Initial findings from a novel population-based child mortality surveillance approach: a descriptive study

Background: Sub-Saharan Africa and south Asia contributed 81% of 5·9 million under-5 deaths and 77% of 2·6 million stillbirths worldwide in 2015. Vital registration and verbal autopsy data are mainstays for the estimation of leading causes of death, but both are non-specific and focus on a single underlying cause. We aimed to provide granular data on the contributory causes of death in stillborn fetuses and in deceased neonates and children younger than 5 years, to inform child mortality prevention efforts. Methods: The Child Health and Mortality Prevention Surveillance (CHAMPS) Network was established at sites in seven countries (Baliakandi, Bangladesh; Harar and Kersa, Ethiopia; Siaya and Kisumu, Kenya; Bamako, Mali; Manhiça, Mozambique; Bombali, Sierra Leone; and Soweto, South Africa) to collect standardised, population-based, longitudinal data on under-5 mortality and stillbirths in sub-Saharan Africa and south Asia, to improve the accuracy of determining causes of death. Here, we analysed data obtained in the first 2 years after the implementation of CHAMPS at the first five operational sites, during which surveillance and post-mortem diagnostics, including minimally invasive tissue sampling (MITS), were used. Data were abstracted from all available clinical records of deceased children, and relevant maternal health records were also extracted for stillbirths and neonatal deaths, to incorporate reported pregnancy or delivery complications. Expert panels followed standardised procedures to characterise causal chains leading to death, including underlying, intermediate (comorbid or antecedent causes), and immediate causes of death for stillbirths, neonatal deaths, and child (age 1–59 months) deaths. Findings: Between Dec 10, 2016, and Dec 31, 2018, MITS procedures were implemented at five sites in Mozambique, South Africa, Kenya, Mali, and Bangladesh. We screened 2385 death notifications for inclusion eligibility, following which 1295 families were approached for consent; consent was provided for MITS by 963 (74%) of 1295 eligible cases approached. At least one cause of death was identified in 912 (98%) of 933 cases (180 stillbirths, 449 neonatal deaths, and 304 child deaths); two or more conditions were identified in the causal chain for 585 (63%) of 933 cases. The most common underlying causes of stillbirth were perinatal asphyxia or hypoxia (130 [72%] of 180 stillbirths) and congenital infection or sepsis (27 [15%]). The most common underlying causes of neonatal death were preterm birth complications (187 [42%] of 449 neonatal deaths), perinatal asphyxia or hypoxia (98 [22%]), and neonatal sepsis (50 [11%]). The most common underlying causes of child deaths were congenital birth defects (39 [13%] of 304 deaths), lower respiratory infection (37 [12%]), and HIV (35 [12%]). In 503 (54%) of 933 cases, at least one contributory pathogen was identified. Cytomegalovirus, Escherichia coli, group B Streptococcus, and other infections contributed to 30 (17%) of 180 stillbirths. Among neonatal deaths with underlying prematurity, 60% were precipitated by other infectious causes. Of the 275 child deaths with infectious causes, the most common contributory pathogens were Klebsiella pneumoniae (86 [31%]), Streptococcus pneumoniae (54 [20%]), HIV (40 [15%]), and cytomegalovirus (34 [12%]), and multiple infections were common. Lower respiratory tract infection contributed to 174 (57%) of 304 child deaths. Interpretation: Cause of death determination using MITS enabled detailed characterisation of contributing conditions. Global estimates of child mortality aetiologies, which are currently based on a single syndromic cause for each death, will be strengthened by findings from CHAMPS. This approach adds specificity and provides a more complete overview of the chain of events leading to death, highlighting multiple potential interventions to prevent under-5 mortality and stillbirths. Funding: Bill & Melinda Gates Foundation. The CHAMPS Network included sites in seven countries: Baliakandi, Bangladesh; Harar and Kersa, Ethiopia; Siaya and Kisumu, Kenya; Bamako, Mali; Manhiça, Mozambique; Bombali, Sierra Leone; and Soweto, South Africa. Site characteristics and selection criteria have been described previously.17 As of Dec 31, 2018, five sites had implemented minimally invasive tissue sampling, and data from these sites are included in this report. Some sites used pre-existing Health and Demographic Surveillance System (HDSS) catchment areas.18, 19, 20, 21 New surveillance areas were developed for CHAMPS within informal settlements in Kisumu, Baliakandi, and Soweto. The total population size for catchment areas within each country (with the exception of South Africa) ranged between 170 000 and 227 219 individuals.21 The Soweto site, in the absence of an established HDSS, used the entire township and surrounding informal settlement areas (population of approximately 1·3 million people) as the catchment area during the majority of the reported period while the HDSS was being established. Sites built relationships with communities22 and health facilities within catchment areas, posting CHAMPS staff at health facilities and developing networks of community reporters to receive notifications of under-5 deaths and stillbirths within the first 24 h.17 At most sites, death notifications were initially mostly received from primary referral health facilities, subse-quently expanding to notifications of deaths occurring