Institutional maternal and perinatal deaths: a review of 40 low and middle income countries

Post Millennium Development Goal global action agendas such as the Sustainable Development Goals (SDGs), Every Newborn Action Plan (ENAP) and Ending Preventable Maternal Mortality continue to measure global progress to reduce the maternal mortality ratio (MMR), the neonatal mortality rate, and now (under ENAP guidance) the stillbirth rate [1‐3]. Understanding the magnitude and clinical causes of maternal and perinatal mortality are basic requirements for policy setting, program design, innovation testing, and the implementation of evidence-based interventions. Understanding maternal and newborn outcomes captured at health facilities presents an opportunity for health care staff and decision-makers to reflect on what they could do better.

High quality data on how many maternal and newborn deaths occur and their causes are notoriously difficult to come by and global estimates come from complex models based on multiple sources: vital registration data, confidential enquiries, large household surveys, reproductive-age mortality studies, research reports, surveillance data, and verbal autopsies [4‐8]. Over the last few decades the Maternal Mortality Estimation Inter-Agency Group produced a series of maternal mortality estimates [7, 9, 10], the Global Burden of Disease Studies contributed important systematic analyses of trends, projections, and causes of maternal and child mortality [6, 11, 12], while the World Health Organization (WHO) produced two systematic analyses of the global causes of maternal death [13, 14]. The authors of these comprehensive systematic reviews shy away from using routine health facility data because of inherent selection bias. However, in the 2003–2009 WHO systematic analysis, the authors consulted health facility data when the institutional delivery rate in that country was 50% or greater during the period reviewed [14].

Globally, the coverage of skilled attendant at birth increased from 59% in 1990 to 71% in 2015 [15]. More countries are adopting a 100% institutional delivery policy and institutional delivery rates are rising. This shift means that proportionately more individuals with peripartum and perinatal complications will access treatment, and mortality events, when they happen, are more likely to occur in facilities than at home. In low and middle income countries facility-based maternal and perinatal mortality figures do not yet substitute population-derived estimates as they reflect only those women and newborns who succeed in accessing facility-based care. Facility-based events are highly specific to local contextualized conditions, and thus, are well-suited to inform local policy makers, clinicians and programs to target specific health system strengthening efforts. Most importantly, they can be used to improve service quality. As the SDGs bring a renewed global focus on improving the quality of routine health management information systems, through the Health Data Collaborative and other initiatives, and as they include the use of new technologies, data quality and availability will increase and the cost of collecting data will decrease.

This paper reviews data from up to 40 low and middle income countries and describes the magnitude of institutional maternal mortality, causes of maternal death and cause-specific case fatality rates, as well as institutional stillbirth and early or pre-discharge neonatal death rates, in most cases, at the national level. This analysis draws on reports produced over the last decade.

This secondary data analysis is based on a review of cross-sectional health facility surveys known as emergency obstetric and newborn care (EmONC) assessments, which focus on routine intrapartum care for women and their newborns as well as more complicated births. These assessments have been driven by the United Nations Fund for Population (UNFPA), the United Nations Children’s Fund (UNICEF), the WHO, and the Averting Maternal Death and Disability (AMDD) program at Columbia University. The methods have been described elsewhere, but a summary follows [16, 17].

Up to 40 country reports (including Zanzibar) provided the number of maternal deaths that took place within health care institutions, 31 from sub-Saharan Africa, and the remaining nine from Latin America and the Caribbean and Asia (Table 1). The scope of EmONC assessments ranged from all nine hospitals in the province of Azuay, Ecuador to 1626 facilities in Burkina Faso, inclusive of all facilities with at least one delivery in the past 12 months. The total number of facilities (15,411) registered 6.5 million deliveries and 17,314 maternal deaths. To contextualize the number of institutional maternal deaths and associated MMR, we included the institutional delivery rate and the percentage of institutional deliveries with a major direct obstetric complication. Both indicators were derived from EmONC assessment data. The final column includes the 2015 population-based MMR estimated by the Maternal Mortality Estimation Inter-Agency Group [7], also included for context, although most assessments occurred before 2015. Institutional delivery rates ranged from 7% in Ethiopia in 2008–9 to 113% in Mongolia (likely explained by a non-standard sampling strategy of 21 hospitals and their catchment areas). The next highest institutional delivery rate was 88% in São Tomé & Príncipe. Institutional MMRs ranged from 2130 maternal deaths per 100,000 deliveries in Chad to 32 in Mongolia.

Countries with relatively low coverage of institutional deliveries such as Haiti, Niger, Sierra Leone, Ethiopia, South Sudan, Angola, and Chad tended to have high institutional MMRs, suggesting that a disproportionate number of women delivering in facilities experienced serious complications. To some extent, the percentage of deliveries with major obstetric complications supports this pattern where high percentages were found in countries with high institutional MMRs. However, countries such as the Democratic Republic of Congo or Afghanistan also exhibited relatively low institutional delivery rates, 10% or more of deliveries with complications, and had institutional MMRs of less than 200, making it difficult to discern any robust pattern. A high percentage of complicated deliveries could also reflect the type of facility surveyed, e.g. Ecuador (30%) and Mongolia (28%), where only hospitals were assessed.

This institutional assessment compiles data from many countries where every country set out with similar objectives, used a similar methodology and data collection instrument, and had common indicators. By design, each assessment captured a complete recording of all maternal deaths by cause and common perinatal outcomes. This is a strength that other multi-country studies have not shared. In this overview, we gathered service statistics from more than 15,400 health care facilities that mirror findings from more complex modeling exercises, regardless of differences in methodology and reference populations.

