RAPID: Real-time Forecasting Models for Hospitalizations of Infectious Disease in the USA

2019冠状病毒病表明，传染病暴发可能在多大程度上给卫生保健系统和整个社会造成沉重负担。为了在疾病高传播时期支持规划和决策工作，必须了解预期的疾病负担和传播模式。该项目将为美国开发住院预测模型，利用新的、高分辨率的公开数据集，即废水和基因组监测数据，以及更传统的流行病学、流动性、人口、社会经济和行为数据。这些模型将被设计用于准确评估美国城市当地医疗保健系统的预期负担，补充目前存在的州和国家层面的建模框架。一群不同的学生将领导模型开发，并通过COVID-19预测中心和流感预警中心向疾病预防控制中心传播结果，进一步扩大已建立的学术与政府伙伴关系。可公开获取的提交材料和编制的综合预测将有助于增进社会对传染病风险的普遍了解，并有助于提高公众的科学翻译和素养。住院预测模型将利用机械建模和统计数据驱动的方法，将不同的数据输入结合到有意义的预测框架中。这项工作将包括开发新的建模技术，以进一步提高预测能力。模型的高分辨率，即社区和城市级别，将填补目前以州和国家级别预测为主的文献和实践空白。高度本地化、更具可操作性的空间尺度将增加我们的模型在实践中的效用，并为地方官员提供一种机制，以区分不同人群的危害，从而为决策者提供更公平和公正的政策指导。明确考虑问题上下文的新评估度量的开发将进一步增加我们模型的效用，并为更广泛的建模社区提供一组新的性能工具。从长远来看，我们对这项研究工作的系统工程方法将有助于建立一套强大的、经过审查的工具，这些工具可用于在呼吸道病毒疾病的季节性周期和大流行期间预测一系列变量。该奖项反映了美国国家科学基金会的法定使命，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。

COVID-19 brought to light the extent to which infectious disease outbreaks can result in a significant burden on the healthcare system and societies in general. In order to support planning and decision-making efforts during periods of high disease transmission, expected disease burden and transmission patterns must be understood. This project will develop hospitalization forecasting models for the United States that exploit novel, high resolution publicly available data sets, namely waste water and genomic surveillance data, alongside more traditional epidemiological, mobility, demographic, socioeconomic, and behavioral data. These models will be designed to accurately assess the expected burden on local healthcare systems for cities in the United States, complementing the state and national level modeling frameworks that currently exist. A diverse group of students will lead the model development and dissemination of the results to the CDC through the COVID-19 Forecast Hub for COVID-19 and FluSight for Influenza, further expanding upon the established academic-government partnership. The publicly accessible submissions and ensemble forecast produced will serve to both enhance societies general understanding of infectious disease risk, and help improve science translation and literacy among the general public.The hospitalization forecasting models will utilize both mechanistic modeling and statistical data-driven approaches that combine disparate data inputs into meaningful predictive frameworks. This work will include the development of novel modeling techniques to further improve predictive capabilities. The high resolution, i.e., community and city-level, nature of the models will fill a gap in both the literature and practice, which to-date is dominate by state and national level forecasts. The highly local, more actionable spatial scales will both increase the utility of our models in practice, and provide a mechanism for local officials to distinguish harm across population groups, enabling more fair and equitable policy guidance for decision makers. The development of novel evaluation metrics that explicitly consider problem context will further increase the utility of our models, and offer a new set of performance tools to the broader modeling community. In the long term, our systems engineering approach to this research effort will contribute to the establishment of a robust, vetted set of tools that can be used for forecasting across a range of variables, during both seasonal cycles of respiratory viral disease and pandemic periods.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.