Determinants of neonatal mortality in Pakistan: secondary analysis of Pakistan Demographic and Health Survey 2006–07

Data used for the current study were derived from the PDHS 2006–07, which used a stratified, multistage cluster sampling method to ensure a sample representative of the population of Pakistan excluding the federally administered northern and tribal areas and restricted military and protected areas. Urban and rural samples were drawn separately and in proportion to the population of each province. At the first stage 1,000 clusters with probability proportional to size (390 in urban and 610 in rural areas) were selected. In Punjab, Sindh, Khyber Pakhtunkhwa and Baluchistan provinces, 440, 260, 180 and 100 clusters were selected, respectively. In urban areas, the clusters were selected from 26,800 enumeration blocks, each including 200 – 250 households. A list of 50,588 villages enumerated in the 1998 population census was used to select clusters in rural areas. In the second stage, 105 households were selected randomly and 10 households in each sampled household were selected using a systematic random sampling technique to conduct interviews with married women of reproductive age [15].

PDHS 2006–07 used six types of questionnaires, the Community, the Short Household, the Long Household, the Women’s, the Maternal Verbal Autopsy, and the Child Verbal Autopsy Questionnaires. The contents of the Household and Women’s Questionnaires were based on model questionnaires developed by the Measure DHS program. The Household Questionnaire listed all the usual members and collected basic information such as age, sex, marital status, education, relationship to the head of the household and household level characteristics. The Women’s Questionnaire collected information from ever-married women 12 – 49 years of age on background characteristics (education, literacy, native language, marriage characteristics etc.), full birth history, history of antenatal care (ANC) for the most recent birth within five years preceding the survey, delivery and postnatal care for all births as well as the survival of live-birth infants.

In the current PDHS 2006–07, 98% of the 97,687 available households were successfully interviewed, and 10,023 women were interviewed, which was 95% of the 10,601 eligible women. Multiple pregnancies (n = 65) were excluded from the analysis due to higher odds of newborn deaths associated with prematurity and pregnancy complications among them [18]. Only last live births in the last five years (2002–2006) preceding the survey were selected for the current study to avoid the violation of the independence assumption and reduce potential recall bias. Further, for last live births in the last five years preceding the survey, antenatal care, delivery, and postnatal care services details were also recorded. A verbal informed consent was taken from each respondent before the interview. The study protocol was approved by the ethics committee of the University of Sydney, Australia.

The conceptual framework proposed by Mosley and Chen [19] with modifications based on the limitations and structure of the DHS data was used (Figure 1).

We defined the death of a live born baby in the first month of life (birth to 30 days) as neonatal mortality. The neonatal mortality rate was calculated as the number of deaths in the first month of life per 1,000 live births. The outcome was neonatal deaths recorded as a binary variable. The age of neonatal death was measured in days and for deaths within 24 hours a value of 0.01 days was used. A list of explanatory variables along with their definitions and categories is presented in Table 2.

Two variables were included at the community level, which were cluster type and region. Six variables were included in the socio-economic status, which were maternal marital status, polygynous marriage, parental education, parental occupation, household wealth index and fuel used for cooking at home. The household wealth index was calculated using an inventory of households assets including presence of television, radio, refrigerator, electricity, type of toilet, condition of housing, and ownership of vehicles, which were weighted using principal components analysis method [20].

The proximate determinants at the individual level were maternal age at childbirth (maternal factor), baby’s gender, mother’s perception of birth size, birth weight and a combined variable of baby’s birth rank and birth interval (neonatal factor), mother’s desire for pregnancy (pre-delivery factor), ANC services by type of providers, antenatal iron-folic acid (IFA) supplements used, delivery complications, delivery assistance, mode of delivery, place of delivery (delivery factor), and postnatal care as a post-delivery factor.

Data analyses was conducted by using STATA 13.1 (Stata-Corp, College Station, TX, USA) with  Svy’ commands to allow for adjustments for the cluster sampling design used in the survey. The frequencies along with weighted percentage were calculated for the selected variables. The neonatal mortality rate was calculated for each category, followed by Cox proportional hazard analysis to identify potential determinants of neonatal mortality without adjusting for other covariates.

All variables with a p-value ≤ 0.2 were included in a multivariate Cox proportional hazard models, which were constructed using step-wise backward elimination procedures. However, variables like ANC services by type of providers and delivery assistance were considered as a priori in the models due to their strong association with neonatal mortality. Hazard ratios (HR) were estimated as the exponential of the regression coefficients, and 95% confidence interval (CI) for the HRs were calculated. Two variables, birth weight and postnatal care, because of higher proportions of missing values were excluded from the analysis.

The current study found that male infants, smaller than average birth size and delivery complications were significantly associated with higher risk of neonatal mortality in Pakistan. Other factors which were associated with neonatal mortality were: infants who were living in province Punjab, first rank baby and belonging to the poor household wealth index quintile. Our findings are important because the level of neonatal mortality in Pakistan is unacceptably high, 54 per 1,000 live births, and the current rate of decline will be insufficient for the country to reach its child survival MDG. Detection of these risk factors for neonatal deaths will help to formulate strategies and program innovations to improve neonatal survival.

We used the nationally representative sample for the current study, which was considered as the major strength of our study findings. Further, PDHS 2006–07 had high response rates at household and individual levels. We used infant survival data for the last five years to reduce recall errors about dates for births and deaths by interviewed mothers [21‐23]. Moreover, we performed Cox proportion hazard models to identify determinants of neonatal mortality, which is a standard method for dealing with censored failure time data and has been widely used in biomedical research [24].