outside health facilities. After receiving notifications of deaths or stillbirths, CHAMPS staff approached families rapidly for eligibility screening. Eligible cases were aged younger than 60 months, and they or their parents resided within the study catchment area. MITS eligibility required that CHAMPS staff were notified within 36 h of death (or ≤72 h if post-mortem refrigeration used) and that the body of the stillborn fetus or deceased child was available for analysis. Families provided consent for the MITS procedure, verbal autopsy, and clinical data abstraction, and we present data on these MITS-eligible fetuses, neonates, or children in this report. Stillborn fetuses and deceased neonates and infants who were not eligible for the MITS procedure (eg, CHAMPS received a notification more than 36 h after their death or consent for MITS was not given) were asked to consent for only verbal autopsy interview and clinical data abstraction, and will be reported elsewhere.17 Parents or guardians of stillborn fetuses or deceased children provided written informed consent before collection of data, specimens, or information on the mothers. Ethics committees overseeing investigators at each site and at Emory University (Atlanta, GA, USA) approved overall and site-specific protocols, as appropriate. The Centers for Disease Control and Prevention (CDC; Atlanta, GA, USA) relied on the Emory University committee to review the overall protocol and on appropriate ethical review committees at the sites where CDC staff were directly engaged. Protocols have been published previously. Trained CHAMPS staff took photographs to identify dysmorphic features and took anthropometric measurements before specimen collection. In sterile conditions, tissue specimens were collected using biopsy needles from the lungs, heart, brain, liver, and bone marrow.23 They also collected peripheral blood, cerebrospinal fluid, stool, and nasopharyngeal secretions. Site laboratories tested post-mortem blood samples for HIV DNA or RNA by PCR, malaria thick and thin smears, and rapid diagnostic tests; blood and cerebrospinal fluid were cultured. Five custom TaqMan Array Cards (ThermoFisher Scientific, Waltham, MA, USA) with specific molecular assays were used to detect 116 pathogen targets from lung tissue, blood, cerebrospinal fluid, rectal brush (collecting stool), and nasopharyngeal swabs.23, 24 Pathology laboratories at each site and at the CDC applied histopathology to all tissues, including routine stains. Special stains, such as tissue Gram stain, targeting microorganisms and immunohistochemistry were done when relevant, at pathology laboratories at the CDC.25 Data were abstracted from all available clinical records of deceased children and relevant maternal health records were abstracted for stillbirths, neonatal deaths, and others with reported pregnancy or delivery complications. Trained interviewers used appropriate-language translations of the 2016 WHO verbal autopsy instrument17, 26 to interview caregivers of enrolled deceased children within 4 weeks of death, when feasible. All data available for each case were reviewed by Determination of Cause of Death (DeCoDe) panels convened at each site, which consisted of paediatricians, obstetricians, other health-care providers, epidemiologists, pathologists, and microbiologists.27 The panels reviewed available case data and determined the chain of events leading to death using WHO ICD-10 and WHO application of ICD-10 to deaths during the perinatal period (ICD-PM) guidelines.28, 29 For cases in which more than one condition led to death, the process included documenting the entire causal chain leading to death, as in a standard death certificate, including underlying cause, intermediate (referred to on the WHO standard death certificate as comorbid or antecedent causes), and immediate cause. To be included in the causal chain, the panels considered whether appropriate management or prevention of that condition could have prevented the death. Additionally, the panels identified any other contributing causes of death and the main maternal factors contributing to perinatal deaths. The immediate cause of death was the most proximal cause leading to death, and antecedent or comorbid causes were others contributing directly to the chain of events leading to death. The underlying cause was the disease or injury that initiated the chain of events that led to the death. Each case could have one immediate cause and multiple comorbid or antecedent causes identified, which could be in the same general category (eg, lower respiratory tract infection could be listed twice in the same causal chain if, for example, a viral pneumonia led to a bacterial superinfection). Additionally, the panels considered all available data for evidence of signs of life (such as postnatal respiration, heartbeat, or movement); the presence of any resulted in a case being categorised as a neonatal death rather than a stillbirth. The DeCoDe process was standardised across the network using diagnosis standards for defining and coding common childhood causes of death and organisation of specific and homogeneous training at all sites.27 Causes were grouped into standardised categories for analysis (appendix p 2). All analyses were conducted using R (version 3.6.3). The funder of the study provided input on study design and site selection and suggested edits on the final versions of the manuscript, but had no role in data collection, data analysis, or data interpretation. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.