Machine learning for maternal risk prediction in resource- constrained antenatal care: evidence-to-implementation gaps and implications for Sustainable Development Goal 3
Résumé
Abstract Objective Machine-learning (ML) models are increasingly proposed for earlier recognition of maternal risk, but strong discrimination does not itself establish clinical usefulness. This Research Note maps recent maternal-health ML evidence against practical requirements for resource-constrained antenatal care (ANC) and examines how prediction could contribute to Sustainable Development Goal (SDG) 3. Results Across the mapped studies and reviews, technical feasibility was common but prospective evidence of clinical benefit was limited. A Rwandan random-forest model reported 90.6% accuracy and an AUROC of 0.85 but only 46.5% recall, illustrating how headline accuracy can obscure missed adverse cases. A northern Ugandan pre-eclampsia model achieved an AUC of 0.849 but depended on laboratory and uterine-artery Doppler inputs that may not be consistently available. Recent syntheses also identify incomplete external validation, calibration, subgroup assessment and implementation evaluation. The evidence supports a prediction-to-action pathway in which predictor availability, temporal validity, calibration, equity, workflow integration, referral capacity and prospective impact are evaluated alongside discrimination. ML can plausibly support SDG 3 only when reliable predictions trigger accessible, timely and effective care.
Citer ce document
Exporter : BibTeX · RIS (Zotero, Mendeley, EndNote)
Accès au document
Texte intégral archivé sur SPHAERO, en accès ouvert : lisible gratuitement en ligne après connexion. Le document reste aussi disponible à sa source. Se connecter
Voir l'article sur le site de la revueLicence et provenance
Licence : CC BY
Notice moissonnée depuis OpenAlex le 16/11/2026. Le document reste hébergé par sa source.
Voir le document à la source →