Introduction

Global efforts to reduce maternal and newborn deaths are expected to fall short of targets set out in the 2030 Sustainable Development Goals (SDGs) [1–3]. In Pakistan, rates of neonatal death (42 per 1,000 live births) and stillbirth (31 per 1,000 total births) remain second highest amongst all nations [4–6]. Similarly, Pakistan’s maternal mortality ratio (186 maternal deaths per 100,000 live births) remains more than double the SDG target (<70 per 100,000) [7]. The majority of maternal, fetal, and newborn deaths in Pakistan are preventable, particularly in remote areas where access to health facilities, skilled birth attendants, and both antenatal and postnatal care are limited [8–10]. For instance, non-sterile delivery environments and lack of umbilical cord care increase the risk of neonatal sepsis, while challenges in identifying early warning signs of infection or hypothermia may result in delayed healthcare seeking and adverse outcomes. To reduce the burden of preventable deaths in the most remote areas, low-cost interventions that can be delivered at-scale through community-based healthcare service models are needed to both prevent serious conditions of newborns and mothers and to provide early warning should these conditions occur [11,12].

Between 2022–2024, a cluster randomized controlled trial (cRCT) in Gilgit-Baltistan, Pakistan aimed to estimate the effect of an integrated newborn care kit on maternal and newborn health outcomes, primarily neonatal mortality (Clinicaltrials.gov NCT04798833). The intervention is designed to be integrated into existing community healthcare infrastructure, and includes tools to be utilized throughout the neonatal period to prevent newborn infection (e.g., clean delivery kit, chlorhexidine gel, sunflower oil emollient), identify and manage hypothermia and fever (ThermoSpot thermoindicator sticker, blanket, and portable heat pack), and teach caregivers on each component’s appropriate usage (pictorial guide) [13]. The kit also includes oral misoprostol tablets to prevent postpartum hemorrhage (PPH), the leading cause of maternal death [14]. The cRCT aimed to enroll >27,000 pregnant women from 2022–2024, with both the intervention and control (local standard of care) delivered by Lady Health Workers (LHW). LHWs are government-sponsored community health workers who are trained to provide reproductive, maternal, newborn, and child health (RMNCH) education as well as basic preventive and curative care [15]. Household LHW coverage has previously been associated with improved RMNCH outcomes [15,16].

In this study, we aimed to describe baseline rates of maternal and newborn health outcomes, determine their association with household LHW coverage, and estimate intracluster correlation coefficients in Gilgit-Baltistan, Pakistan to inform the design of future cluster randomized trials in this population.

Materials and methods

Study design and procedures

We conducted a cross-sectional, representative household survey between June 14–August 31, 2021. The primary objective of this survey was to collect baseline outcome parameters to inform the cRCT. Specifically, baseline data regarding cluster-level neonatal mortality and health facility deliveries were utilized as inputs for the trial’s covariate constrained randomization, and will also be adjusted for in the trial’s pre-specified multivariable analyses [13].

The baseline survey was conducted through a two-stage, stratified sampling design (Fig 1). The sampling frame included villages with pre-existing LHW coverage, and selection was stratified by each of the 77 clusters available for randomization in the cRCT. Clusters were defined as Union Councils (or sub-Union Councils), which are geographic subdivisions that are directly involved in local government health administration. To estimate neonatal mortality with equal precision across all clusters, three villages were selected from each cluster with the probability of selection proportional to village population size. Within each selected village, data collection teams comprising three data collectors and one supervisor conducted a standard household listing procedure to enumerate and map occupied households. Forty-eight households per village were selected (or fewer if <48 occupied households were enumerated), using an equal probability systematic selection approach.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study sampling design and flowchart of households included in baseline survey.

