The Heartbeat Nobody Wrote Down
Expert panels assigned a cause to 94% of 2,492 stillbirths across seven CHAMPS sites. In the facilities where almost all of those babies were delivered, the file often lacked the observations that would have made prevention possible.
Ninety-five per cent of the stillbirths in a study published on 27 October were delivered in a health facility. That single number reframes the way the paper is usually read. This is not a story about families who never reached care. Across seven countries in sub-Saharan Africa and South Asia, between December 2016 and December 2023, teams from the Child Health and Mortality Prevention Surveillance network (CHAMPS) investigated 2,492 stillbirths that, in the overwhelming majority of cases, happened under the eyes of health workers.
The study, in PLOS Global Public Health, is the largest systematic post-mortem investigation of stillbirth causes yet assembled: 4,876 eligible deaths approached, 2,731 sampled, 2,492 with a completed cause. Local expert panels named a fetal or maternal condition in 94% of them. What they could not always do was say what the clinical record should have told them.
What the panels found
Each case reached a determination panel — neonatologists, obstetricians, pathologists, microbiologists, epidemiologists — carrying histology from minimally invasive biopsies of lung, brain and liver, multiplex PCR and culture results, placental sections where a placenta existed, photographs, abstracted antenatal records, and the findings of a verbal autopsy interview conducted with the 2016 WHO instrument. Causes were coded in ICD-10 and in the WHO application of ICD-10 to perinatal deaths, with fetal and maternal conditions recorded separately along the causal chain.
Intrauterine hypoxia dominated: three-quarters of the series, at every site. In jargon this is close to a description of a death rather than a disease — no specific fetal abnormality, but a pregnancy or delivery that cut off the supply. Where the panels could attribute it, a maternal condition explained 49% of those deaths and a placental cause 33%. A primary maternal condition was identified in 72% of the whole series, most often placental complications (18%) and maternal hypertension (17%).
The rest of the distribution is where geography shows up. Congenital infections accounted for about 9% of stillbirths, but 31% at the Soweto site in South Africa, with group B streptococcus, Escherichia coli and Enterococcus faecalis the commonest organisms. Congenital birth defects were likewise around 9% overall and 24% in Ethiopia. Complications of placenta, cord and membranes were most frequent in Mali (43%). Bangladesh and South Africa carried the highest share of maternal medical and surgical conditions.
The file is thinner than the diagnosis
Three gaps run through the paper, and all three are gaps in observation rather than in knowledge. In 55% of cases there was no documentation of a fetal heartbeat during labour — which left the panels unable to separate a death that happened before labour began from one that happened during it, and therefore unable to say which prevention opportunity had been missed. The timing of fetal death was unknown in 24% of cases. Gestational age was unknown in 21%.
Then the specimen. A placenta reached the panel in 49% of cases, ranging from 28% in Ethiopia to 75% in Bangladesh. The reason is mundane and systemic: placentas are not routinely preserved in these facilities and were often discarded before consent for the study could be obtained. The consequence is measurable. Where a placenta was examined, 37% of deaths were attributed to placental, membrane and cord complications; where none was available, that figure fell to 21%, with the cause resting on maternal clinical records and the family interview alone.
What an interview can carry, and what it cannot
Verbal autopsy was not an afterthought here — trained teams interviewed caregivers for every enrolled death, and panels read those findings. But the paper is explicit about the instrument’s reach in this domain: for stillbirths, verbal autopsy consists of only a few questions about the events a family reports around the delivery, and neither vital registration nor an interview involves reviewing a mother’s records or examining the fetus and the placenta. Those are the two processes the panels needed most.
The interview reconstructs a sequence a family witnessed. It cannot report histology, and it cannot report a heartbeat that no one auscultated. Where the biological record was missing, the earlier work on hypertensive and placental deaths in Mozambique made the same point from a different angle: without the specimen, cause attribution drifts toward whatever the available evidence supports.
Who reaches the table
Consent shaped the series as much as pathology did. Of 4,876 eligible stillbirths, 56% of families agreed to tissue sampling, from 32% in Bangladesh to 82% in Kenya. Sampling had to happen within 24 hours of delivery, or 36 if the body was refrigerated, which means deaths in the community were structurally less likely to be enrolled.
The authors are candid about the resulting bias: 95% of their stillbirths were delivered in facilities, close to the 96% of all births delivered in facilities in South Africa and 88% in Kenya, but far above national figures such as Ethiopia’s 48% and Bangladesh’s 65%. The paper therefore describes facility stillbirths in high-burden settings with unusual precision, not the national burden. Cause was completed for 91% of enrolled cases; 239 were still pending when the analysis closed.
Prevention, priced by site
Panels judged 72% of these deaths preventable given the care that already existed locally — a counterfactual judgement, not a measured effect. Among those, better antenatal and obstetric care was the recommended route in 85% of cases. Where panels wrote detailed notes, on 1,099 deaths, 69% could have been prevented through improved antenatal care (quality of care, management of medical conditions, health education), 33% through better obstetric management, and 10% through delivery by a qualified provider, concentrated in Sierra Leone and Kenya.
One recommendation is a data recommendation: in Kenya, the panels judged that 26% of preventable stillbirths could have been avoided through better documentation of medical records. The geography of the rest resists a single global message. Ethiopia’s share points at folate fortification for neural tube defects; Kenya’s at HIV prevention and control; South Africa’s at group B streptococcus vaccination and repeat syphilis screening, given pathogens with no current maternal screening pathway — 40% of congenital-infection deaths were classed as not preventable for exactly that reason. Mozambique’s points at referral, emergency transport and intrapartum monitoring.
What the study does not settle
The dominant category is also the vaguest: intrauterine hypoxia is a pathway, not an aetiology, and how much of it a panel assigns depends on how much placental and intrapartum evidence it holds — which varied by site. Comparability was managed by an annual cross-network review of five or six cases per site, a quality-assurance ritual rather than an inter-rater statistic. Preventing 72% of deaths is a professional judgement about optimal use of existing care; nothing in the design tests whether that care would in fact have been delivered.
What the paper does establish is a boundary. For the deaths that fill a health system’s maternity ward, the answers sit in the placenta, in the labour record and in the mother’s chart — objects that a surveillance system built on interviews can count but cannot interrogate. Half the placentas in this study were gone before anyone thought to look at them.
Sources
- Rahman A, Lee KH, El Arifeen S, et al. Causes of stillbirth in sub-Saharan Africa and South Asia: Findings from Child Health and Mortality Prevention Surveillance, 2016–2023. PLOS Global Public Health 2026;6(10):e0005235 (published 27 October 2026) journals.plos.org
- Full text (open access) — PMC13712675, PubMed Central pmc.ncbi.nlm.nih.gov
- WHO verbal autopsy instrument (2022 revision), World Health Organization who.int