Causes of Death Among Stillbirths and Children Aged <5 Years in Africa and South Asia — Child Health and Mortality Prevention Surveillance, Seven Countries, 2016–2024

Introduction

In 2023, an estimated 4.9 million deaths among children aged <5 years and 1.9 million stillbirths occurred worldwide; approximately 80% were in sub-Saharan Africa and South Asia, where deaths are likely underestimated and accurate cause-of-death determination remains challenging (1). Despite progress in reducing deaths, limited access to high-quality obstetric and pediatric care, health care workforce shortages, and inadequate diagnostic capacity contribute to high rates of preventable fetal, neonatal, infant, and child deaths in these regions (2,3). Accurate identification of underlying causes and contributing conditions is essential for understanding causal pathways and guiding prevention strategies. In many low- and middle-income countries (LMICs), death registration and medical certification remain incomplete, and many deaths occur outside health care facilities or shortly after arrival, with little opportunity for diagnostic evaluation (4). Conventional methods for cause-of-death attribution in LMICs, including verbal autopsy, rely largely on interviews with family members or other lay respondents and are designed primarily to assign a single probable underlying cause on the basis of limited information (5,6). These approaches generally do not identify multiple coexisting or contributory conditions. As a result, routine mortality data often record only one underlying cause and, without diagnostic confirmation, might be nonspecific or imprecise and obscure contributory factors relevant to prevention.

The Child Health and Mortality Prevention Surveillance (CHAMPS) network was established in 2016 to generate accurate cause-of-death data and guide prevention strategies in high-mortality settings. CHAMPS investigates deaths among neonates, infants, and children aged <5 years and stillbirths using standardized postmortem methods, including laboratory evaluation and minimally invasive tissue sampling (MITS) (7–9). After a death is identified, CHAMPS teams collect tissue and body fluid specimens for microbiologic, molecular, and histopathologic analyses, enabling identification of infectious, nutritional, perinatal, and other causes of death while maintaining cultural acceptability (10–12). These findings are integrated with verbal autopsy, clinical records, and contextual information on social determinants and health care system factors. At each site, multidisciplinary determination of cause of death (DeCoDe) panels review all available data using standardized World Health Organization (WHO) International Classification of Diseases (ICD) frameworks to assign underlying, antecedent, and immediate causes and to identify contributing maternal conditions, comorbidities, and health care system factors across the pregnancy-to-childhood continuum (13,14).

During 2016–2024, CHAMPS operated in defined catchment areas in Bangladesh and six African countries: Ethiopia, Kenya, Mali, Mozambique, Sierra Leone, and South Africa. CHAMPS was implemented through partnerships with national research and public health institutions and coordinated through the CHAMPS Program Office in collaboration with CDC and the Task Force for Global Health’s Public Health Informatics Institute. CDC provided scientific leadership, core mortality surveillance methods, laboratory approaches, and ongoing pathology and diagnostic support (15,16). CHAMPS findings were routinely shared with ministries of health and other stakeholders to support prevention activities and contribute to initiatives related to antimicrobial resistance, vaccine-preventable diseases, and maternal and child health surveillance (13,17,18).

This report summarizes surveillance findings during December 2016–December 2024 across CHAMPS catchment areas in seven countries and describes CHAMPS methods, laboratory systems, and findings on causes and preventability of stillbirths and deaths among neonates, infants, and children aged <5 years. Building on approximately 100 CHAMPS publications, including reports focused on specific countries, age groups, and conditions (19–22), this report highlights cross-site patterns, shared causal pathways, and health care system gaps that can guide improvements in mortality surveillance and strategies to reduce preventable deaths.

Methods

Surveillance System Overview

CHAMPS sites were selected through a multistage process on the basis of mortality rates, geographic representation, existing research infrastructure, and government commitment (7). During 2016–2024, surveillance sites operated in Bangladesh (Baliakandi and Faridpur), Ethiopia (Haramaya, Harar, and Kersa), Kenya (Karemo and Manyatta), Mali (Banconi, Djicoroni, and Sebenicoro), Mozambique (Manhiça and Quelimane), Sierra Leone (Bo and Makeni), and South Africa (Freedom Park, Soweto, and Thembelihle) (Figure 1). Site leadership teams were based at national research and public health institutions, including ICDDR,B (Dhaka, Bangladesh); College of Health and Medical Sciences, Haramaya University (Harar, Ethiopia); Kenya Medical Research Institute-Center for Global Health Research (Kisumu, Kenya); Centre pour le Développement des Vaccins-Mali (Bamako, Mali); Centro de Investigação em Saúde de Manhiça (Maputo, Mozambique); The Africa Research Collaborative for Health (Freetown, Sierra Leone); and Vaccines and Infectious Diseases Analytics Research Unit of the South African Medical Research Council, University of the Witwatersrand (Johannesburg, South Africa). Pakistan (Al Akbar Shah, Bhains Colony, and Karachi) and Nigeria (Bauchi and Calabar) joined during 2023–2024, but their data are excluded because those sites were in the early start-up phase of implementation.

Surveillance was implemented in collaboration with ministries of health within population-defined catchment areas with approximately 70,000–2,000,000 residents. Deaths were identified through health care facilities, community-based death-reporting networks, demographic surveillance systems, mortuaries, and civil registration offices (5,23). Most CHAMPS sites have public health and demographic surveillance systems that provide vital denominators and mortality notification frameworks (23). Ongoing community engagement supported culturally appropriate implementation, disclosure of cause-of-death findings to families, and dissemination of results to community leaders and public health officials. Sites also routinely shared data with health authorities to guide public health action (5,23,24). Network coordination, protocol standardization, and data management were conducted through the CHAMPS Program Office at Emory University (2016–2025) in partnership with CDC and the Task Force for Global Health’s Public Health Informatics Institute; in July 2025, the CHAMPS Program Office transitioned to the Task Force for Global Health. Surveillance protocols, laboratory procedures, data systems, and analytic workflows were standardized across sites (14). Detailed methods for CHAMPS field procedures, laboratory platforms (including custom syndromic TaqMan array cards), pathology protocols, and quality assurance have been published elsewhere (5,11–16,23,24). Additional information is available (CHAMPS resources).

