From risk factors to predictive modelling: applying machine learning to childhood malaria surveillance in resource-limited settings

Comprehensive Summary

This study analyzed malaria risk among 693 under-five children living in Nigerian internally displaced persons camps using traditional epidemiological methods and machine learning models. Malaria prevalence was high (68.5%), with key risk factors including caregiver education level, occupation, and camp of residence. Among four models tested, the Random Forest algorithm performed best (AUC = 0.892), highlighting its ability to capture complex, non-linear relationships. Despite high caregiver knowledge of malaria prevention, bed net use was extremely low, revealing a major knowledge-practice gap.

Outcomes and Implications

The findings demonstrate that machine learning, particularly Random Forest models, can significantly improve malaria risk prediction in resource-limited settings. These tools can support precision public health by identifying high-risk populations for targeted interventions such as vector control, bed net distribution, and focused health education. Integrating ML models into malaria surveillance systems may enhance resource allocation and improve disease control outcomes in humanitarian settings.