Predicting schistosome transmission in rural Uganda using water contact data from wearable GPS devices

Abstract Granular spatial data and models of human behaviour are currently lacking for schistosomiasis. We collected ten days of wearable GPS logger data from 452 individuals in rural Uganda to model water contact as a proxy indicator of usage at 69 georeferenced open-water sites and 32 public taps or boreholes. Among participants, 63.9 and 33.2% visited at least one water site and tap or borehole, respectively. Exponential spatial decay models accurately predicted site-specific open-water contact (area under the receiver operating characteristic curve = 0.87) and tap or borehole usage (0.92). There was no evidence of tap or borehole usage influencing open-water contact. Incorporating mobility terciles did not improve simple spatial decay models. Integrating spatial decay-based estimates of open-water-site usage into an individual-based transmission model produced realistic estimates of one-year Schistosoma mansoni reinfection and provided a ranking of water sites contributing to transmission. Our spatial decay models offer scalable tools for focal interventions for schistosomiasis.