Trained data collectors then visited households to administer the study questionnaire; one representative per household was asked to participate and self-report all elements of the questionnaire. The questionnaire was administered primarily in Urdu; however, at least one member of each data collection team was proficient in the local language of each village visited, and translated the study questions as needed into Balti (spoken in Baltistan), Brushiski (spoken in Nagar), or Shina (spoken in Astore). Participants were asked about: a) household reproductive history, including the number of women who gave birth, the number who experienced PPH, and the number of live births and stillbirths delivered; b) live birth history, including the time and place of birth and any neonatal deaths; c) LHW household coverage, including the number of total visits, visits for prenatal care, and visits for postnatal care; and d) household experiences with COVID-19, which will be reported elsewhere. Respondents were asked to report reproductive and live birth outcomes in the past five years, and LHW household coverage in the past 12 months. Finally, quality control officers recontacted approximately 18% of households to verify accuracy of the data collection.

https://doi.org/10.1371/journal.pgph.0002693.g001

We calculated the required sample size of the survey by assuming an NMR of 35 per 1,000 live births, desired precision of 6 per 1,000 live births, and design effect value of 2.03 [5,18]. To achieve this level of precision, we estimated that 7,619 women would need to participate from 11,231 households, assuming a similar response rate and proportion of eligible households as the 2017–18 Pakistan Demographic and Health Survey [5].

Results

In total, 10,264 households agreed to participate in the survey, including 62.7% with ≥1 pregnant women and 62.1% with ≥1 live birth in the past five years (Fig 1, Table 1). Pregnancy histories were collected for 7,600 women, while 12,376 births were recorded including 11,863 live births and 513 stillbirths (Table 2). The median household size was seven occupants (IQR 5–9). Overall, 66.2% of households were visited by LHWs in the past year but this differed by district, ranging from 20.9% in Diamer to 83.5% in Nagar. Among households with any LHW visits, the median number of home visits was five (IQR 4–7). Among households with any pregnant women, 52.3% were visited by LHWs who provided prenatal care, while 77.4% of households with any live birth were visited by LHWs who provided postnatal care.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Household characteristics at baseline.

https://doi.org/10.1371/journal.pgph.0002693.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Five-year reproductive, delivery, and newborn outcomes at baseline.

https://doi.org/10.1371/journal.pgph.0002693.t002

A history of PPH was self-reported by 13.4% of women who had given birth in the past five years (Table 2). Meanwhile, 30 maternal deaths were reported, corresponding to a maternal mortality ratio of 225 deaths per 100,000 live births (95% CI 137–369 per 100,000). Prenatal LHW home visits were associated with a significant 11% reduction in PPH per each additional visit (e.g., one vs. zero visits, two vs. one visit, etc.) (risk ratio [RR] 0.89, 95% CI 0.83–0.96, Fig 2). The association between prenatal LHW home visits and maternal death was RR = 0.79 per each additional visit (95% CI 0.48–1.30).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Associations between maternal, delivery, and newborn outcomes with village-level frequency of Lady Health Worker home visits.

Each line represents the unadjusted risk ratio between prenatal or postnatal LHW visits and the outcome of interest, from an unadjusted mixed effects Poisson regression. Clustering at the Union Council and village levels were accounted for using Stata’s svyset command, while within-household clusters was accounted for by including household ID as a random effect in the regression. We assessed for non-linearity by including quadratic terms for the number of LHW visits–interaction terms were removed if not significant using a Wald test. Solid lines indicate a p-value <0.05, while dashed lines indicate p≥0.05.

https://doi.org/10.1371/journal.pgph.0002693.g002

From all reported births in the past five years, 4.1% were stillborn, corresponding to a stillbirth rate of 41.4 per 1,000 births (95% CI 36.8–46.7). Among all live births, 64.6% were delivered at a health facility and 204 neonatal deaths (1.6%) were reported, including 151 early neonatal deaths (i.e., first seven days of life; 77.5%) and 53 late neonatal deaths (i.e., days 8–28 of life; 22.5%). The NMR was 16.2 per 1,000 live births (95% CI 13.6–19.3), early NMR was 12.4 per 1,000 live births (95% CI 10.1–15.2), and late NMR was 3.7 per 1,000 live births (95% CI 2.6–5.3). Meanwhile, the perinatal mortality rate was 53.0 per 1,000 total births (95% CI 47.6–59.0). Postnatal LHW home visits were associated with a significant 20% reduction in late neonatal mortality per each additional visit (RR 0.80, 95% CI 0.67–0.97, Fig 2). The magnitude of association was smaller between LHW home visits and stillbirth (RR 0.95 per visit, 95% CI 0.88–1.03), neonatal death (RR 0.95 per visit, 95% CI 0.87–1.03), early neonatal death (RR 0.99 per visit, 95% CI 0.86–1.14), perinatal mortality (RR 0.97, 95% CI 0.90–1.04), and health facility delivery (RR 0.98 per visit, 95% CI 0.95–1.02). Across the five-year recall period, the greatest number of neonatal deaths occurred in the month of January (15.9%) while the fewest occurred in May (3.7%; S1 Fig). Neonatal mortality rates were similar between children born at home versus those born in a health facility (S1 Table).