Community Engagement and Feasibility

Before implementation, social and behavioral sciences (SBS) teams conducted formative research to understand local cultural, religious, and social norms related to death and burial and to develop community engagement and consent procedures (5,12,24). Activities included participatory meetings with community, traditional, and religious leaders; information sessions for caregivers; and monitoring of community concerns, misinformation, and rumors about MITS so they could be identified and addressed promptly. During surveillance, SBS staff members supported bereavement-sensitive consent, coordinated with local leaders to respect funeral timing and practices, and maintained community feedback mechanisms to identify barriers and adapt messages and procedures. These activities began before introduction of MITS and continued throughout implementation to support acceptability and guide operational adjustments. After DeCoDe review, families were offered an in-person results disclosure visit during which a clinician, often accompanied by SBS staff members, explained findings in plain language, provided grief counseling and psychosocial support, answered questions, and facilitated referral for clinical or mental health services when indicated. At many sites, families were offered more than one follow-up visit, and disclosure visits also provided opportunities to address identified preventable factors and connect families with relevant health care services.

Case Eligibility and Enrollment

CHAMPS enrolled stillbirths whose mothers met the catchment area–residency requirement and deaths among neonates, infants, and children aged <5 years who had lived in the catchment area for at least 4 months before death or since birth if aged <4 months (5). Eligibility for MITS required timely notification so that postmortem specimen collection could occur within 24 hours of death, or within 72 hours if the body had been refrigerated (5). Stillbirths were defined as fetal deaths at ≥28 weeks’ gestation or a birth weight of ≥1,000 g and no signs of life at delivery (25). Neonatal deaths were defined as deaths among live-born infants aged <28 days (9), and infant and child deaths were defined as deaths among children aged 28 days to <60 months (8).

Deaths were identified through notifications from health care facilities, including maternity wards, pediatric and emergency units, and mortuaries, and through mobile-platform reports submitted by designated community death reporters (12). Deaths were classified as facility deaths if they occurred in health care facilities, including referral hospitals, community hospitals, and clinics; deaths occurring outside health care facilities were classified as community deaths. Site teams contacted families to obtain consent for MITS and related data collection, including clinical record abstraction and verbal autopsy. When MITS was declined or was not feasible, such as after delayed notification or because the body had not been refrigerated, families were invited to participate in a non-MITS pathway that included verbal autopsy and, when available, clinical record abstraction. Community-based mobile teams were established to support timely local response to deaths that occurred in homes. Enrollment stopped at all sites during the early months of the COVID-19 pandemic and restarted at different times as communities lifted restrictions.

Minimally Invasive Tissue Sampling

Before specimen collection, postmortem anthropometric measurements, including weight, length, and mid-upper arm circumference (MUAC), were obtained using calibrated scales, length boards, and MUAC tapes. z scores for weight-for-age (WAZ), length-for-age (LAZ), weight-for-length (WLZ), and MUAC-for-age (MUACZ) were calculated according to WHO Child Growth Standards to support standardized cross-site comparisons (26,27). z scores were categorized as normal (≥ −2 SD), moderate (−3 SD to < −2 SD), and severe (< −3 SD). Indicators of undernutrition were defined as underweight (WAZ < −2 SD), stunting (LAZ < −2 SD), and wasting (WLZ < −2 SD or MUACZ < −2 SD).

MITS protocols were age-specific and standardized across sites (11). For all deaths, specimens included whole blood, cerebrospinal fluid (CSF), nasopharyngeal and rectal swabs, and tissue biopsies from the lung, liver, and brain. For stillbirths and neonatal deaths, placental disc, membranes, and umbilical cord specimens also were collected to assess maternal-fetal and perinatal conditions (5). Specimen collection was performed by trained personnel, including pathologists, clinicians, and dedicated nonphysician MITS technicians, using sterile technique and 14- to 16-gauge biopsy needles (11,16).

Each site maintained a biosafety level 2 mortuary laboratory for specimen processing (16). Whole blood and CSF were collected in ethylenediaminetetraacetic acid (EDTA) tubes; specimens for culture and molecular testing were stored at 2°C–8°C (32.6°F–46.4°F) and transported within 4 hours to the appropriate laboratory for analysis. Formalin-fixed specimens were used for histopathologic evaluation. Standardized procedures were supported through international training, quality-assurance review, and telepathology mentorship from CDC pathology experts (16).

Laboratory Diagnostics

CHAMPS laboratories conducted standardized postmortem diagnostic testing to identify infectious and noninfectious causes of death (15,28,29). Blood and CSF specimens underwent bacterial culture testing; in South Africa, lung tissue also was cultured for bacteria and fungi, and blood and CSF were cultured for fungi. Molecular testing was performed on lung tissue, blood, CSF, nasopharyngeal and oropharyngeal swabs, and rectal swabs using CDC-developed multiplex syndromic TaqMan array cards that detected approximately 100 bacterial, viral, fungal, and parasitic targets (15). Because certain TaqMan array card targets are shared by Escherichia coli and Shigella spp./enteroinvasive E. coli, these detections are reported as “E. coli/Shigella spp.”; detailed methods for pathotype differentiation have been described (30). Mycobacterium tuberculosis was detected by GeneXpert MTB/RIF Ultra, and malaria infection was confirmed by microscopy, rapid diagnostic testing, or molecular assays (13,15,31). HIV infection was assessed by qualitative polymerase chain reaction (PCR) for viral RNA or DNA using whole blood or dried blood spot specimens (32).

