Ecological interventions are needed to control schistosomiasis, a neglected tropical disease transmitted to humans via freshwater snails. Schistosomiasis affects more than 200 million people worldwide, mostly in tropical low-income countries. Infected snails release infectious larvae into the water where humans wade, bathe, fish, and do other household activities. History shows that an integrated approach combining snail control and drug administration works best to reduce transmission risk long-term. However, molluscicides and other snail control methods are not feasible unless they can be targeted at the precise areas that pose the greatest risk to humans. Here we use drone-based aerial imagery of snail habitat validated with fine-scale aquatic snail and parasite distribution data, and human re-infection data from 16 villages tracking 1460 children in the Senegal River Basin to ask: what environmental variables best predict human schistosomiasis risk? We use machine-learning habitat classification models to identify snail habitat from drone imagery and precisely quantity the total coverage of good snail habitat at each village water point where people are exposed. We compare the fit of models aimed at predicting human reinfection through combinations of variables measuring suitable snail habitat, snail density, snail infection prevalence, human risk factors (age, sex), habitat characteristics (lake site vs. river site), and random effects of village and time.
Results/Conclusions
The total area of snail habitat within 30m of shore was the best predictor of disease risk (marginal R2=0.41, conditional R2=0.53), outperforming snail density and snail infection prevalence, which are traditionally the metrics of choice in public health efforts, as alternative risk predictors. We hypothesized that some residual variation in the model is due to unsampled snail habitat in deeper water beyond the accessible water access point. Therefore, we sampled deeper water and found dense populations of intermediate host snails inhabiting aquatic vegetation farther offshore than can be sampled manually. The next phase of this work will analyze snail habitat at increasing distances from shore to determine whether variation in human risk is best predicted by a particular “sphere of influence” (spatial scale) of snail habitat. Understanding the spatial scale of disease transmission will inform the development and deployment of new, targeted snail control methods to interrupt the cycle of infection between snails and humans. This is especially salient in the Senegal River Basin, where an intense schistosomiasis epidemic has been sustained for several decades despite mass drug administration campaigns to curb disease burden.