Large-scale agent-based epidemiological modeling

The collective behavior that results from large numbers of atoms interacting in a material (mainly each atom only with its immediate neighbors for metals) determines that material's response to impulsive loading, such as a shock wave passing through a solid. Similarly, a pandemic can spread throughout a country or even worldwide as the result of a series of individual human-to-human contacts. By combining a stochastic agent-based model of disease spread among individuals at the local community level with detailed U.S. Census and Department of Transportation data on population demographics and mobility, we have extended our “SPaSM” (Scalable Parallel Short-range Molecular dynamics) code into a powerful epidemiological modeling tool for studying the spatiotemporal dynamics of regional to national-scale outbreaks. This simulation model, developed in the early 2000s, has been used to assess potential mitigation strategies – school closures, vaccination or antiviral prophylaxis campaigns, restricted air or highway travel, quarantines, ..., and various combinations thereof – in the event of an influenza pandemic in the United States. The arrival of COVID-19 presented additional challenges, in turning a simulation platform designed for “what-if” scenario exploration into one used for real-time response to an emerging pandemic outbreak.