Histopathologic evaluation was performed at site laboratories and CDC reference laboratories (16). Hematoxylin and eosin staining was used to characterize tissue injury and inflammation. When histopathologic findings or clinical context indicated the need for additional evaluation, ancillary testing was performed on a case-by-case basis and guided by CHAMPS syndromic algorithms and molecular and microbiologic results. Ancillary testing included special stains, immunohistochemistry, formalin-fixed paraffin-embedded tissue–based 16S ribosomal RNA PCR with sequencing, targeted PCR or reverse–transcription PCR, and other site-specific assays. Telepathology enabled remote case review using whole-slide imaging and the CHAMPS telepathology platform (slides scanned at sites using NanoZoomer-SQ digital slide scanners and at CDC using a Leica Aperio AT2, then uploaded to the CHAMPS central server for shared viewing and annotation). CDC-coordinated quality assurance, including standardized histopathology checklists, diagnostic algorithms, and telepathology-based case review, supported consistency in pathology interpretation across site and reference laboratories (16).

Verbal Autopsy

When consent was provided by a parent, caregiver, or other close family member, trained interviewers administered a locally translated WHO 2016 verbal autopsy questionnaire for deaths occurring in health care facilities or the community (5,33). The questionnaire was adapted for CHAMPS to include site identifiers and updated content, skip logic, and corrected units of measurement. Interviews were conducted after burial and at the family’s convenience, typically 2–4 weeks after death and usually within 12 weeks, to balance mourning practices with recall accuracy.

Clinical and Maternal Record Abstraction

For deaths occurring in health care facilities, clinical and maternal records were abstracted using standardized abstraction forms to collect demographic characteristics, admission diagnoses, vital signs, anthropometric measurements, laboratory findings, treatments, and clinical course (5,13). Neonatal abstraction forms included labor and delivery circumstances, Apgar scores, resuscitation, early feeding, and neonatal complications. Stillbirth abstraction forms included maternal age, parity, antenatal care visits, ultrasound findings, hypertensive disorders, maternal infections, and intrapartum events.

Determination of Cause of Death

Multidisciplinary DeCoDe panels, including clinicians, pediatricians, obstetricians, pathologists, microbiologists, and epidemiologists, reviewed laboratory, pathology, clinical, maternal, and verbal autopsy data compiled in standardized electronic dossiers. Panels assigned causes of death using WHO ICD, Tenth Revision (ICD-10) and ICD for Perinatal Mortality (ICD-PM) codes, which classify perinatal deaths by timing and links them to five maternal condition groups: M1, complications of the placenta, cord, or membranes; M2, maternal complications of pregnancy; M3, other complications of labor and delivery; M4, maternal medical or surgical conditions; and M5, no maternal condition identified. A phased transition to ICD, Eleventh Revision began in 2022. For this report, maternal conditions were summarized by using standard CHAMPS groupings that map DeCoDe-assigned ICD diagnoses to the ICD-PM maternal condition framework and predefined subcategories, consistent with previous CHAMPS perinatal analyses. Each DeCoDe record could include approximately 120 laboratory results and thousands of discrete clinical, maternal, pathology, and verbal autopsy data elements per case (14).

Panelists evaluated the sequence of events leading to death, identified one underlying cause and, when applicable, identified one or more antecedent causes and an immediate cause (8,13). The underlying cause was the condition that initiated the chain of events leading to death; antecedent causes were intermediate conditions in that chain, and the immediate cause was the final condition directly resulting in death. Panels integrated laboratory and pathology findings with clinical and contextual information to determine whether identified pathogens or other conditions were causal and to identify deaths in which multiple conditions contributed (17,30). When deciding whether to include a condition in the causal chain, panelists applied counterfactual reasoning: they considered whether the death would have occurred, when it occurred, in the absence of that condition (13). Network quality assurance included periodic cross-site review of a small subset of cases from each site; results were analyzed centrally to assess interpanel concordance and guide retraining and revisions to diagnosis standards (14). Any diagnosis classified as an underlying, antecedent, or immediate cause was counted as a cause; therefore, cause-specific proportions reflect multicausal attribution and can sum to >100%.

Preventability Assessment

For each case with a determined cause of death, DeCoDe panels completed a structured preventability assessment (3,14). Panelists classified a death as 1) preventable when they assessed that one or more plausible changes in health care system performance, access to care, or family response could have averted the death; 2) possibly preventable under certain circumstances when they assessed that the death might have been averted only under specified circumstances; or 3) not preventable when no such change was considered likely to avert the death. Classifications were assigned within the context of services, resources, and policies available in the facility and community and focused on circumstances immediately surrounding the death rather than broader political, financial, or social determinants. As part of the preventability assessment, panelists could identify one or more standardized health care system improvement categories and subcategories, including antenatal care, obstetric care, pediatric clinical management and quality of care, health care–seeking behavior, health education, infection prevention and control, nutritional support, HIV infection prevention and treatment, vaccination, and transportation. These categories were derived from early DeCoDe free-text recommendations and subsequently applied across sites. A subset of DeCoDe-reviewed cases underwent secondary cross-site review as a quality-control measure; however, the interrater reliability specifically for preventability classifications was not calculated. Individual preventability recommendations were shared with participating families when relevant for the health of the parents (e.g., pregnancy-related conditions that might reoccur) or other children (e.g., malnutrition or vaccine-preventable conditions); aggregate preventability findings and recommendations were shared with health care facilities, communities, ministries of health, and other public health stakeholders to guide locally appropriate prevention activities.

Data Management and Quality Assurance

Data were collected by using standardized electronic systems, including Research Electronic Data Capture (REDCap [version 16.1.4; Vanderbilt University]) and LabKey, with embedded range and logic checks, automated audit trails, and secure data transfer protocols (11,14). The CHAMPS Program Office conducted routine audits and cross-site data harmonization. Deidentified data were stored in a centralized repository under standardized quality-control procedures. CHAMPS maintained a centralized, web-accessible global data repository governed by a data governance committee; access was role based, and deidentified datasets could be made available after registration and approval, with additional protections for datasets that included potentially identifying variables.

