Urban settings are central to maintaining dengue virus transmission. Their pronounced spatial heterogeneity challenges our understanding of transmission dynamics. In particular, urban human density remains a neglected axis despite two main potential modes of influence on this vector-borne infection. On the one hand, because humans are the main producers of breeding sites for the mosquito vector Aedes aegypti, their spatial distribution can affect dengue suitability at fine scales and emerging temporal patterns of incidence at coarser levels. On the other hand, because human density is related to human movement, it should affect the spatial spread of dengue within a city. We focus specifically on the arrival of an infected individual to a spatial unit, which triggers local transmission and we call here a 'spark'. We examine the dependence of the spark rate on both local human density and global incidence, by analyzing case data from surveillance in Rio de Janeiro at a resolution of 250m x 250m over a period of five years. The ability of this function to explain the differential timing of local transmission initiation is then assessed by building a stochastic metapopulation model coupled through global incidence.
Results/Conclusions
We show that the spark rate per unit is strongly determined by local human density and the total number of cases in the city. Numerical simulations of transmission dynamics that resolve space finely and incorporate the global temporal incidence of the city as a covariate, show that the proposed spark function explains the delay in the arrival time of infections observed in the least dense units. In addition, the aggregation of spatial units according to human density produce clear, distinct temporal patterns of incidence. In contrast, standard aggregation based on contiguous space, such as neighborhood of belonging, fail to exhibit any consistent pattern. The proposed parameterization of the spark rate provides an effective approach to incorporate fine-scale density effects, as well as spatial coupling without its explicit representation. A novel metapopulation model that is based on this notion is discussed.