Quantifying Model Uncertainty for Forecasting the Spread of Infectious Diseases

描述(由申请人提供)：我们建议的主要目标是通过量化基于代理的模型不确定性和行为改变对传染病传播的影响来增强国家的公共卫生应对能力。我们的目标是提高对突发行为对疾病传播预测模型的准确性和适用性的影响的理解。我们将在数学模型公式中评估人类和人群对大流行的行为反应的不确定性的影响。这一基础性认识将有助于改进所有现有的流行病学模型，从而加强公共卫生从业人员和政策制定者有效管理迅速蔓延的流行病的能力，无论使用何种工具。我们将利用现有的流行病学和行为模拟基础设施来开发新的数学方法，以单独纳入不同类型的疾病和行为变化，并与其他干预策略相结合。我们将验证这些模型，并量化计算模型对参数和已知疾病传播模式的敏感性。这些模型将用于估计流行率和发病率，并将使我们能够系统地比较预防措施的相对效果，如行为变化、隔离、接触者追踪、隔离和接种疫苗。首先，我们将开发新的方法来表征紧急行为，并扩展数学基础和软件基础设施，以模拟应对流行病的行为变化。其次，我们将量化由行为反应的分布引起的疫情进展的不确定性。这些行为模型将在现有的基于高保真代理的活动模拟器模型中实现和验证。最后，我们将传播这些进展，以便它们可以在其他流行病学模拟中有用和使用。 相关性：持续流行的数据稀少、不准确，而且往往无法获得。因此，量化参数和计算不确定性对于预测疾病传播的影响至关重要。我们不能假设不确定参数的影响可以忽略不计；特别是在基于模型的决策将影响无数人的生活的情况下。当前模型的根本局限性之一是，它们在多大程度上捕捉到了人类行为的变化，以应对持续的地方病。

DESCRIPTION (provided by applicant): The primary goal of our proposal is to enhance the nation's public health response capability by quantifying agent-based model uncertainty and the impact of behavioral modification on the spread of infectious diseases. Our goal is to improve the understanding of the impact of emergent behavior on the accuracy and applicability of predictive models of disease spread. We will evaluate the implications of uncertainty in human and population behavioral response to a pandemic in mathematical model formulations. This foundational understanding will help improve all existing epidemiological models thereby potentiating the ability of public health practitioners and policy-makers to effectively manage a burgeoning epidemic regardless of the tool being used. We will leverage existing epidemiologic and behavioral simulation infrastructure to develop new mathematical approaches to incorporate different types of diseases and behavioral changes alone and in combination with other intervention strategies. We will validate the models and quantify sensitivity of computational models to parameters, and known disease spread patterns. These models will be constructed for use in estimating prevalence and incidence and will allow us to compare systematically the relative effects of preventive measures, such as behavioral changes, isolation, contact tracing, quarantine, and vaccination. First, we will develop novel approaches to characterize emergent behavior and extend the mathematical foundation and software infrastructure for modeling behavior changes in response to an epidemic. Secondly, we will quantify the epidemic progression uncertainty caused by the distribution of behavioral responses. These behavioral models will be implemented and validated in an existing high-fidelity agent-based activity simulator model. Finally, we will disseminate these advances so they can be useful, and used, in other epidemiological simulations. RELEVANCE: The data for an ongoing epidemic is sparse, inexact, and often just unavailable. Therefore, quantifying parameter and computational uncertainties is crucial for forecasting the impact of disease spread. We cannot assume impact of the uncertain parameters is negligible; especially when decisions based on the model will impact the lives of countless people. One of the fundamental limitations of the current models is in how well they capture changes in human behavior in response to an ongoing endemic.