Ethics

Activities were approved by institutional review boards or national ethics committees in each participating country and by Emory University. This activity was reviewed by CDC and conducted consistent with applicable federal law and CDC policy.* CHAMPS used tiered informed consent to balance scientific value with family autonomy and local practices. Trained staff members proficient in the local language approached families soon after death notification to explain procedures and obtain written informed consent before enrollment and data collection. Community engagement and formative social and behavioral research guided the way that teams explained MITS, approached families during bereavement, coordinated with funeral and burial practices, and adapted site-specific consent procedures while preserving voluntary participation. Ongoing feedback from families, community leaders, and local teams was used to identify concerns and refine messaging, timing of family approach, and data-collection procedures. Materials supporting cause-of-death determination (e.g., clinical documents and pathology images) were handled under approved protocols (14). Activities were coordinated with community leaders to respect burial practices and support timely release of the body. Data were deidentified, stored securely, and shared under data-use agreements with ministries of health.

Results

Enrollment and MITS Completion

During December 2016–December 2024, a total of 18,784 eligible stillbirths and deaths among neonates, infants, and children aged <5 years were reported across seven CHAMPS countries (Bangladesh, Ethiopia, Kenya, Mali, Mozambique, Sierra Leone, and South Africa); 15,612 (83.1%) of these deaths were enrolled (Figure 2) (Supplementary Table 1). Consent for MITS was obtained for 9,518 deaths, and MITS analysis was completed for 9,415 (98.9%) of these deaths. A total of 8,500 (90.3%) deaths with completed MITS analyses also underwent complete DeCoDe review; these 8,500 deaths were included in the analysis.

Demographic, Clinical, and Anthropometric Characteristics

Among the 8,500 deaths included in the analysis, 3,199 (37.6%) were stillbirths, 3,230 (38.0%) were neonatal deaths, and 2,071 (24.4%) were infant or child deaths (Table 1). The proportion of stillbirths ranged from 57.7% (761 of 1,318) in Ethiopia to 25.6% (303 of 1,183) in Sierra Leone. Males accounted for 55.5% of deaths. Overall, 88.2% of deaths occurred in health care facilities and 11.8% in the community. Facility deaths accounted for most deaths in all sites, ranging from 74.3% in Kenya and 83.7% in Ethiopia to >90% in Bangladesh, Mozambique, and Sierra Leone. Median duration of stay in a health care facility before death (excluding stillbirths) was 31 hours (IQR = 9–94); country-specific medians ranged from 18 hours in Kenya to 81 hours in South Africa. The overall median interval from death to MITS completion was 9 hours (IQR = 3–17); by country, the median interval ranged from 1 hour in Bangladesh to 23 hours in South Africa.

Postmortem anthropometry of bodies indicated a high prevalence of undernutrition among infants and children, with country-level variation; these measures did not necessarily correspond to malnutrition being included in the causal chain. Among infants and children with available measurements, 61.3% (1,241 of 2,026) were moderately or severely underweight, 43.0% (864 of 2,008) were stunted, and 61.3% (1,218 of 1,987) were wasted (Table 1). Severe underweight was most frequent in Ethiopia (77.5%), severe stunting in Ethiopia (76.1%) and Mozambique (38.9%), and severe wasting in Ethiopia (66.9%), Mali (51.6%), and Kenya (45.6%).

Causes of Death

Multiple conditions were commonly identified in the causal chain leading to death, particularly among neonates (44.3% with two or more conditions) and infants and children (67.9% with two or more conditions). Among 3,199 stillbirths, the conditions most commonly identified anywhere in the causal chain were perinatal asphyxia or hypoxia (79.1%), followed by congenital birth defects (9.0%) (Table 2) (Supplementary Table 2). Infectious causes were less common; congenital infections (2.9%) and syphilis (0.8%) were concentrated in South Africa and Mozambique. Cause-of-death distributions varied by site and country: asphyxia or hypoxia accounted for >85% in Bangladesh, Kenya, and Sierra Leone, versus 74.1% in Ethiopia and 60.7% in South Africa (Supplementary Figure 1). The proportion of stillbirths with congenital birth defects was highest in Ethiopia (23.7%), primarily reflecting neural tube defects, including anencephaly or craniorachischisis, spina bifida, encephalocele, and congenital hydrocephalus.

Among 3,230 neonatal deaths, leading causes identified anywhere in the causal chain were preterm birth complications (39.7%), perinatal asphyxia or hypoxia (37.7%), and sepsis (36.5%); lower respiratory infections (12.0%), meningitis or encephalitis (7.7%), congenital birth defects (7.7%), and congenital infections (5.7%) were also common (Table 2). Preterm birth complications (34.7%) and asphyxia or hypoxia (32.4%) were the most common underlying causes, whereas sepsis (27.0%) and lower respiratory infections (9.4%) more often appeared as antecedent or immediate causes (Supplementary Table 2). The proportion of neonatal deaths with sepsis included anywhere in the causal chain varied across sites (ranging from 21.6% in Kenya to 62.9% in Ethiopia) (Supplementary Figure 1). Preterm birth complications were most frequent in South Africa (60.7%) and Bangladesh (53.2%); perinatal asphyxia or hypoxia in Sierra Leone (56.1%) and Ethiopia (53.0%); congenital birth defects in Mali (16.1%) and South Africa (11.7%); and aspiration syndromes in Ethiopia (15.4%). Across all age groups, when congenital birth defects were present in the causal chain, they were typically assigned as the underlying cause (overall, 668 of 704 [94.9%]).

Among 2,071 infant and child deaths, leading conditions identified anywhere in the causal chain of death included lower respiratory infection (37.4%), sepsis (36.9%), malnutrition (27.3%), malaria (22.1%), diarrheal disease (17.3%), anemia (11.3%), congenital birth defects (8.0%), HIV infection (7.7%), and meningitis or encephalitis (7.4%) (Table 2). The most frequent underlying causes were malnutrition (20.8%), malaria (15.5%), and lower respiratory infection (9.1%), whereas sepsis (32.4%), lower respiratory infection (28.8%), and diarrheal disease (9.9%) more often occurred as antecedent or immediate causes (Figure 3) (Supplementary Table 2). Malaria frequently contributed to deaths in Sierra Leone (40.1%), Kenya (31.5%), and Mozambique (20.1%). In Ethiopia, sepsis (72.5%), malnutrition (76.1%), and diarrheal disease (45.8%) each contributed to a substantial proportion of deaths (Supplementary Figure 1).

