Agent-Based Simulation for Localized COVID-19 Intervention Decision

Abstract

- Disease models have been a helpful resource which have guided health organizations in choosing appropriate interventions during the COVID-19 pandemic. However, most current models simulate disease spread on a countrywide/statewide level, lacking specificity for localities such as towns or counties. As a result, one-size-fits-all policies are being instituted for entire states despite localities being heterogeneous in many important factors (population density, age demographics, and vaccination rate). Models tailored to individual localities are necessary to facilitate local level health action. In this research, a novel agent-based disease model was created using NetLogo to simulate localized COVID-19 disease dynamics. Individual agents represent each member of a population, and their individual traits (vaccination status, age, etc.) conform to the model input (vaccination rate, age distribution, etc.). Interactions between these agents produce the model outputs, which include predicted infections and deaths. The model was validated using data from state and local health agencies for Westchester County, NY (84.2% accuracy). Using the model, this research aims to answer the following question: what local factors affect COVID-19 outbreak severity and intervention impact? To accomplish this, a sensitivity analysis was conducted for three local variables (vaccination rate, age distribution, intervention applied) and a comparison of locality simulation was conducted for four different U.S. counties. From the results attained, this research concluded that vaccination rate, age distribution, and intervention applied in a locality all contribute significantly to risk level differences between localities, and that higher risk localities are impacted harder by interventions than those with lower risk. Localities can use this model to make health related decisions, and a website (www.localcovidmodel.org) has been created for model access.