RAPID: Modeling the Severity and Transmissibility of COVID-19 in the USA with Intrinsic Behavior Change

随着新冠肺炎在世界各地的社区中传播，特别是在美国，许多问题仍然没有答案。特别重要的是，不同的缓解和遏制战略在多大程度上影响由此产生的重症监护病房病例和/或死亡人数？随着社区开始不同程度地放松目前的“封锁”命令，这个问题将变得更加微妙。此外，天气的时空变化(温度、降水和湿度)以及紫外线辐射在多大程度上调节了疾病的演变？通过独特的数据收集、模型改进和科学调查相结合，这项研究可以为这些问题提供有价值的见解。代码和派生数据将通过GitHub存储库、CRAN包和门户网站提供给科学界，并将为潜在感兴趣的利益相关者提供非正式培训，如县公共卫生部门、疾控中心和国防部机构。此次调查将使用现有的最先进的建模和预测框架，相互作用社区流行病动力学(DICE)，以研究新冠肺炎动态的人文生态。DICE是一种独特的工具，可以帮助揭示不同遏制和非药物缓解策略以及气候强迫对新冠肺炎传播的影响。独一无二的是，它是一个任意规模的混合空间集合种群模型，在该模型中，单个社区经历了确定性的疾病动力学，但在该模型中，一个社区在另一个社区播种疫情的过程是随机的。骰子可以在县、州、地区或国家级别运行，或者这些子单元的各种组合可以结合在一起，这取决于可用的数据。骰子解决SE1…系统Eni1…ImRX方程产生模拟的发病概况和作为时间、R(T)、暴发严重程度的函数的再现次数的估计，以及量化干预效果的参数。DICE已经具备了整合学校假期数据的能力，并使用了来自NASA和NOAA的气候数据，特别是比湿度，这已被证明在预测流感的演变方面非常重要。最近使用一个补充的单一人口原型工具(草案)对一系列纳入干预措施的方法进行了测试和探索，这些方法包括学校关闭、社会距离和原地安置命令，该工具是专门为快速探索可纳入骰子的改进而开发的。DICE既可以模拟未来可能发生的情况，也可以根据现有数据来估计不同干预方案的效果，还可以捕获严重程度(SEV)和传播率(R(T))的联合估计。随着新冠肺炎在美国的传播，可以在R-Sev空间对社区传播进行评估，这将提供关键和战略信息，帮助政策制定者做出更明智的决策。通过使用蒙特卡洛马尔可夫链(MCMC)方法，DICE对预测中的不确定性进行了稳健的估计。此外，DICE是一种多模型算法，允许生成32个以上模型变体的预报，这不仅提供了对各种因素(例如气候)可能产生的影响的估计，而且还产生了模型实现的超级集合，这反过来又提供了对不确定性的额外估计。这一快速奖项是由环境生物学部门传染病生态学和进化计划利用冠状病毒援助、救济和经济安全(CARE)法案的资金颁发的。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

As COVID-19 spreads through communities across the world, and particularly within the USA, a number of questions remain unanswered. Of particular importance is to what extent different mitigation and containment strategies affect the resulting number of ICU cases and/or deaths? This question will become ever more nuanced as communities begin to relax current “lockdown” orders to varying degrees. Additionally, to what extent do spatial and temporal changes in weather (temperature, precipitation, and humidity) as well as UV radiation modulate the disease’s evolution? Through a combination of unique data collection, model refinement, and scientific investigation, this study can shed valuable insight on these questions. The codes and derived data will be made available to the scientific community through GitHub repositories, CRAN packages, and web portals, and informal training will be provided for potentially interested stakeholders, such as county public health departments, the CDC, and DoD agencies.This investigation will use an existing state-of-the-art modeling and forecasting framework, Dynamics of Interacting Community Epidemics (DICE), to examine the human ecology of COVID-19 dynamics. DICE is a unique tool that can help reveal the impact of different containment and non-pharmaceutical mitigation strategies, as well as climate forcing, on the transmission of COVID-19. Uniquely, it is an arbitrarily scaled hybrid spatial metapopulation model in which individual communities experience deterministic disease dynamics, but between which the process of one community seeding an outbreak in another community is stochastic. DICE can be run at the county, state, region, or national level, or, various combinations of these sub-units can be coupled, depending on what data are available. DICE solves the system of SE1…EnI1…ImRX equations producing a modeled incidence profile and estimates of the reproduction number as a function of time, R(t), the severity of the outbreak, and parameters quantifying the efficacy of interventions. DICE already has the capability of incorporating school vacation data, and uses climate data from NASA and NOAA, and specific humidity, in particular, which has been shown to be important in forecasting the evolution of influenza. A range of methodologies for incorporating interventions, such as school closures, social distancing, and shelter-in-place orders have been recently tested and explored using a complementary single-population prototype tool (DRAFT), specifically developed to rapidly explore refinements that can be incorporated into DICE. DICE can both simulate possible future scenarios as well as fit to available data to estimate the efficacy of different intervention profiles, and also captures joint estimates of severity (Sev) and transmissibility (R(t)). As COVID-19 spreads across the U.S., community transmission can be evaluated in R-Sev space, which will provide crucial and strategic information to assist policymakers in making more informed decisions. Through the use of a Monte Carlo Markov Chain (MCMC) approach, DICE produces robust estimates of the uncertainties in the projections. Additionally, DICE is a multi-model algorithm, allowing the generation of forecasts for more than 32 model variants, which provides not only an estimate of the impact that various factors may play (e.g., climate), but also produces hyper-ensembles of model realizations, which, in turn provide additional estimates of uncertainty. This RAPID award is made by the Ecology and Evolution of Infectious Disease Program in the Division of Environmental Biology, using funds from the Coronavirus Aid, Relief, and Economic Security (CARES) Act.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.