Predictive models for stillbirths and neonatal deaths: Protocol for a global scoping review with focus on sub-Saharan Africa.

<ns3:p>Background Sub-Saharan Africa (SSA) accounts for nearly half of the global burden of stillbirths and neonatal deaths. Many of these deaths are preventable through improved quality of maternal and newborn care, particularly timely risk recognition and clinical response. Advancements in artificial intelligence and its applications offer opportunities to improve care through real-time monitoring, assisting clinical decisions, and enhancing public health management, thereby contributing to better care and outcomes. This scoping review will map prediction models and modelling approaches for predicting stillbirths and neonatal deaths globally with focus on SSA. We will compare the performance of classical, machine learning (ML) and artificial intelligence (AI) methods, and examine model development, validation and transportation to SSA. Methods The scoping review will follow the framework proposed by Arksey and O’Malley. We will search five databases (EMBASE, Medline, Scopus, Web of Science, Global Health) and grey literature sources. Two reviewers will independently screen titles, abstracts, and full studies for inclusion based on inclusion criteria, focussing on three predictive approaches (classical, ML, and AI) for stillbirths and neonatal deaths globally, with emphasis to models developed in or transported to SSA. Data extraction will follow a conceptual framework encompassing six domains: predictors, data sources, model types, evaluation metrics, model performance, and model validation. Risk of bias will be assessed using PROBAST+AI. Findings will be summarised using tables, figures, and maps. Anticipated findings The review will identify and characterise predictive models for stillbirths and neonatal deaths, including classical statistical, ML, and AI approaches. We will map predictors, data sources used, evaluation metrics reported, and compare model performance across methodological approaches, and identify which models have been developed in or transported to SSA. Conclusions Findings will identify methodological gaps and inform future model development to support targeted interventions for reducing perinatal mortality in Sub-Saharan Africa.</ns3:p>