Under-five mortality estimation methods: A methodological systematic review

This methodological systematic review aimed to identify and synthesise the existing under-five mortality (U5M) estimation methods globally.

We searched seven databases including Medline, Embase, Scopus, Web of Science, CINAHL, Global Health, and ProQuest Central, as well as grey literature sources from inception to September 25, 2025. The review protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42023465476).

Fifty-six studies were included in this review. The most frequently employed child mortality estimation method was the indirect method (n = 16), followed by the Global Burden of Disease (GBD) study method (n = 12) and the direct method (n = 11). The most commonly applied models were spatiotemporal Gaussian process regression and the Bayesian B-spline bias-reduction model. Substantial variation was observed across studies in geographical scope, temporal coverage, data sources, uncertainty quantification, statistical modelling, and bias adjustment.

There are substantial variations in U5M estimation methods, with challenges in data availability, uncertainty estimation, and bias adjustment. These findings highlight the need to harmonise methodological approaches and refine estimation methods. Strengthening vital registration systems is essential to ensure accurate, reliable data to inform evidence-based decision-making and track progress towards U5M reduction targets.