RAPID: Real-time updating of an agent-based model to inform COVID-19 mitigation strategies.

SARS-CoV-2病毒是世纪以来最严重的大流行病的罪魁祸首。由于目前还没有疫苗，非药物干预措施（NPI）是控制病毒的唯一途径。这些干预措施，包括社交距离，关闭学校和庇护所，如果时机适当并广泛采用，可能是有效的。与此同时，非营利机构造成严重的社会和经济破坏，这意味着必须尽可能少地使用它们。为了就何时采用非盈利机构以及何时放松非盈利机构作出决定，必须了解这些行动可能产生的后果。这项研究将提高数学模型的能力，以深入了解这些后果。研究人员将使用统计方法来估计关键的未知数，例如以前或活跃感染的人数。展望未来，研究人员将使用美国各地社区的模拟模型来评估替代行动的后果。将这些方法结合起来将充分利用每种方法的优势，从而改善对缓解美国COVID-19大流行的替代战略后果的预测。该项目将以美国SARS-CoV-2传播的地理现实的、基于病原体的模型为特色。该模型的优点包括其对美国特定县的人口密度、人口统计和移动模式的详细描述，以及通过代理人的行为修改直接实施NPI的能力。这些特征使得能够对SARS-CoV-2的传播和NPI对其的影响进行本地定制的预测。与此同时，基于代理的模型的计算需求对在流行病的快节奏背景下使用它们提出了挑战。为了克服这一挑战，研究人员将使用计算要求较低的统计方法来估计基于代理的模型中使用的输入，直到大流行的给定点。然后，基于代理人的模型将在有关使用NPI的替代方案下，从那时起进行模拟。为了进一步防止计算需求成为产生及时结果的限制因素，研究人员将利用高性能计算资源对基于代理的模型进行批量模拟，以考虑随机性和参数不确定性。研究结果将在整个项目过程中定期公布，研究人员将与利益相关者协调，以确保考虑中的缓解方案随着大流行的发展而保持相关性。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The SARS-CoV-2 virus is responsible for the most significant pandemic in a century. With a vaccine not yet available, non-pharmaceutical interventions (NPIs) offer the only way to control the virus at this time. Those interventions, which include social distancing, school closures, and sheltering in place, may be effective if timed appropriately and adopted widely. At the same time, NPIs cause serious social and economic disruption, meaning that they must be used as sparingly as possible. To inform decisions about when to adopt NPIs, and when to relax them, it is important to understand what the consequences of those actions might be. This research will advance the capability of mathematical models to provide insight into those consequences. Looking to past data, the researchers will use statistical approaches to estimate key unknowns, such as numbers of people previously or actively infected. Looking into the future, the researchers will use simulation modeling of communities across the United States to evaluate the consequences of alternative actions. Merging these approaches will capitalize on the strengths of each, resulting in improved projections of the consequences of alternative strategies for mitigating the COVID-19 pandemic in the United States.This project will feature a geographically realistic, agent-based model of SARS-CoV-2 transmission in the United States. Advantages of this model include its detailed portrayal of the density, demography, and movement patterns of people in specific counties across the United States, and its ability to directly implement NPIs through behavior modification of agents. These features enable locally tailored projections of SARS-CoV-2 transmission and impacts of NPIs thereon. At the same time, the computational demands of agent-based models pose a challenge to using them in the fast-paced context of a pandemic. To overcome that challenge, the researchers will use less computationally demanding statistical approaches to estimate inputs for use in the agent-based model up to a given point in the pandemic. The agent-based model will then simulate forward from that time under alternative scenarios about use of NPIs. To further safeguard against computational demands being a limiting factor for producing timely results, the researchers will make use of high-performance computing resources to perform batches of simulations of the agent-based model that account for stochasticity and parameter uncertainty. Results will be publicly disseminated on a regular basis over the course of the project, and the researchers will coordinate with stakeholders to ensure that mitigation scenarios under consideration remain relevant as the pandemic progresses.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.