The time- and space-varying roles of human mobility in shaping urban dengue epidemics

Dengue outbreaks continue to cause large morbidity and mortality globally. Previous work has shown that human movement alongside weather, climatic, ecological, and socioeconomic factors influence outbreak size, persistence, and geographical spread. However, the relative importance of human movement is unclear for the establishment, rapid expansion, and persistence of dengue in urban settings. Relatedly, the extent to which neighbourhoods differentially influence outbreaks over time is unclear. To address these gaps, we developed a multi-model framework that integrates Bayesian hierarchical modelling and data-driven, deep-learning-based approaches to parameter inference and out-of-sample probabilistic predictions of dengue in Fortaleza, Brazil. We apply our framework to dengue outbreaks at the neighbourhood level, using epidemiological surveillance data alongside public transportation data. We use information criteria, scoring rules, and explainability metrics to measure the models performance and ultimately explain how human mobility shapes disease dynamics. We find that human mobility is a consistent driver of dengue outbreaks over the five-year period studied. We also find that human movement is a more important driver of transmission dynamics between neighbourhoods than transmission within them, both within and across dengue seasons, and that spatially, the relative importance of different neighbourhoods to transmission elsewhere is relatively constant over time. Our framework highlights the time- and space-varying roles of human mobility and is applicable to outbreaks of other infectious diseases and to other questions of relative epidemic drivers.