A graph-theoretic framework for integrating mobility data into mathematical epidemic models.

Advances in modeling the spread of infectious diseases have allowed modellers to relax the homogeneous mixing assumption of traditional compartmental models. The recently introduced synthetic network model, which is an SIRS type model based on a non-linear transmission rate, effectively decouples the underlying population network structure from the epidemiological parameters of disease, and has been shown to produce superior fits to multi-wave epidemics. However, inference from case counts alone is generally problematic due to the partial unidentifiability between probability of person to person transmission and the average number of contacts per individual. An alternate source of data that can inform the network alone has the potential to improve overall modeling results. Aggregate cell phone mobility data, which record daily numbers of visits to points of interest, provide a proxy for the number of contacts that people establish during their visits. In this paper, we link the contact rate from an epidemic model to the total number of contacts formed in the population. Inferring the latter from Google Community Mobility Reports data, we develop an integrated epidemic model whose transmission adapts to population mobility. This model is illustrated on the first four waves of the COVID-19 pandemic.