Estimating the timing of stillbirths in countries worldwide using a Bayesian hierarchical penalized splines regression model

TL;DR

This paper introduces a Bayesian hierarchical model to estimate the timing of stillbirths worldwide, addressing data variability and regional differences to inform prevention strategies.

Contribution

It develops a novel Bayesian penalized splines regression framework that accounts for data heterogeneity and regional pooling to estimate stillbirth timing globally.

Findings

Intrapartum stillbirth proportion is generally decreasing over time.

Progress in reducing intrapartum stillbirths is slower in Sub-Saharan Africa.

The model effectively handles data variability and regional differences.

Abstract

Reducing the global burden of stillbirths is important to improving child and maternal health. Of interest is understanding patterns in the timing of stillbirths -- that is, whether they occur in the intra- or antepartum period -- because stillbirths that occur intrapartum are largely preventable. However, data availability on the timing of stillbirths is highly variable across the world, with low- and middle-income countries generally having few reliable observations. In this paper we develop a Bayesian penalized splines regression framework to estimate the proportion of stillbirths that are intrapartum for all countries worldwide. The model accounts for known relationships with neonatal mortality, pools information across geographic regions, incorporates different errors based on data attributes, and allows for data-driven temporal trends. A weighting procedure is proposed to account for unrepresentative subnational data. Results suggest that the intrapartum proportion is generally decreasing over time, but progress is slower in some regions, particularly Sub-Saharan Africa.