Humans vs. COVID-19: Data-driven Modeling of Disease Spread on College Campuses

Mathematical modeling in systems in which we have census population data on disease in close-knit communities can help us understand the spread, peak of infections, and key drivers. We consider two such instances of census data in the setting of a small residential liberal arts campus. In the first study, we use data from a game of tag called Humans vs. Zombies, played on a small liberal arts campus. In this study, we develop five disease models and utilize three parameter fitting techniques to find the optimal parameter set for each model. We find that an SIR model with added multiple susceptibility classes and a sleep cycles modification proves to be the best fit, showing that the immune resistance and daily routines of individuals are two essential elements in disease modeling. We conclude that it is important to incorporate human behaviors, such as sleep and highly protective behaviors to ward off disease, into disease models, particularly in these settings. We then implemented these insights in modeling the spread of COVID-19 on small liberal arts college campuses in New England. This shows us that data from games and simulations can be a powerful tool for drawing insights about modeling the spread of real-life diseases