Infant-level and child-level predictors of mortality in low-resource settings: the WHO Child Mortality Risk Stratification Multi-Country Pooled Cohort
Ahmed T., Ali W., Argaw A., Bahl R., Bailey J., Baqui A., Becquey E., Berkley JA., Brown K., Chisti MJ., Chowdhury R., Diallo HA., Duggan C., Evans D., Fawzi W., Goga A., Grais R., Guindo O., Hamer D., Hess S., Huybregts L., Isanaka S., Jeha F., Kabakyenga J., Kabore P., Kaldenbach S., Kerac M., Khan MA., Khanam R., Kissoon N., Kounnavong S., Lachat C., LaGrone L., Le Port A., Lelijveld N., Leroy J., Mahfuz M., Manary M., Manji KP., Marconi S., McGrath M., Mohan VR., Moore S., Mugisha NK., Mupere E., Mwaringa S., Natarajan SK., Ngari M., Nisar I., Nisar YB., Olney D., Ouedraogo JB., Prentice A., Prost A., Roberfroid D., Rocker P., Rollins N., Ruel M., Saleem A., Sazawal S., Schwinger C., Singa B., Stobaugh H., Strand TA., Timbwa M., Toe LC., Trehan I., Trilok-Kumar G., Voskuijl W., Walson JL., Wang D., Wiens M., Ayushi None.
Background: Despite impressive reductions in overall global child mortality, the rate of decline has slowed during the past decade. Current guidelines for the care of paediatric patients in low-resource settings mostly focus on broad clinical syndromes or undernutrition rather than children's individual contextualised risk. We aimed to identify readily assessable child-level characteristics that can predict mortality risk in a range of community and health-care settings in high-burden settings. Methods: The WHO Child Mortality Risk Stratification Multi-Country Pooled Cohort (WHO-CMRS) included pooled data from individual children enrolled in observational or randomised controlled trials in low-income and middle-income countries. The criteria for inclusion of a dataset were documentation of age, weight, vital status, and date of death, and at least two observations per participant younger than 60 months. To calculate odds ratios, we built generalised linear mixed effects regression (glmer) models with each child and each study as random intercepts and time interval as the offset. In all analyses, the outcome was defined as death within the respective observation period of the child. From the glmer models, we predicted absolute risk of death per child-month associated with risk exposures separately and combined with anthropometry according to the following age groups: 0–5 months, 6–11 months, 12–23 months, and 24–59 months. Studies were grouped according to population types studied: the general population, populations selected based on anthropometric criteria, and populations selected based on the presence of illness. Findings: We analysed pooled data from WHO-CMRS, including 75 287 children from 33 studies done in 17 countries between Jan 1, 2001, and Dec 31, 2021. During a total of 69 085 child-years of follow-up, 2805 (3·7%) children died. Age younger than 24 months, low anthropometry, preterm birth, low birthweight, and absence of breastfeeding (either was breastfeeding not offered or an underlying illness interfered with breastfeeding practices) were each associated with increased mortality: risks declined with increasing age. The highest absolute mortality risk was among the youngest children (age 0−5 months), with a weight-for-age Z score of less than −3 (ie, a predicted absolute risk of 11·0 [95% CI 6·2−19·5] per 1000 child-months in general population studies). Risks were additive: underlying risk exposures such as low birthweight and preterm birth added to the mortality risks in children with anthropometric deficit. For example, children aged 0−5 months with a weight-for-age Z score of less than −3 and a history of preterm birth had a predicted absolute mortality risk of 40·1 (95% CI 22·0−72·1). However, overall mortality and the association between child-level characteristics and mortality differed according to the type of study population and child age. Interpretation: Risk assessments combining individual child-level characteristics including anthropometry can enable programmes to identify children at high and lower risk of mortality and, thereafter, differentiate care accordingly. Such a strategy could reduce mortality and optimise health system efficiency and effectiveness. Funding: US Agency for International Development. Translations: For the Spanish and French translations of the abstract see Supplementary Materials section.