We saw hypertensive diseases as the predominant cause of institutional maternal death in the Latin America and Caribbean region and hemorrhage highlighted in Asian countries, despite the small number of surveys in each of those regions. Meanwhile, hemorrhage was the predominant cause of institutional maternal deaths in sub-Saharan African countries, a region also distinguished by its large proportion of indirect maternal deaths.

In Table 4 below we compare the overall distribution of institutional causes of 14,785 deaths from 26 sub-Saharan African countries with the population-based distribution found in the WHO 2003–2009 systematic review (in both cases, unknown causes of death were excluded). Cases of ruptured uterus and obstructed labor were included in “other direct causes” for the EmONC Assessments while these cases were likely assigned to hemorrhage or sepsis in the WHO study [14]. The degree of similitude in the distribution of causes is both validating and reassuring but may also point to possible data quality issues and/or differences between all deaths versus just those occurring in facilities.

The larger proportion of indirect causes found in this paper is noteworthy but it also may be underreported especially where comorbidities were common. During the training, data collectors were instructed to classify a maternal death as direct if there was evidence of both direct and indirect causes. For programmatic purposes, indirect causes of maternal mortality require a greater focus of attention, not just for purposes of reporting but also for health service delivery organization to intervene early to prevent these deaths.

We also observed that institutional stillbirth rates tended to be substantially higher than early neonatal death rates, and that countries with high institutional stillbirth rates also tended to exhibit high institutional MMRs. According to other studies, we might have expected the ratio of stillbirths to early neonatal deaths to be approximately 1.3 to 1, but these institutional data suggest a lower ratio, i.e., more stillbirths than expected [20, 21]. Unfortunately, given the uneven reporting of whether the stillbirth was macerated or intrapartum, the often-cited ratio of 1 intrapartum stillbirth to 3 macerated stillbirths could not be assessed [22].

The 2030 ENAP target for the stillbirth rate is 12/1000 total births and the target neonatal death rate is the same, but among live births. At this time, most countries in this overview are far from reaching the stillbirth target and many countries would fail to reach the neonatal target of 12, although this is more difficult to ascertain given the censoring of data since so many women and their newborns are discharged within 12 h of delivery. A recent six-country study of early neonatal mortality showed that neonatal deaths in the first six and 24 h account for one-third and 46%, respectively, of all neonatal deaths [23]. Therefore, a doubling or tripling of the early neonatal deaths observed in this overview might provide a rough estimate of the actual neonatal death rate. But like the MMR, it is unclear whether institutional rates and ratios are likely to be higher or lower than the population-based rates. Nevertheless, high stillbirth rates observed in the EmONC assessments give pause; the global stillbirth rate for 2015 was 18.4 per 1000 births, while the rate for sub-Saharan Africa was 28.7 [4]. According to the authors of recent trend data for stillbirth rates, when compared to high quality vital registration data, facility data tend to overestimate the stillbirth rate due to selection bias [4, 5].

Despite evidence for reductions in maternal and perinatal mortality over the last two decades, this multi-country overview leads to recommendations for clinical practice and policy if we are to move towards the goal of ending preventable maternal and newborn deaths. From the clinical perspective, although fewer in absolute numbers than hemorrhage or hypertensive disorders, uterine rupture and maternal sepsis were the most lethal complications. The literature consistently shows the elevated risk of mortality from ruptured uterus [24‐27]. High case fatality rates for uterine rupture suggest poor diagnostic skills, inadequate patient monitoring after admission and delays in appropriate treatment [28], or perhaps inappropriate or overuse of augmentation or induction. Several studies point to high rates of rupture after admission [25, 29]. Ruptured uterus is also an indication that women with obstructed labor or at risk of rupture, e.g. having a previous uterine scar, still experience difficulties in accessing surgical care in a timely manner.

Considerable international investment has focused on reducing deaths due to hemorrhage and hypertensive disorders, given how many deaths are attributable to these complications. Both have well-known pharmacological solutions as well as effective preventative measures with active management of the third stage of labor and the potential to detect high blood pressure and proteinuria during antenatal care. Ruptured uterus might be viewed as requiring more complex multi-sectoral fixes – improved road networks, better communication and transportation options, as well as the human resources who can and will monitor the progression of labor, follow protocol, and perform cesarean delivery. Sepsis may require more of a professional culture change towards infection prevention, more accessible water, sanitation and hygiene infrastructure as well as antenatal screening.

To optimize the investment of an EmONC assessment, it should be followed by multilevel planning and implementation phases. In 2016, only six of the countries mentioned in this publication have set up such processes that include maternal and newborn care monitoring in EmONC facilities (Burkina Faso, Cambodia, Haiti, Madagascar, Niger and Togo). However, this number is likely to increase significantly in 2017. The production, analysis and utilization of data by providers with the support of coaches also contribute to improve quality of care.

As skilled delivery coverage increases and maternal mortality declines, women who die in facilities may no longer represent the tip of an iceberg, but most maternal deaths. With appropriate reflection, institutional stillbirth and early neonatal death rates, causes of maternal death and case fatality rates can guide management on how to improve health workers’ capacity to meet the demand for emergency care, including record-keeping, and identify hotspots of where and what is needed to reduce delays in seeking, reaching and receiving care. Facility-level data will become all the more important and thus efforts to improve data quality are crucial.