Travel restrictions have often been used as a measure to combat the spread of disease -- in particular, they have been extensively applied in 2020 against coronavirus disease 2019 (COVID-19). How to best restrict travel, however, is unclear. Most studies and policies simply constrain the distance r individuals may travel from their home or neighbourhood. However, the epidemic risk is related not only to distance travelled, but also to frequency of contacts, which is proxied by the frequency f with which individuals revisit locations over a given reference period. Inspired by recent literature that uncovers a clear universality pattern on how r and f interact in routine human mobility, this paper addresses the following question: does this universal relation between r and f carry over to epidemic spreading, so that the risk associated with human movement can be modeled by a single, unifying variable r * f? To answer this question, we use two large-scale datasets of individual human mobility to simulate disease spread. Results show that a universal relation between r and f indeed exists in the context of epidemic spread: in both of the datasets, the final size and the spatial distribution of the infected population depends on the product r * f more directly than on the individual values of r and f. The important implication here is that restricting r (where you can go), but not f (how frequently), could be unproductive: high frequency trips to nearby locations can be as dangerous for disease spread as low frequency trips to distant locations. This counter-intuitive discovery could explain the modest effectiveness of distance-based travel restrictions and could inform future policies on COVID-19 and other epidemics.
Download