Pathogens Identified in the Causal Chain of Death

Among 3,199 stillbirths, a total of 230 (7.2%) had at least one pathogen included in the causal chain of death, most often a single organism (182 of 230 [79.1%]; mean = 1.3 pathogens per death). Gram-negative bacteria (131 of 230 [57.0%]) and gram-positive bacteria (108 of 230 [47.0%]) were the pathogen groups most often included in the causal chain; percentages exceed 100% because certain deaths involved multiple pathogens. The pathogens most often included in the causal chain were E. coli/Shigella spp. (65 of 230 [28.3%]), Streptococcus agalactiae (54 of 230 [23.5%]), Treponema pallidum (29 of 230 [12.6%]), and Klebsiella pneumoniae (22 of 230 [9.6%]) (Table 3) (Supplementary Table 3) (Supplementary Figure 2). Among stillbirths with a pathogen included in the causal chain, the proportion with S. agalactiae infection varied by site, ranging from 18.2% in Mozambique (four of 22) and 18.4% in Ethiopia (nine of 49) to 33.0% in South Africa (34 of 103). Viral causes were uncommon; cytomegalovirus (24 of 230 [10.4%]) was the virus most often included in the causal chain, and no fungal pathogens were included.

Among 3,230 neonatal deaths, at least one pathogen was included in the causal chain of death for 1,234 (38.2%), with one pathogen in 786 (63.7%), two in 283 (22.9%), and three or more in 165 (13.4%) (mean = 1.5 per death). Gram-negative bacteria were the most frequently identified pathogen group in the causal chain (1,057 of 1,234 [85.7%]), particularly K. pneumoniae (596 of 1,234 [48.3%]), Acinetobacter baumannii (319 of 1,234 [25.9%]), and E. coli/Shigella spp. (179 of 1,234 [14.5%]) (Table 3). These bacteria were typically assigned as antecedent or immediate causes and were most often implicated in sepsis, lower respiratory infection, and meningitis and encephalitis (Supplementary Table 3) (Supplementary Figure 3). Gram-positive bacteria were included in the causal chain in 285 (23.1%) deaths, most often S. agalactiae (69 of 1,234 [5.6%]), Staphylococcus aureus (52 of 1,234 [4.2%]), and Streptococcus pneumoniae (37 of 1,234 [3.0%]) (Table 3). Among neonatal deaths with a pathogen included in the causal chain, the proportion with S. agalactiae varied by site, ranging from 5.1% in Mali (seven of 138) and 8.0% in Mozambique (11 of 137) to 11.4% in South Africa (41 of 360). Fungal pathogens were included in the causal chain in 68 (5.5%) deaths, predominantly Candida albicans (34 of 1,234 [2.8%]). Viral pathogens were included in the causal chain in 59 (4.8%) deaths, most often cytomegalovirus (21 of 1,234 [1.7%]). Across sites, K. pneumoniae was the pathogen most frequently included in the causal chain, whereas A. baumannii was disproportionately frequent in sites in South Africa (207 of 360 [57.5%]) and Bangladesh (67 of 165 [40.6%]) (Supplementary Figure 4).

Among 2,071 infant and child deaths, at least one pathogen was included in the causal chain for 1,515 (73.2%). Among these 1,515 deaths, the mean number of pathogens included in the causal chain was 2.0 per death (range = 1–7): one pathogen was identified in 739 (48.8%) deaths, two in 362 (23.9%), and three or more in 414 (27.3%). Most infant and child deaths with a pathogen included in the causal chain occurred in health care facilities (1,075 of 1,515 [71.0%]). Gram-negative organisms were included in the causal chain in 831 (54.9%) deaths, most often K. pneumoniae (474 of 1,515 [31.3%]), E. coli/Shigella spp. (237 of 1,515 [15.6%]), and Haemophilus influenzae (detected by a pangenome H. influenzae detection assay; H. influenzae type b [Hib] target not detected) (128 of 1,515 [8.4%]). Gram-positive bacteria were included in the causal chain in 522 (34.5%) deaths, most often S. pneumoniae (305 of 1,515 [20.1%]) and S. aureus (95 of 1,515 [6.3%]) (Table 3) (Supplementary Figures 5 and 6). Viral pathogens were included in the causal chain in 480 (31.7%) deaths, most commonly cytomegalovirus (129 of 1,515 [8.5%]) and adenovirus (89 of 1,515 [5.9%]); parasitic pathogens were included in 420 (27.7%) deaths, primarily Plasmodium falciparum (408 of 1,515 [26.9%]); and fungal pathogens were included in 108 (7.1%) deaths, including Pneumocystis jirovecii (52 of 1,515 [3.4%]) and C. albicans (28 of 1,515 [1.8%]). Among 408 deaths with P. falciparum in the causal chain, 92 (22.5%) also had at least one bacterial pathogen in the causal chain. Among infant and child deaths with a pathogen included in the causal chain, the proportion with S. pneumoniae was highest in Ethiopia (60 of 125 [48.0%]), Mali (50 of 134 [37.3%]), and Mozambique (62 of 195 [31.8%]).

Maternal Conditions Associated with Stillbirths and Neonatal Deaths

A main maternal condition was assigned for 2,273 of 3,199 stillbirths (71.1%) and 1,879 of 3,230 neonatal deaths (58.2%) (Table 4) (Supplementary Tables 4 and 5). Among stillbirths, the most frequent ICD-PM maternal condition groups were M1, complications of the placenta, cord, or membranes (891 of 3,199 [27.9%]); M4, maternal medical or surgical conditions (748 of 3,199 [23.4%]); M3, other complications of labor and delivery (345 of 3,199 [10.8%]); and M2, maternal complications of pregnancy (168 of 3,199 [5.3%]). Among neonatal deaths, the most frequent maternal condition groups were M4 (484 of 3,230 [15.0%]), M3 (468 of 3,230 [14.5%]), M2 (455 of 3,230 [14.1%]), and M1 (309 of 3,230 [9.6%]).

