Bayesian spatiotemporal modelling of neonatal, infant and under-5 mortality (2000-2022) in 41 Asian countries: a...

Child mortality continues to pose a major public health challenge across Asia. This study examines trends in under-5, infant and neonatal mortality and identifies key determinants, spatial risk patterns and projections through 2030 using spatiotemporal modelling.

We used national-level data from 41 Asian countries, representing over 80% of Asia's population, between 2000 and 2022, incorporating 26 health, environmental and sociodemographic indicators. A hierarchical Bayesian model using Integrated Nested Laplace Approximation, incorporating fixed effects, spatially structured and unstructured random effects, and temporal smoothing, was used. Model performance was assessed via the Deviance Information Criterion, Watanabe-Akaike Information Criterion, coefficient of determination (R²), root mean squared error (RMSE) and mean absolute error metrics.

Under-5 mortality decreased significantly (p<0.001) from 46.73 to 18.53 per 1000 live births. Strong negative associations were observed between child mortality and vaccination coverage-rubella (r=-0.79), Diphtheria, Tetanus, and Pertussis (DTP) (r=-0.74), hepatitis B (r=-0.58) and rotavirus (r=-0.20). Female literacy (r=-0.20) and life expectancy (r=-0.30) also contributed to improved outcomes. Spatial analyses identified Afghanistan (under-5: 77.12), Bangladesh (57.12) and Myanmar (63.16) as high-risk hotspots, while Japan, Sri Lanka and the UAE maintained low predicted rates (≈0). Neonatal mortality patterns were flatter across time and space, peaking in Bangladesh (6.30), Indonesia (5.08) and Azerbaijan (5.00). Predictive accuracy was highest for neonatal mortality (R²=0.99; RMSE=1.93). Some countries, such as Yemen and the UAE, displayed near-zero or negative forecasts, suggesting sensitivity to spatial priors in sparse-data contexts.

The study highlights the critical role of immunisation and maternal education in reducing mortality, and the need for more targeted neonatal interventions. The white-box modelling framework enables both interpretability and reliable forecasting, supporting data-driven policy planning toward achieving advanced, equitable child survival, as outlined in Sustainable Development Goal 3.2.