AI models uncover factors influencing scorpionism in Northern Brazil

Abstract

Envenomation by scorpion stings is a serious public health problem in tropical regions of the world. In Brazil’s Northern region, there has been a significant increase in cases over the last decade, accompanied by a rise in the fatality rate. Climate change and intensive land use are altering the distribution of species that pose health risks and may be associated with the increased incidence of accidents. We integrated species distribution models (SDMs) of three medically important species (Tityus obscurus, T. metuendus, and T. silvestris), bioclimatic data, and land use to predict scorpionism incidence and quantify the importance of predictors in Northern Brazil. We used these predictors to build a model to predict the incidence of scorpion envenomations using the XGBoost artificialintelligence (AI) algorithm and assessed the importance of the predictor variables with the Shapley method. Our models demonstrated good performance in predicting incidence, with a MAE of 7.17 and an RMSE of 10.62. The analysis identified that climatic factors are the main determinants of incidence but also highlighted the relevance of the distribution of T. obscurus and T. silvestris species, pasture areas, and rural population density. The study showed that integrating SDMs and AI techniques is effective for predicting scorpionism incidence and assisting in the formulation of prevention as well as management strategies.