RAPID: Planning for the present and future management of COVID-19

COVID-19的出现和快速传播在几个月内改变了我们的世界。在一年的时间里，与新冠病毒有关的疾病夺去了地球仪各地200多万人的生命，在美国夺去了40多万人的生命。疫情暴露了我们医疗系统的关键漏洞，并在设备，疫苗，测试和其他资源的分配方面带来了前所未有的物流挑战。此外，决策者发现自己在非治疗性干预措施的时机和严重程度方面面临困难的选择，这些措施减缓了疾病的传播，但也对经济和下游健康产生了负面影响。最优自适应决策策略使用累积数据来告知资源应如何在空间和时间上分配，以在控制成本的同时最大化累积健康。在这种情况下，最佳决策策略有可能挽救成千上万人的生命，更有效地利用资源，并减少负面的经济影响。该研究项目开发了使用异质数据流（包括人类流动数据，空中交通以及疾病传播和疫苗传播信息）估计最佳决策策略的新方法。该项目还为学生和一名博士后学者提供培训机会。该研究项目通过将行为变化纳入随机房室模型来推进统计和数学疾病模型。多个数据流的整合，包括人口流动性，人口普查，发病率和检测报告，疫苗接种传播和航空旅行允许更丰富和更现实的模型比那些适合使用单一的数据模式。此外，与人口普查数据相联系的流动性数据将产生关于异质性的新知识，以应对各种干预措施，如在人口统计中的就地避难所。这些模型被集成到新的时空强化学习方法中，用于估计最佳决策策略，该策略建议如何在多个分辨率上分配资源和非治疗性干预（例如，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 emergence and rapid spread of COVID-19 has transformed our world in a matter of months. In the course of a year, COVID-related illness has claimed the lives of more than 2 million people across the globe and more than 400,000 people in the United States. The outbreak has exposed critical vulnerabilities in our healthcare system and created unprecedented logistical challenges in the distribution of equipment, vaccines, tests, and other resources. Furthermore, policymakers have found themselves facing difficult choices regarding the timing and severity of non-therapeutic interventions which slow the spread of disease but also incur negative economic and downstream health impacts. An optimal adaptive decision strategy uses accumulating data to inform how resources should be allocated over space and time to maximize cumulative health while controlling cost. In this context, an optimal decision strategy has the potential to save tens of thousands of lives, use resources more efficiently, and reduce negative economic impacts. This research project develops new methodologies for estimation of optimal decision strategies using heterogeneous data streams including human mobility data, air traffic, as well as information on disease spread and vaccine dissemination. This project also provides training opportunities for students and a postdoctoral scholar. This research project advances statistical and mathematical disease models by including behavior-change into stochastic compartmental models. The integration of multiple data streams including human-mobility, the census, incidence and testing reports, vaccination dissemination, and air-travel allows for richer and more realistic models than those fit using a single data modality. Furthermore, mobility data linked with census data will generate new knowledge about heterogeneity in response to interventions such as shelter-in-place across demographics. These models are integrated into novel spatio-temporal reinforcement learning methods for estimation of an optimal decision strategy that recommends how resources and non-therapeutic interventions should be allocated across multiple resolutions (e.g., locally and nationally) to minimize disease spread subject to constraints on economic and other downstream impacts.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.