Intracluster correlation coefficients (ICC) and coefficients of variation (k) are shown in Table 3. There was a high degree of clustering for health facility deliveries, with 24% of the variation explained by Union Councils and 27% explained by villages. ICC values were lower for mortality and maternal outcomes. Covariance between all RMNCH indicators, aggregated to the cluster level, are shown in S2 Table. Rates of PPH were significantly correlated with maternal death (ρ = 0.26), perinatal mortality (ρ = 0.29), neonatal mortality (ρ = 0.35), and early neonatal mortality (ρ = 0.34).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Intracluster correlation coefficients (ICC) and coefficients of variation (k).

https://doi.org/10.1371/journal.pgph.0002693.t003

Discussion

In this representative, cross-sectional study of >10,000 households in Gilgit-Baltistan, we identified a high burden of maternal, perinatal, and neonatal death. Despite some variability between study districts, six districts exhibited higher maternal mortality ratios (75–580 deaths per 100,000 births) than the 2030 SDG target (<70 deaths per 100,000 live births) while five districts exhibited higher NMR than the SDG target (<12 deaths per 1,000 live births) [2,3,9]. Local indicators of RMNCH including postpartum hemorrhage, maternal death, and perinatal death were also correlated with one another, suggesting some clustering of the poorest outcomes. These areas are most likely to benefit from targeted, community-based interventions such as those delivered in the cluster randomized trial.

We estimated a higher maternal mortality ratio in Gilgit-Baltistan (225 deaths per 100,000 live births, from pregnancy-related causes within 45 days of delivery) than the 2019 Pakistan Maternal Mortality Survey (157 per 100,000 live births, from any cause except accidents or violence, within 42 days of delivery) [7]. Indeed, the highest MMRs occurred in the most remote districts of Diamer (356 per 100,000), Kharmang (580 per 100,000), and Shigar (496 per 100,000), supporting distance to health facilities as a primary contributor to mortality risk [22,23]. We also estimate that approximately one in eight women (13.4%) experienced PPH, the leading cause of maternal death globally and in Pakistan [14,24]. Few studies have estimated the rate of PPH in Pakistan. However, a 2005–2006 cross-sectional household survey found that 21.3% of women experienced PPH in Khyber Agency, and similarly a randomized trial in Chitral, Khyber Pakhtunkhwa found that 21.9% of women in the control arm experienced PPH [25,26]. Importantly, the diagnosis of PPH is limited in community-based settings where medical attendance of births is not universal, suggesting that data sources relying on self-reported perceived excessive blood loss may overestimate the true burden of PPH. Indeed, a facility-based study of 16 referral healthcare facilities across Pakistan found that only 1.6% of women experienced PPH [27]. Nevertheless, home-based delivery of interventions which reduce the risk of PPH such as oral misoprostol is warranted and has been demonstrated to have efficacy without associated safety concerns [26,28].

Compared to the 2017–18 Pakistan Demographic and Health Survey (DHS), we found lower neonatal mortality (16.2 deaths per 1,000 live births vs. 47 deaths per 1,000 live births in the DHS) but higher perinatal mortality (53.0 deaths per 1,000 total births vs. 43 deaths per 1,000 total births in the DHS) in Gilgit-Baltistan [5]. This is suggestive of a high degree of pregnancy outcome misclassification in our study, whereby early intrapartum neonatal deaths were self-reported as stillbirths, particularly among home births [29,30]. This conclusion is further supported by our high stillbirth estimate (41.4 stillbirths per 1,000 total birth) relative to other Pakistani data sources (30.6 stillbirth per 1,000 total births) [31]. Nevertheless, it remains possible that true declines in neonatal mortality have occurred, alongside those occurring at the national level between the 2012–13 DHS (55 deaths per 1,000 live births) and 2017–18 DHS (42 deaths per 1,000 live births) [5]. Indeed, prior regional estimates from a quasi-experimental study in Gilgit estimated NMRs of 26.0 (in urban areas) and 39.8 (in rural areas) per 1,000 live births from 2002–2003, and declines over the ensuing decades are plausible [32]. Finally, while a third of households lacked any recent LHW coverage, restriction of the sampling frame to villages with anticipated LHW coverage may also have contributed to the lower mortality estimates.

