increase the proportion of women with antenatal care, facility delivery or delivery by skilled delivery personnel [56]. Adapting family planning services for women to increase birth interval timing could reduce the proportion of pregnancy among younger or older mother. Delaying age at marriage could also reduce mortality, which is associated with younger maternal age [49]. To reduce premature births, several behavioral (e.g., smoking cessation) and medical interventions (e.g., progesterone supplementation) are effective [57]. Moreover, Kangaroo Mother Care (KMC) could be an effective and low-cost intervention to prevent preterm or small size newborns’ mortality [58, 59]. These components should be included in newborns’ survival strategy. Furthermore, strengthening of the Afghan health system could facilitate better management of preterm or larger babies in addition to other sick newborns, as well as increase utilization and quality of such services. Further research is required to recognize other unexamined factors in the context of this country. A common platform could be made with researchers and policymakers from all South Asian countries, as most of the countries in this region share this common problem of high early neonatal mortality rates [1, 2].