Improving Lassa fever risk mapping using self-organizing maps and spatial determinants.

Abstract

Lassa fever is a zoonotic disease endemic to West Africa, primarily transmitted through contact with Mastomys natalensis, its rodent reservoir. Existing risk models often rely on regression-based approaches, which may overlook interactions with human settlement and environmental conditions. This study introduces a Self-Organizing Maps (SOM) classification approach to improve Lassa fever risk prediction. SOM included high-dimensional environmental data, outbreak zones based on rodent suitability, and human settlement density to improve classification accuracy. Google Earth Engine was used to extract environmental predictors (1980-2022), including land surface temperature, vegetation index, potential evapotranspiration, elevation, and built-up areas [i.e., rodent occurrence data]. The model showed a high accuracy f > 0.89 during training and validation, confirming recurrent hotspots that aligned with outbreaks. The model demonstrated moderate to high-risk prediction for Sierra Leone and Liberia [100 %], Guinea [97.5 %], Nigeria [45.8 %], and Cameroon [77.6 %]. This research demonstrated the potential of innovative analytics to predict risk areas and outbreaks, offering a data-driven framework to improve Lassa fever monitoring and response strategies, thereby enhancing public awareness, control, and prevention in West Africa.