Increased frequency of LHW home visits was associated with significantly lower risk of PPH (11% reduction per additional visit) and late neonatal mortality (20% reduction per additional visit). Moreover, LHW home visits were associated with a 21% reduction in maternal death per additional visit, though this relationship was limited by a small number of maternal deaths and was not sufficiently powered to detect statistical significance. We hypothesize that LHWs are well equipped to identify early warning signs associated with these conditions such as prenatal bleeding, abdominal pain or dizziness, neonatal infection, and hypothermia. This aligns with LHW program policy, whereby LHWs aim to conduct between 4–12 home visits during pregnancy, including for identification of danger signs and health facility referrals [16]. LHWs also aim to complete a first mandatory postnatal visit within the first week of life, plus an additional 1–2 home visits per month to assess the child’s nutrition and weight status and deliver first polio vaccinations. In this study, we found a median of four postnatal care visits among households with any LHW coverage, perhaps indicating high motivation among LHWs to provide additional care during this critical period [33]. LHW home visits were not associated with declines in perinatal mortality, reflecting the need for skilled birth attendance to address intrapartum complications. These findings are consistent with a household survey among >26,000 children in Balochistan, Punjab, and Sindh provinces, where LHW covered households experienced reduced post-neonatal mortality and similar rates of neonatal mortality compared to LHW uncovered households between 2015–2017 [34]. Importantly, we were unable to conduct multivariable analysis, and therefore these findings may be confounded by factors such as distance to the nearest health facility and maternal education. Nevertheless, several prior studies have shown improved RMNCH outcomes in LHW covered households, as well as multiple community-based trials demonstrating effective interventions (e.g., on mortality, nutrition, development) when delivered by LHWs [16,35,36].

The intracluster correlation coefficients and coefficients of variation estimated in this study help to characterize between-cluster variability and may be utilized by researchers for sample size estimation for outcomes related to maternal, perinatal, and neonatal health. Representative baseline data such as these should be used to aid in study planning of comprehensive, high-quality designs.

This study has several limitations. Firstly, the sample frame for the study was limited to areas with anticipated LHW coverage (i.e., to match the delivery of trial interventions) and therefore may exclude more remote areas where RMNCH outcomes are poorest. This may partially explain differences between our study and other household surveys. Moreover, only three Union Councils in Diamer district were included in our study (Goharabad, Gonarfarm, and MC Chilas) and findings may not be representative of the entire district. Second, PPH was collected as a cumulative measure and may not approximate a true incidence rate. Third, place of birth was not collected for stillbirths, and we were therefore unable to further assess differential pregnancy outcome misclassification. Fourth, recall periods for maternal and newborn outcomes (five years) differed from LHW home visits (12 month), and we did not collect data on possible confounders such as distance to the nearest health facility. Fifth, all elements were ascertained by self-report and may have been affected by recall bias, particularly outcomes which used a five-year recall period. Finally, the study recall period included the first 1.5 years of the COVID-19 pandemic, a period when antenatal and perinatal care services may have been less commonly available and when an unknown proportion of the population were infected with SARS-CoV-2. We are unable to comment on what degree these factors affected the maternal and newborn outcomes presented in this study.

Conclusions

In this study, we confirmed high rates of maternal, perinatal, and neonatal death in Gilgit-Baltistan, Pakistan. Increased frequency of prenatal and postnatal LHW home visits were most strongly associated with declines in postpartum hemorrhage, maternal death, and late neonatal mortality. These findings may be used to plan ongoing and future studies of RMNCH outcomes, and emphasize the need for scalable, low-cost, and community-based interventions to reduce the burden of preventable mortality in the region.

Supporting information

Acknowledgments

We gratefully thank the data collectors, field supervisors, drivers, and other team members for their commitment to this study and for tirelessly administering this survey. We also thank the participating families who welcomed our teams into their homes and shared their experiences.