Real-time predictive modeling for public health departments to control infectious diseases

项目概要 公共卫生部门越来越多地使用预测模型来指导美国传染病控制的决策和资源分配，特别是在 COVID-19 大流行期间。这些新颖的预测模型有望更好地识别高危人群，以精确部署疫苗接种等干预措施，但关于公共卫生部门如何使用这些模型以及它们是否转化为减少传染病的政策的证据有限。我寻求解决的主要科学问题是确定预测模型是否以及在何种程度上可以纳入公共卫生实践，并由公共卫生部门转化为政策，以改善对传染病的控制。通过利用与加州公共卫生部 (CDPH) 的重要合作和丰富的流行病学数据源，我将解决一个关键的公共卫生挑战，即如何优化分配有限的资源，用于针对百日咳、季节性流感和甲型肝炎的有针对性的疫苗接种。目标是将疫苗瞄准最高风险的地点和人群，以减少疫情和感染的数量。我的假设是，公共卫生部门可以有效地将有关疫苗接种最佳目标的预测数学模型纳入其政策决策中。我将运用我在预测模型和传染病方面的专业知识，为县公共卫生部门开发开源的预测模型工具，为最高风险人群分配有针对性的疫苗接种，并研究这些模型在公共卫生中的逐步实施。我的广泛而长期的目标是评估使用预测模型来指导公共卫生部门的疫苗接种决策对公共卫生的因果影响。在目标 1 中，我将开发和验证预测模型，以最佳方式针对百日咳、季节性流感和甲型肝炎的高风险地区和人群（年龄、人口和风险因素）接种疫苗。该模型将提供各种有针对性的疫苗接种策略的比较有效性和成本，以及针对县和传染病的总体疫苗建议。在目标 2 中，我将应用实施科学中的方法来优化公共卫生官员的用户体验，以最大限度地提高基于模型的疫苗建议的可用性、沟通和采用。在目标 3 中，我将在加州公共卫生部门实施定向疫苗接种的预测模型，并在试点研究中衡量实施结果。这项工作将为未来与 CDPH 进行的创新试验奠定基础，该试验对县公共卫生部门进行随机分组，并评估使用基于模型的最佳疫苗分配预测是否可以因果关系减少病例和疫情。这项拟议的工作有可能开启新的科学方向，将预测模型转化为公共卫生的常见实践，然后可以应用于许多传染病。

PROJECT SUMMARY Public health departments increasingly use predictive modeling to guide decisions and resource allocation for the control of infectious diseases in the United States, especially during the COVID-19 pandemic. These novel predictive models offer promise to better identify high-risk populations to precisely deploy interventions such as vaccination, yet there is limited evidence on how these models are used by public health departments and whether they translate into policy that reduces infectious diseases. The major scientific problem I seek to address is to identify whether, and to what degree, predictive models can be incorporated into public health practice and translated into policy by public health departments to improve the control of infectious diseases. By leveraging a key collaboration with the California Department of Public Health (CDPH) and rich epidemiologic data sources, I will address a key public health challenge of how to optimally allocate limited resources for targeted vaccination against pertussis, seasonal influenza, and hepatitis A. The goal is to target vaccines to the highest-risk locations and populations to reduce the number of outbreaks and infections. My hypothesis is that public health departments can effectively incorporate predictive mathematical models on optimal targeting of vaccination into their policy decisions. I will apply my expertise in predictive modeling and infectious diseases to develop open-source, predictive modeling tools for county public health departments to allocate targeted vaccination to the highest-risk populations, and study the step-by-step implementation of these models in public health use. My broad, long-range goal is to evaluate the causal public health impact of using predictive models to guide decisions on vaccination in public health departments. In Aim 1, I will develop and validate predictive models to optimally target vaccines to high-risk locations and populations (age, demographic and risk factor) for pertussis, seasonal influenza, and hepatitis A. The model will provide comparative effectiveness and costs of various targeted vaccination strategies, and an overall vaccine recommendation specific to the county and infectious disease. In Aim 2, I will apply methods from implementation science to optimize the user experience for public health officials to maximize usability, communication, and uptake of model-based vaccine recommendations. In Aim 3, I will implement the predictive models of targeted vaccination in California public health departments and measure implementation outcomes in a pilot study. This work will provide the foundation for a future innovative trial with CDPH that randomizes county public health departments and evaluates whether using model-based predictions on optimal vaccine allocation can causally reduce cases and outbreaks. This proposed work has the potential to unlock new scientific directions of translating predictive models into common practice in public health, which can then be applied across many infectious diseases.