Linking models and policy: Using active adaptive management for optimal control o

为了控制疾病的爆发，在面临不确定性的情况下，必须作出关键决定。虽然我们可以使用 决策支持模型，任何特定流行病的关键不确定性，农业和人类 健康环境，不能先验地解决。通过监测疫情对管理的反应， 通过干预，人们可以了解模型结构和参数值，以便为决策提供信息。等 对相互竞争的模型的评价和评估通常是在事后进行的，很少用于实时评估 施政纲要而自适应管理（AM）不是专注于识别“最佳模型”，而是将 实时模型拟合，基于动态监测数据，随机优化选择最佳 管理行动，以最大限度地提高管理目标的条件下，目前的支持竞争 模型AM的根本创新是主动学习的结合， 根据行动对实现目标的内在效益及其贡献来评价行动 解决不确定性，这些不确定性限制了对当前疫情采取最佳行动的选择。虽然 以前应用于自然资源管理，AM还没有被推广用于处理 传染病动态管理。在此，我们提出一项多年努力， 使用AM进行流行病应对的基于模型的结构化决策基础设施。证明 建模和决策之间的反馈，我们建议开发一个回顾性分析， 2001年，英国爆发口蹄疫。通过每年与机构股东的互动， 研讨会，我们将制定具体的FMD模型方案，研究空间不确定性的相互作用， 美国环境中决策和口蹄疫爆发反应的动态。我们将制定方法， 软件来研究FMD的案例研究，我们将采用更普遍的调查AM的其他 面对各种空间和后勤不确定性来源， 管理利用理论模型，我们将研究应用实时监控数据来解决 空间位置、传输网络以及相互竞争的地方和全球目标的关键不确定性， 制定适应性战略，以最佳方式应对特定疫情。

To control disease outbreaks, critical decisions are necessary in the face of uncertainty. Though we can use models for decision support, key uncertainties about any specific epidemic, in both agricultural and human health settings, cannot be resolved a priori. By monitoring the response of an outbreak to management interventions, one can learn about both model structure and parameter values, to inform decisions. Such evaluation and assessment of competing models is often done in retrospect, and is rarely of use to real-time policies. Rather than focusing on identification of a "best model", Adaptive Management (AM) combines real-time model fitting, based on dynamic surveillance data, with stochastic optimization to select the best management action to maximize management objectives conditional on the current support for competing models. The fundamental innovation of AM is the incorporation of active learning, whereby management actions are evaluated based on their inherent benefit to achieving the objective, as well as their contribution to resolving uncertainties that limit the selection of the best action for the outbreak at hand. Though previously applied in natural resource management, AM has not been generalized for dealing with the management of infectious disease dynamics. Here we propose a multi-year effort to develop an infrastructure for model-based structured decision-making using AM for epidemic response. To demonstrate the feedback between modeling and decision-making, we propose to develop a retrospective analysis of the 2001 UK foot and mouth disease (FMD) epidemic. Through interactions with agency stake-holders in annual workshops, we will develop specific FMD model scenarios to study the interaction of uncertainties in spatial dynamics with decision-making and FMD outbreak response in the US setting. We will develop methods and software to study the FMD case study, which we will employ more generally to investigate AM of other livestock and human outbreaks in the face of various sources of spatial and logistical uncertainties that limit management. Using theoretical models, we will study the application of real-time surveillance data to resolve key uncertainties in spatial locations, transmission networks, and competing local and global objectives for the development of adaptive strategies that can optimally respond to specific outbreak settings.