Among stillbirths, the most frequent main maternal conditions were placental complications (539 of 3,199 [16.8%]), primarily placental abruption or other placental separation with hemorrhage, and hypertensive disorders of pregnancy (537 of 3,199 [16.8%]). Hypertensive disorders were most frequent in South Africa (102 of 374 [27.3%]), Kenya (72 of 381 [18.9%]), and Bangladesh (98 of 543 [18.0%]) (Supplementary Table 4). Placental complications were most frequent in Mali (92 of 276 [33.3%]), South Africa (80 of 374 [21.4%]), and Sierra Leone (61 of 303 [20.1%]). Chorioamnionitis and membrane complications accounted for 6.9% of stillbirths (220 of 3,199) and umbilical cord complications for 3.5% (112 of 3,199).

Among neonatal deaths, the most frequent main maternal conditions were hypertensive disorders of pregnancy (294 of 3,230 [9.1%]), often linked to asphyxia or hypoxia, followed by multiple gestation (227 of 3,230 [7.0%]), premature rupture of membranes (171 of 3,230 [5.3%]), and preterm labor or delivery (117 of 3,230 [3.6%]) (Supplementary Table 5). The proportion of neonatal deaths with placenta, cord, or membrane complications (group M1) as the assigned maternal condition was highest in South Africa (89 of 616 [14.4%]), Sierra Leone (39 of 374 [10.4%]), Mozambique (67 of 655 [10.2%]), and Mali (31 of 305 [10.2%]) (Supplementary Table 5).

Across stillbirths and neonatal deaths, perinatal asphyxia or hypoxia was the most common underlying cause among deaths with placenta, cord, or membrane complications (group M1; 853 of 1,200 [71.1%]) and other labor and delivery complications (group M3; 602 of 813 [74.0%]) (Table 4). Preterm birth complications were the leading underlying cause among deaths with maternal complications of pregnancy (group M2; 264 of 623 [42.4%]). Among deaths with no main maternal condition identified (group M5), congenital birth defects accounted for 19.2% (437 of 2,277).

Preventability of Deaths

Among 8,500 deaths analyzed, 7,558 (88.9%) had an overall preventability classification; among these, 6,103 (80.7%) were classified as preventable (n = 5,323) or as possibly preventable under certain circumstances (n = 780). Specific health care system improvement recommendations were documented for 6,276 deaths and are summarized overall (Table 5) and by site and age group (Supplementary Figure 7). The most frequently identified opportunities for improvement were in antenatal care (2,684 of 6,276 [42.8%]), obstetric care and management (2,460 of 6,276 [39.2%]), health care–seeking behavior (2,285 of 6,276 [36.4%]), and pediatric clinical management and quality of care (2,094 of 6,276 [33.4%]). Additional gaps included health education (1,563 of 6,276 [24.9%]), infection prevention and control (955 of 6,276 [15.2%]), nutritional support (653 of 6,276 [10.4%]), preconception counseling and family planning (566 of 6,276 [9.0%]), HIV infection prevention and control (225 of 6,276 [3.6%]), and transport systems (225 of 6,276 [3.6%]). These recommendations often reflected gaps in diagnostic capacity, timely referral, and access to recommended treatment, particularly for severe infections.

Prevention opportunities varied by age group. Among stillbirths with recommendations (n = 2,244), recommendations most often focused on antenatal care (1,542 of 2,244 [68.7%]) and obstetric care and management (1,246 of 2,244 [55.5%]) (Table 5), including clinical management of pregnancy complications, screening and diagnosis, and fetal monitoring during labor and delivery. Among neonatal deaths with recommendations (n = 2,431), the most frequently identified priorities were obstetric care and management (1,185 of 2,431 [48.7%]), antenatal care (1,087 of 2,431 [44.7%]), and pediatric clinical management and quality of care (1,086 of 2,431 [44.7%]), reflecting susceptibilities spanning late pregnancy, delivery, and the immediate postnatal period. Infection prevention and control was identified for 24.8% (603 of 2,431) of neonatal deaths, particularly at sites where K. pneumoniae and A. baumannii were major causes of sepsis. Among infant and child deaths with recommendations (n = 1,601), recommendations most often addressed pediatric clinical management and quality of care (920 of 1,601 [57.5%]), health care–seeking behavior (773 of 1,601 [48.3%]), and health education (686 of 1,601 [42.8%]), as well as gaps in nutritional support (422 of 1,601 [26.4%]) and infection prevention and control (318 of 1,601 [19.9%]).

Discussion

This analysis of CHAMPS mortality surveillance data from Bangladesh and six African countries during 2016–2024 provides standardized, multisource postmortem evidence on causes and preventability of stillbirths and deaths among neonates, infants, and children aged <5 years in high-mortality settings. Among 8,500 deaths investigated by MITS and reviewed by multidisciplinary DeCoDe panels, infections, intrapartum complications, preterm birth complications, and malnutrition often occurred together within multistep causal chains that also involved maternal conditions, comorbidities, and gaps in care. By integrating histopathology, microbiology, clinical and maternal record review, verbal autopsy, and expert interpretation, CHAMPS complements verbal autopsy, facility records, and routine mortality reporting by identifying not only broad causes of death but also specific pathogens, maternal contributors, and modifiable factors relevant to prevention.

