Prioritizing Health Care Strategies to Reduce Childhood Mortality

Importance: Although child mortality trends have decreased worldwide, deaths among children younger than 5 years of age remain high and disproportionately circumscribed to sub-Saharan Africa and Southern Asia. Tailored and innovative approaches are needed to increase access, coverage, and quality of child health care services to reduce mortality, but an understanding of health system deficiencies that may have the greatest impact on mortality among children younger than 5 years is lacking. Objective: To investigate which health care and public health improvements could have prevented the most stillbirths and deaths in children younger than 5 years using data from the Child Health and Mortality Prevention Surveillance (CHAMPS) network. Design, Setting, and Participants: This cross-sectional study used longitudinal, population-based, and mortality surveillance data collected by CHAMPS to understand preventable causes of death. Overall, 3390 eligible deaths across all 7 CHAMPS sites (Bangladesh, Ethiopia, Kenya, Mali, Mozambique, Sierra Leone, and South Africa) between December 9, 2016, and December 31, 2021 (1190 stillbirths, 1340 neonatal deaths, 860 infant and child deaths), were included. Deaths were investigated using minimally invasive tissue sampling (MITS), a postmortem approach using biopsy needles for sampling key organs and fluids. Main Outcomes and Measures: For each death, an expert multidisciplinary panel reviewed case data to determine the plausible pathway and causes of death. If the death was deemed preventable, the panel identified which of 10 predetermined health system gaps could have prevented the death. The health system improvements that could have prevented the most deaths were evaluated for each age group: stillbirths, neonatal deaths (aged <28 days), and infant and child deaths (aged 1 month to <5 years). Results: Of 3390 deaths, 1505 (44.4%) were female and 1880 (55.5%) were male; sex was not recorded for 5 deaths. Of all deaths, 3045 (89.8%) occurred in a healthcare facility and 344 (11.9%) in the community. Overall, 2607 (76.9%) were deemed potentially preventable: 883 of 1190 stillbirths (74.2%), 1010 of 1340 neonatal deaths (75.4%), and 714 of 860 infant and child deaths (83.0%). Recommended measures to prevent deaths were improvements in antenatal and obstetric care (recommended for 588 of 1190 stillbirths [49.4%], 496 of 1340 neonatal deaths [37.0%]), clinical management and quality of care (stillbirths, 280 [23.5%]; neonates, 498 [37.2%]; infants and children, 393 of 860 [45.7%]), health-seeking behavior (infants and children, 237 [27.6%]), and health education (infants and children, 262 [30.5%]). Conclusions and Relevance: In this cross-sectional study, interventions prioritizing antenatal, intrapartum, and postnatal care could have prevented the most deaths among children younger than 5 years because 75% of deaths among children younger than 5 were stillbirths and neonatal deaths. Measures to reduce mortality in this population should prioritize improving existing systems, such as better access to antenatal care, implementation of standardized clinical protocols, and public education campaigns. CHAMPS collects standardized, population-based, surveillance data from sites with high child mortality to understand and track preventable causes of death. CHAMPS currently includes sites in 7 countries: Bangladesh, Ethiopia, Kenya, Mali, Mozambique, Sierra Leone, and South Africa. CHAMPS is not intended to be representative of entire countries but rather focuses on regions where mortality rates are known to be highest. To that end, understanding causes of death (and what it would take to prevent them) at these sites could have the greatest impact in terms of reducing mortality. By design, CHAMPS does not reflect low mortality areas. The CHAMPS database contains comprehensive data on all stillbirths and deaths among children younger than 5 years enrolled at each of the surveillance sites. These data include demographic characteristics, extensive postmortem diagnostic results, clinical medical record abstraction data for each child and, when appropriate, maternal antenatal records and verbal autopsy data (as well as social autopsy data in Sierra Leone). Site characteristics, selection criteria, catchment areas, death notification methods, eligibility screening, and specimen and data collection methods have been previously described.10,11 Limitations of the CHAMPS methodology have been documented elsewhere10,12,13 and include the inability to include all deaths within catchment areas, disparate population characteristics between sites, and overrepresentation of health care facility–based deaths. Ethical approval was obtained for use of CHAMPS data by each site’s ethical review board and by the Emory University Rollins School of Public Health. Parents or guardians of stillborn fetuses or deceased children provided written informed consent before collection of data, specimens, or information on the mothers. All cases were anonymized prior to review. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cross-sectional studies were followed. Details regarding the cause of death determination and standardization across sites processes have been described elsewhere.11,12 Briefly, deaths are investigated with minimally invasive tissue sampling (MITS), a postmortem approach using biopsy needles for sampling key organs and body fluids. The samples undergo testing using conventional microbiology and multiplexed polymerase chain reaction (PCR) assays using TaqMan array cards; tissues are also examined by pathologists and subject to more advanced histopathological tests. Any available data regarding the terminal events are abstracted from medical records and verbal autopsy and recorded from caregiver recollection. A determination of cause of death (DECODE) panel consisting of pediatricians, obstetricians, epidemiologists, pathologists, microbiologists, and other health care professionals review case data at each surveillance site to assign causes of death. CHAMPS uses the WHO International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) and the WHO application of ICD-10 deaths during the perinatal period (ICD-PM).14,15 For deaths in which only a single cause led to death, that cause is listed as the underlying cause. For deaths in which multiple causes led to the death, the panel determines the causal chain including the underlying, antecedent, and immediate causes leading to death.13 The underlying cause usually occurred before immediate or antecedent conditions and may have predisposed the child to an immediate cause or comorbid illnesses that then led to death; the immediate cause was closest to the death, and the antecedent causes were in between the underlying and immediate causes. Each death has only 1 underlying cause, zero or 1 immediate cause, and zero or more antecedent causes. At the site level, a subset of cases that underwent DECODE review are shared with the other sites for secondary review as a quality control measure. For each death, the DECODE panel determined whether the death was preventable (yes, no, or under certain circumstances) by considering all the information available for each case, which may include demographic, clinical, pathological, microbiological, verbal autopsy, photography, and anthropometric measurements. The definition of preventability mainly captures the conditions immediately surrounding the death of that particular child and not the broader global political, financial, and social influences. If the death was deemed potentially preventable, the panel identified predetermined health system gaps (Table) and recommended improvements based on those gaps that could have prevented the death. These 10 categories emerged from categorization of the free text responses derived from the first DECODE panel evaluations from 2016 to 2017. These categories are still evolving. Each death could have multiple prevention categories listed. For each preventable death, the panel also had the option to provide specific public health action recommendations beyond the 10 categories in an open text field, which were subsequently categorized as well. There were 10 high-level categories of health system improvements (Table), and each death could have multiple prevention categories listed. Although some health system improvement categories primarily target specific age groups (eg, improvements in antenatal and obstetric care), any category implemented at a given site may affect children in other age groups to varying degrees. We evaluated health system improvement categories across all sites by age group. We defined stillbirths as the death of a baby before or at delivery, neonates as those aged 0 to 27 days, and infants and children as those aged 28 days to younger than 5 years.16 To determine which health system improvements could have prevented the most deaths regardless of cause, we generated every combination of 1 to 10 categories (1023 combinations) and calculated how many deaths could have been prevented for each combination under the assumption (A1) that all health system improvement categories recommended for a single death are necessary to prevent that death. We also conducted sensitivity analyses assuming (A2) deaths would be reduced proportionally to the number of categories implemented for deaths with multiple health system improvement categories noted, and (A3) any single category among categories recommended for each death is sufficient to prevent the death. For example, if 4 health system improvement categories were recommended for a set of deaths and only 1 was implemented, we calculated that, according to the 3 assumptions: (A1) those deaths would not be prevented, (A2) 25% of those deaths would be prevented, and (A3) 100% of those deaths would be prevented (eMethods in Supplement 1). All analyses were done in R version 4.1.2 (R Foundation for Statistical Computing).