However, in PDHS 2006–07 only surviving mothers were interviewed, which was one of the limitations of the current study. It may have led to an underestimate of the neonatal mortality rate, because of the association of neonatal deaths with maternal deaths, and could also have led to an underestimate of the effect of some of the associated factors, such as delivery complications [22]. Further, some variables, such as parental occupation which represented the employment status within the last 12 months preceding the survey, were not infant specific because these only presented the most recent conditions. We did not consider the initiation of breastfeeding variable because of the low number of neonatal deaths in the late neonatal period (8–30 days), which was hypothesized as the time when colostrum would start to provide protection to the infant for infectious diseases. The variables like birth weight and postnatal care were excluded from the multivariate analysis due to high proportions of missing values. Genetic and some environmental factors which are also possible determinants of neonatal mortality were not available in the PDHS dataset. Nevertheless, these limitations were unlikely to have had an important influence on the validity of our findings.

The current study revealed that smaller than average birth size was an independent risk factor for neonatal mortality in Pakistan. We did not use the birth weight variable in the analyses as only 12% of infants were reported to have been weighed at the time of birth in the survey. We considered smaller than average birth size as a proxy variable for low birth weight. Our findings agree with several other studies which have identified low birth weight as a risk factor for neonatal mortality [2, 25‐29]. An in-depth analysis of 2006–07 PDHS data by National Institute of Population Studies (NIPS) Pakistan reported a strong association between small birth size and neonatal mortality. However, the authors did not adjust the analysis for antenatal, delivery and postnatal care variables [30]. Low birth weight arises through preterm birth or in-utero growth restriction, or both [31]. In a hospital based retrospective study from Pakistan, 68% of mortality in newborn was contributed by low birth weight, 74% of them being preterm, suggesting high mortality among low birth weight-preterm infants [32]. Preterm births with complications are considered as the leading cause of neonatal deaths in Pakistan [1, 8].

Infants whose mothers had delivery complications had a higher risk of neonatal death in our sample. The delivery complications included vaginal bleeding, presence of fever or convulsions, and these complications need to be managed by a skilled birth attendant in a well equipped health facility. However, only 36% of deliveries were conducted by health professionals in our sample. Research has shown that deliveries in a health facility with a skilled birth provider reduces early neonatal deaths [33, 34].

Male gender was identified as a predictor of neonatal mortality in Pakistan. Several studies have identified male gender as a risk factor for neonatal deaths [25, 35‐38]. Biologically male children have higher risk of getting infectious diseases due to higher prevalence of immune deficiency [37], higher prevalence of respiratory illnesses and congenital malformations of urogenital system due to late maturity [38], probably all these lead to higher neonatal mortality among males.

Our study highlighted an association between first birth order and neonatal mortality in Pakistan. Previous studies have reported a U-shaped curve relationship between birth order and neonatal mortality as neonatal mortality tends to be higher for the first born child, and higher order births of order 4 and above compared with second and third order births [39‐41]. First birth infants were at a 33% increased risk of dying during the neonatal period than births of second through third order reported by others [41]. There is a biological basis for the poor survival experience of first births during the neonatal period. First births in developing countries take place before a woman has reached full physical and reproductive maturity. Furthermore, it could be due to poor preparation by first-time mother to handle new roles and responsibilities in her life [40].

The poorest household wealth index quintile was identified as a risk factor for neonatal mortality in our analyses. Similar to our finding, secondary analysis of Sudan DHS has also found lower household wealth index as a risk factor for neonatal mortality [42]. Household poverty has been reported to increase neonatal mortality either by increasing the prevalence of risk factors like maternal infections or through reducing access to effective care [2].

Our sample also showed regional variation in neonatal mortality in Pakistan. In this context, living in Punjab province was identified as a predictor of neonatal mortality in Pakistan. Punjab is the most populated province of Pakistan with rural and urban population. In the Punjab, there are about 2,800 primary health facilities (basic health units and rural health centres) in the public sector serving the rural communities with trained staff and heavily subsidized medicines. Nevertheless, these facilities are distributed unevenly compared to the population catchment area. Hence, there are more doctors per facility in some areas at the expense of others. In 2011 a survey [43] was conducted to evaluate the current status of basic health units in Punjab. Out of 850 selected basic health units, 7.2% were closed, and 52.4% did not have the essential staff at the time of survey. Further, the monitoring system of these facilities was found to be weak and non-availability of drugs at these facilities were also considered a major problem [43]. The Punjab has had the highest neonatal mortality rates over the last two decades (Table 1) despite an increase in the percentage of women receiving ANC and being delivered by a skilled provider. There could be several reasons to explain this discrepancy, such as lack of staff at the primary level health facilities, and shortages of essential drugs. However, we also analysed the subset of the sample from PDHS 2006–07 of babies whose mother were living in the Punjab and found that smaller than average birth size and delivery complications were significantly associated with higher risk of neonatal mortality (data not shown). Hence, there is a need to conduct further research to examine factors associated with neonatal mortality in the Punjab.

Infants whose mothers were not working and father was employed had lower risk of mortality in neonatal period in the current study compared to both working. Involvement of mothers in work outside home may adversely affect the care provided to the newborn. Others have reported an increased odds of neonatal deaths due to lack of personal and timely care including infrequent breastfeeding experienced by infants born to working mothers [44]. However, parental employment status in PDHS only showed working during the last 12 months preceding the survey.

The current study findings have implications for the maternal and newborn health programs in Pakistan. To achieve national MDG-4 target, it is important to reduce neonatal deaths substantially as 57% of under-five child deaths are in neonatal period in Pakistan [15]. In the light of the current analyses, there is a need to implement cost effective interventions through community based trials to see their effectiveness in the Pakistani context. However, at the same time, further research in various regions of Pakistan is required to indentify the obstacles, such as socio-cultural issues, practices etc., for poor newborn health and survival.