Overall Mortality Findings

Cause-of-death patterns varied by age group, reflecting transitions in risk from pregnancy and delivery through the neonatal period and early childhood. Most stillbirths were attributed to perinatal asphyxia or hypoxia, often in the context of maternal or placental conditions such as hypertensive disorders of pregnancy, placental complications, and chorioamnionitis. These findings highlight opportunities for improved antenatal detection, intrapartum monitoring, timely referral, and emergency obstetric management (34,35). Neonatal deaths were primarily attributed to preterm birth complications, perinatal asphyxia or hypoxia, and sepsis. Gram-negative bacteria, especially K. pneumoniae and A. baumannii, were frequently included in the causal chain of neonatal deaths. Previous analyses of CHAMPS data suggested that many infections involving these pathogens were likely associated with health care, although community acquisition also occurred (especially for K. pneumoniae) and the acquisition setting could not be established for every death (9,17,36). These findings reinforce the importance of infection prevention and control, timely recognition and management of neonatal infection, and quality improvement in maternity and newborn health care units (36,37).

Among infants and children, deaths commonly involved both infectious and nutritional causes, including lower respiratory infections, sepsis, diarrheal disease, and malnutrition. Malaria and HIV infection contributed substantially to the number of deaths at certain sites, indicating persistent gaps in access to, uptake of, or quality of established prevention and treatment services (31,32). Malnutrition affected approximately half of children with anthropometric measurements and was frequently included in the causal chain, underscoring the close relationship between undernutrition and infectious mortality (26,38,39). These findings support integrated clinical and public health approaches that can identify and manage overlapping risks, including infection, prematurity, growth failure, HIV exposure or infection, and barriers to timely care, rather than single-disease approaches alone (3,13,40).

The frequent identification of multiple conditions in the causal chain is a central finding. Conventional mortality data often assign a single underlying cause, which can obscure preventable contributors such as maternal complications, malnutrition, delayed care-seeking, missed antenatal screening, inadequate intrapartum monitoring, or health care–associated infection risks. CHAMPS provides a more complete picture of causal pathways by distinguishing underlying, antecedent, and immediate causes of death and by evaluating whether pathogens or other findings plausibly contributed to death. This approach is particularly important in high-mortality settings, where children often seek care late with severe illness and multiple coexisting conditions.

Interpretation in the Context of Global Evidence

Cause-of-death patterns across CHAMPS sites broadly align with global estimates from the United Nations Inter-agency Group for Child Mortality Estimation and WHO, while providing greater etiologic specificity and additional evidence on comorbidities, maternal contributors, and health care system gaps (1). The predominance of intrapartum-related stillbirths and early neonatal deaths and the substantial number of deaths attributed to infectious diseases among neonates, infants, and children are consistent with previous studies (41,42). CHAMPS links these broad categories to specific pathogens and modifiable circumstances. For example, the prominence of K. pneumoniae across sites highlights the importance of infection prevention and control, diagnostic capacity, prompt empiric management of suspected severe bacterial infection, and development of preventive tools, including vaccines against major neonatal and pediatric bacterial pathogens (17,36).

CHAMPS findings also underscore the continuing contribution of maternal hypertensive disorders and infections such as HIV and syphilis to fetal and neonatal deaths (9,25). Stillbirths and neonatal deaths comprised approximately three fourths of MITS-investigated deaths, emphasizing the concentration of deaths during pregnancy, labor, delivery, and the early newborn period. As deaths among older children decline, preventing fetal and neonatal deaths will require sustained attention to timely and high-quality antenatal, obstetric, and newborn care. These findings are consistent with WHO and UNICEF priorities, including the Every Newborn Action Plan and the Global Strategy for Women’s, Children’s, and Adolescents’ Health, and are most useful when translated into district- and facility-level action through routine dissemination to health authorities, linkage with child and perinatal death review processes, and audit and feedback to address modifiable gaps in care and referral (43–45).

CHAMPS also demonstrates the value of postmortem surveillance as a complement to routine mortality systems. By operating within defined catchment areas and, in certain sites, linking with public health and demographic surveillance systems, CHAMPS can support interpretation of mortality patterns and comparison with model-based estimates (23,46). The system also generates evidence that routine data sources often cannot provide, including histopathologic evidence of disease processes, pathogen-specific attribution, maternal contributors to stillbirth and neonatal death, and structured assessments of preventability.

Preventability and Public Health Implications

Approximately four in five deaths with a preventability assessment were considered preventable or possibly preventable under certain circumstances. This finding does not imply that every death could have been prevented by a single intervention; rather, it indicates that many deaths occurred along pathways in which earlier recognition, timely referral, improved quality of care, or better implementation of existing interventions might plausibly have altered the outcome. The most common opportunities identified by DeCoDe panels included strengthened antenatal care, improved intrapartum and emergency obstetric management, infection prevention and control, improved pediatric case management, nutritional support, and earlier health care–seeking for pregnancy complications and childhood illness.

These priorities are consistent with evidence that expanding access to and improving delivery of established maternal, newborn, and child health interventions can prevent many stillbirths and neonatal and child deaths. Relevant interventions include high-quality antenatal and intrapartum care; screening for and management of hypertensive disorders, diabetes, HIV infection, and syphilis; antenatal corticosteroids when indicated for threatened preterm birth; timely obstetric referral; neonatal resuscitation and supportive care; prevention and treatment of severe bacterial infection; oxygen and other supportive therapies for severe pneumonia; nutritional support; malaria prevention and treatment; prevention of mother-to-child transmission of HIV; and equitable receipt of routine childhood vaccines (36,47–53). CHAMPS also identified neural tube defects as an important contributor to stillbirth and neonatal mortality in Ethiopia, supporting the potential value of periconceptional folic acid intake and food fortification where feasible (54).

The high proportion of preventable deaths suggests that the major gaps often are not the absence of known interventions but incomplete implementation, uneven quality, delayed access, shortages of staff or commodities, weak referral systems, and limited accountability (55). Social and behavioral findings from CHAMPS have also highlighted barriers to timely care, including distance, cost, mistrust, and reliance on traditional medicine, emphasizing that clinical improvements must be paired with strategies that reduce barriers to access and strengthen community trust (24,56).

