Host behaviour driven by awareness of infection risk amplifies the chance of superspreading events

We demonstrate that heterogeneity in the perceived risks associated with infection within host populations amplifies the chances of superspreading during the crucial early stages of an epidemic. Under this behavioural model, individuals less concerned about the dangers from infection are more likely to be infected and attend larger-sized (riskier) events. For directly transmitted diseases such as COVID-19, this leads to infections being introduced at rates above the population prevalence to the events most conducive to superspreading. We develop an interpretable computational framework for evaluating within-event risks and derive a small-scale reproduction number measuring how the infections generated at an event depend on transmission heterogeneities and the number of introductions. This quantifies how event-scale patterns relate to population-level characteristics and generalises previous frameworks. As event duration and size grow, our reproduction number converges to the basic reproduction number. We illustrate that even moderate levels of heterogeneity in the perceived risks from infection substantially increase the likelihood of disproportionately large clusters of infections occurring at larger events, despite fixed overall disease prevalence. We show why collecting data linking host behaviour and event attendance is essential for accurately assessing the risk posed by an invading pathogen in the emerging stages of an outbreak.