An agent-based model approach to computational epidemiology

Abstract: The aim of this work was to compare different Non-pharmaceutical interventions (NPIs) in a simulated environment modeling Sweden with 10 000 000 (ten million) agents. Simulating the performance of these NPIs can be beneficial to policymakers and other government officials in order to give the power to make the right decision in the face of an epidemic. In this work, an agent-based model (ABM) was used, rather than estimating an ordinary differential equation, which is often done in the field of computational epidemiology. An existing ABM was used to perform this task. We found that using various interventions could greatly decrease the number of infected agents as well as fatalities at the end of the simulations. Five different types of interventions were simulated, these were; baseline, lockdown, home quarantine, case isolation and mask usage. The results from the interventions were compared against a baseline of no intervention. The most effective intervention that was simulated by far was total lockdown, reducing infected agents by 99.4% and fatalities by 99.6% from baseline. In second place, home quarantine reduced the number of infected agents by 38.0% and fatalities by 45.1%. We found that the reduction in infected agents was much greater in the simulated environment than in the real world, when compared with the results from another study on humans rather than simulated agents. There was a certain similarity between how well the methods ranked in effectiveness in the simulation and in reality. However, they were not equal. The results and methods of this work are probably not suitable as a proxy for estimating real-world effectiveness of NPIs, but may be used to compare the different NPIs against each other.