Across CHAMPS sites, findings have guided preventive and quality-improvement activities. Beginning in 2022, CHAMPS findings informed a package of interventions in Siaya County, Kenya, including expansion of Kangaroo Mother Care and training in emergency obstetric care and neonatal triage, assessment, and treatment. From 2022 to 2023, Kangaroo Mother Care admissions nearly doubled, while crude mortality among combined newborn-unit and Kangaroo Mother Care admissions declined concurrently with these interventions (57). In South Africa, CHAMPS helped expand access to antenatal ultrasound services; ultrasound scans were performed for 696 pregnancies during June 2023–April 2024 (58). In addition, aspirin as a preventive treatment for pregnancy-related hypertensive disorders was added to the South African national formulary in response to CHAMPS data showing the burden of stillbirths and neonatal deaths related to these disorders. In Sierra Leone, CHAMPS findings that demonstrated that malnutrition was associated with 48% of deaths among children aged <5 years supported the country’s first domestic procurement of ready-to-use therapeutic food. The Sierra Leone government purchased 1,700 cartons (for US $99,000), and UNICEF secured an additional 1,700 cartons through a matching funding mechanism; supplies were distributed across all 16 districts (59). Moreover, CHAMPS findings on HIV infection supported an additional US $8 million investment from the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) for a population-based HIV impact assessment and the reinitiation of national HIV viral-load testing in Sierra Leone. These examples and others document changes in service delivery and resource allocation. Integrating CHAMPS findings with national surveillance, mortality review, and routine health information systems can help guide program response and strengthen accountability for achieving the United Nations’ goal of ending preventable deaths among newborns and children aged <5 years by 2030.

Implementation of CHAMPS also strengthened public health capacity. Site teams conducted bereavement-sensitive postmortem investigations in partnership with communities, demonstrating that MITS-based mortality surveillance can be feasible and acceptable when supported by sustained community engagement, transparent consent processes, and family-centered disclosure of results (5,12,24). Network investments strengthened laboratory, pathology, data, and analytic systems, including microbiology and molecular testing, histopathology, quality assurance, and standardized cause-of-death review (15,16,43,60). These capacities extend beyond the analyses in this report by supporting surveillance for priority pathogens, outbreak investigations, and local quality-improvement activities.

Limitations

The findings in this report are subject to at least four limitations. First, CHAMPS catchment areas might not be nationally representative, and the analytic population was shaped by site selection, death notification, family consent, and MITS eligibility requirements. Because MITS required timely notification and specimen collection, deaths occurring in health care facilities were more likely to be investigated. Consequently, community deaths, deaths with delayed notification, deaths evaluated primarily through verbal autopsy, and certain injury-related deaths might be underrepresented (47,61).

Second, data completeness varied across sites and over time. Maternal and child clinical records, antenatal screening results, placental data, premortem diagnostics, and documentation of treatments were sometimes incomplete. These missing data could have limited assessment of upstream contributors to stillbirth, prematurity, fetal growth restriction, intrauterine hypoxia, congenital conditions, and chronic or noninfectious diseases. Certain maternal infections, chronic maternal conditions, child comorbidities, and structural abnormalities might therefore have been underdiagnosed or misclassified.

Third, pathogen attribution in postmortem surveillance is complex. CHAMPS used standardized laboratory, pathology, and DeCoDe procedures to distinguish pathogens considered causal from organisms detected incidentally or through contamination; however, attribution could still vary because of true epidemiologic differences, differences in case mix and enrollment, previous antimicrobial exposure, timing of specimen collection, and residual differences in interpretation across panels despite quality-assurance procedures (13,14). This broad surveillance summary presents pathogens included in the causal chain of death but does not provide detailed, case-level analyses of antimicrobial susceptibility, antibiotic access and use, treatment failure, or infection source. More detailed pathogen- and syndrome-specific CHAMPS analyses have addressed these problems for priority organisms and clinical syndromes, including K. pneumoniae, pneumonia, and neonatal bacterial infections (17,22,28,36). Therefore, aggregate pathogen findings in this report should be interpreted as an overall surveillance summary and considered in addition to organism-specific analyses of antimicrobial resistance, antibiotic treatment options, and community-associated versus health care–associated acquisition.

Finally, preventability assessments depended on available documentation and expert judgment. DeCoDe panels evaluated whether changes in health care–seeking behaviors, health care access, quality of care, or health care system performance within the local context might plausibly have averted the death, but these assessments were not designed to quantify the population-level impact of specific interventions. Preventability findings should therefore be interpreted as indicators of actionable gaps and opportunities for improvement, not as causal estimates of how many deaths would be prevented by any single intervention.

Future Directions

CHAMPS demonstrates that standardized postmortem surveillance can generate actionable information for families, clinicians, health care systems, and public health programs. In the future, CHAMPS will measure changes in causes of death as health policies evolve and will adapt surveillance methods to align with changing country priorities, local governance, and mortality review and civil registration systems. Although the full CHAMPS model requires specialized infrastructure and trained personnel, select components (e.g., standardized MITS training, targeted sampling, tiered diagnostics, streamlined DeCoDe review, and structured linkage to perinatal and child death review) could be adapted for sentinel mortality surveillance in more locations, special studies, and outbreak investigations.

Conclusion

These findings indicate that many fetal, neonatal, and child deaths in high-mortality settings remain preventable with timely, evidence-based interventions. Important contributors varied by age group and included perinatal asphyxia or hypoxia, sepsis, preterm birth complications, lower respiratory infections, congenital birth defects, malnutrition, malaria, and maternal conditions associated with stillbirth and neonatal death. Many of these contributors can be addressed through established maternal, newborn, and child health programs when delivered with sufficient quality, timeliness, and equity. By providing standardized, laboratory-supported mortality surveillance, CHAMPS has improved understanding of when, where, and why these deaths occur and has identified modifiable gaps in care and prevention. As countries work toward the United Nations’ targets for maternal, newborn, and child survival, integrating pathology, microbiology, epidemiology, and mortality review within national systems can help bridge diagnostic gaps, strengthen local capacity, and support prevention of avoidable deaths.