Founder effects arising from gathering dynamics systematically bias emerging pathogen surveillance

This is the authors' abstract. We don't add key points for this paper.

Models of infectious disease transmission have shown the importance of heterogeneous contact networks for epidemiology; the most connected individuals are most likely to be infected early. Yet it is cumbersome to parameterize and incorporate such networks into simple models. We introduce an alternative model framework that explicitly includes attendance at and disease transmission within gatherings of different sizes, which disaggregates sequential epidemics moving from the most to least social subpopulations that underly the overall, single-peaked infection curve. This can systematically bias initial estimates of the growth rate for emerging variants and their severity, if vulnerable populations avoid large gatherings. Finally, we show that how often similarly social individuals preferentially interact (i.e., homophily, or assortative mixing) tunes the magnitude and duration of these biases. Together, we provide a simple framework for incorporating socialization and behavior in epidemic models, which can help contextualize surveillance of emerging infectious agents.