Ai-Driven Predictive Models for Maternal Mortality Reduction in Rural Nigeria: A Systematic Review
Abstract
Maternity mortality in rural Nigeria is a pressing social health problem because the delay in seeking care, poor referral services, and absence of risk categorization still compromise maternal health. Predictive models developed using artificial intelligence (AI) have turned out to be a promising solution to the disadvantages of maternal health, although their role in reducing maternal mortality situations in rural Nigerian settings has not been systematically synthesised. This research paper conducted a PRISMA-style 2020 systematic review to determine the relevance of AI-driven predictive models to maternal mortality prevention in rural Nigeria and similar low- and middle-income countries. The search of PubMed, Scopus, Web of Science, IEEE Xplore, and African Journals Online included the research that used machine learning, deep learning, ensemble, or hybrid AI to predict maternal mortality or similar outcomes. Nine qualitative studies were considered. No research directly predicted maternal mortality in rural Nigeria. Rather, AI models mainly dealt with determinants and clinical antecedents of maternal death, anticipating antenatal care utilisation, skilled birth attendance, home delivery, pregnancy loss, maternal near miss events, obstetric complications, and maternal deterioration. The most commonly used models were based on random forests and tended to provide moderate-to-high levels of predictive validity, with reported AUC scores exceeding 0.80, but should not be generalised, as the models most frequently incorporated internal validation and few health-system delay factors. On the whole, the presented evidence suggests that the area of AI application in maternal health improvement has significant potential and presents a significant gap in translating predictive modelling into explicit maternal mortality reduction in rural Nigeria. Subsequent studies must continue to focus on context-conscious, mortality-centering AI models incorporating clinical, sociodemographic, and delay-in-layer health-system extrinsic to the present study, with dependable extrinsic validation and ethical rules and principles.