Authors
O Eneanya1; O Cano2; T Garske1; C Donnelly1; 1 Imperial College London, UK; 2 London School of Hygiene & Tropical Medicine, UK Discussion
Lymphatic filariasis (LF) is a mosquito-borne parasitic disease and a major cause of disability worldwide. It is one of the neglected tropical diseases identified by the WHO for elimination as a public health problem by 2020. Maps of disease distribution and environmental suitability are a necessary tool to delineate endemic foci and target scarce control resources. Here, we used pre-intervention occurrence data from 717 out of a possible 774 implementation units (IUs) collected during extensive mapping surveys by the Health Ministry. Using an ensemble of machine learning modelling algorithms (generalised boosted models (GBM) and random forest (RF)), we predicted the ecological niche of LF at a spatial resolution of 1 km. By overlaying modelled population distribution maps, we estimated population living in LF risk areas on a cell-by-cell scale. Our maps demonstrate that there is a heterogeneous distribution of LF across Nigeria, with large portions of northern Nigeria having more environmentally suitable climate able to drive disease transmission. Here we estimated that approximately 92 million individuals are at risk of transmission. Machine learning and ensemble modelling are a powerful tool to map disease risk, and are known to construct more accurate predictive models while decreasing the uncertainty of single models. We hope that this map will help target and assess the potential impacts of LF control measures. 
