People Meet People - A Microlevel Approach to Predicting the Effect of Policies on the Spread of COVID-19

Governments worldwide have been adopting diverse and nuanced policy measures to contain the spread of Covid-19. However, epidemiological models usually lack the detailed representation of human meeting patterns to credibly predict the effects such policies. We propose a novel simulation-based model to address these shortcomings. We build on state-of-the-art agent-based simulation models, greatly increasing the amount of detail and realism with which contacts take place. Firstly, we allow for different contact types (such as work, school, households or leisure), distinguish recurrent and non-recurrent contacts and allow the infectiousness of meetings to vary between contact types. Secondly, we allow agents to seek tests and react to information, such as experiencing symptoms, receiving a positive test or a known case among their contacts, by reducing their own contacts. This allows us to model the effects of a wide array very targeted policies such as split classes, mandatory work from home schemes or test-and-trace policies. To validate our model, we show that it can predict the effect of the German November lockdown even if no similar policy has been observed during the time period that was used to estimate the model parameters.