Small area estimates of under-five, under-one and neonatal mortality in low- and middle-income countries from 2000 to 2021

Erin May, Institute for Health Metrics and Evaluation, University of Washington
Nathaniel Henry, Henry Spatial Analysis
Haley Comfort, University of Washington - IHME
Darwin Jones, Institute for Health Metrics and Evaluation, University of Washington
Stefanie Watson, University of Washington
Denny Wang, University of Washington
Nicholas Verghese, University of Washington
Reiner Robert, University of Washington, Seattle
Austin E. Schumacher, University Of Washington

Reducing preventable child mortality is listed in the Sustainable Development Goals as a critical target for ensuring healthy lives and well-being at all ages; as such, it is important that policy makers have an accurate understanding of where child mortality is highest at a small spatial granularity (such as the district level) to effectively target interventions. To meet this demand, we implement small area estimation (SAE) models to estimate sex-specific under-five, under-one and neonatal mortality rates at the second administrative level for low and middle income countries from 2000-2021. This builds upon previous work to estimate child mortality in sub-Saharan Africa. We synthesized complete birth histories (CBH) primarily from the Demographic and Health Surveys Program between 2000 and 2021 as model inputs. We deviate from previous models by using SAE methods to implement a Bayesian hierarchical spatiotemporal model with correlated autoregressive (CAR) random effects. Furthermore, we use regional models to borrow strength across countries and incorporate covariates such as lag distributed income and education to inform our estimates. Our results show substantial heterogeneity in child mortality across and within countries. Overall, these estimates are an indispensable tool policy makers can use to gain a more granular understanding of child mortality.

Keywords: Spatial Demography, Children, Adolescents, and Youth, Mortality and Longevity, Geo-referenced/geo-coded data

See extended abstract.