Explainable Artificial Intelligence in Multimodal Malaria Prediction: A Systematic Review and Roadmap Integrating Climate Change, Parasite Genomics, and Public Health Decision Support

Malaria remains a persistent global public health burden, particularly in sub-Saharan Africa, requiring predictive systems that are both accurate and clinically interpretable. Although artificial intelligence (AI) has significantly improved malaria forecasting and diagnosis, real-world deployment remains constrained by opaque “black-box” decision pathways and fragmented modelling approaches that analyse climatic, genomic, and clinical drivers in isolation. This systematic review synthesises contemporary evidence on Explainable Artificial Intelligence (XAI) applications in malaria prediction, integrating climate vulnerability and parasite genomic insights. The review was conducted and reported in accordance with PRISMA 2009 guidelines. Four databases (PubMed, IEEE Xplore, SpringerLink, and ScienceDirect) were systematically searched, yielding 161 records; 12 studies published between 2020 and 2025 met the inclusion criteria. The restricted timeline reflects the recent surge in model-agnostic interpretability methods such as SHAP and LIME. Findings indicate that ensemble learning models, particularly Random Forest and XGBoost, demonstrate robust predictive performance and strong post hoc interpretability in climate-driven and clinical forecasting contexts. In contrast, deep learning architectures, including Convolutional Neural Networks and Transformers, excel in image-based and genomic classification tasks but require interpretability overlays to mitigate opacity. Despite methodological advances, Causal AI remains underutilised, limiting current models' capacity to simulate intervention scenarios and inform policy decision-making. Furthermore, no fully unified multimodal framework that integrates climatic, genomic, and clinical features into a single explainable architecture was identified. This review not only synthesises trends in algorithmic performance and interpretability but also establishes a strategic roadmap for developing unified, policy-integrated multimodal XAI systems that support transparent malaria risk prediction amid accelerating climate change and parasite evolutionary dynamics.