Building a Predictive Epidemiological Model for Lassa Fever Outbreak Risk in Nigeria Integrating Climate, Rodent Ecology, and Human Vulnerability Indices

Lassa fever causes annual epidemics in Nigeria with case fatality rates of 15 to 30 percent among hospitalized patients, yet operational outbreak risk prediction tools integrating multi-domain drivers do not exist for the Nigerian health system. This interdisciplinary study constructed and validated a spatiotemporal predictive epidemiological model for Lassa fever outbreak risk across all 36 Nigerian states using machine learning-integrated epidemiological modeling. Twelve years of Nigeria Centre for Disease Control confirmed Lassa case records (2012 to 2023) were combined with high-resolution satellite-derived climate variables including normalized difference vegetation index, rainfall anomalies, and land surface temperature representing Mastomys natalensis habitat suitability, human vulnerability indices from National Bureau of Statistics health access and WASH data, and rodent trapping density data from published ecological surveys. Random forest and gradient boosted tree ensemble models were trained to predict district-level outbreak probability 4 and 8 weeks ahead. Model performance was assessed by temporal cross-validation. A web-based early warning dashboard was prototyped with NCDC input. The ensemble model achieved area under the ROC curve of 0.91 for 4-week and 0.87 for 8-week outbreak prediction. Vegetation index peaks in the North Central and South South zones were the strongest predictors of subsequent outbreak activity. Human WASH deficit and hospital density were the most influential vulnerability covariates. This original predictive tool provides an operational basis for pre-positioned pre-outbreak response across Nigeria and represents a scalable framework for other rodent-borne viral hemorrhagic fevers in West Africa. Keywords: Lassa fever, outbreak prediction, machine learning, spatiotemporal model, Nigeria.