Abstract
BACKGROUND: Despite decades of epidemiological theory making relatively simple assumptions about host movements, it is increasingly clear that non-random movements drastically affect disease transmission. To better predict transmission risk, theory is needed that quantifies the contributions of both fine-scale host space use and non-independent, correlated host movements to epidemiological dynamics. METHODS: We developed and applied new theory that quantifies relative contributions of fine-scale space use and non-independent host movements to spatio-temporal transmission risk. Our theory decomposes pairwise spatio-temporal transmission risk into two components: (i) spatial overlap of hosts-a classic metric of spatial transmission risk - and (ii) pairwise correlations in space use - a component of transmission risk that is almost universally ignored. Using analytical results, simulations, and empirical movement data, we ask: under what ecological and epidemiological conditions do non-independent movements substantially alter spatio-temporal transmission risk compared to spatial overlap? RESULTS: Using theory and simulation, we found that for directly transmitted pathogens even weak pairwise correlations in space use among hosts can increase contact and transmission risk by orders of magnitude compared to independent host movements. In contrast, non-independent movements had reduced importance for transmission risk for indirectly transmitted pathogens. Furthermore, we found that if the scale of pathogen transmission is smaller than the scale where host social decisions occur, host movements can be highly correlated but this correlation matters little for transmission. We applied our theory to GPS movement data from white-tailed deer (Odocoileus virginianus). Our approach predicted highly seasonally varying contributions of the spatial and social drivers of transmission risk - with social interactions augmenting transmission risk between hosts by greater than a factor of 10 in some cases, despite similar degrees of spatial overlap. Moreover, social interactions could lead to a distinct shift in the predicted locations of transmission hotspots, compared to joint space use. CONCLUSIONS: Our theory provides clear expectations for when non-independent movements alter spatio-temporal transmission risk, showing that correlated movements can reshape epidemiological landscapes, creating transmission hotspots whose magnitude and location are not necessarily predictable from spatial overlap.