Improving Detection of Outbreaks Due to Aerosol Attacks

拟议的研究涉及及早发现生物恐怖袭击的问题，即生物恐怖分子将生物剂以气雾剂的形式释放到大气中。此类袭击有可能导致数万至数十万人死亡。及早发现风媒攻击可以及早做出反应，反应越早，挽救的生命就越多。在过去的三年里，我们开发了一种专门为检测由风媒攻击引起的疫情而设计的算法。该算法的潜在优势包括：1.利用天气数据和对气溶胶释放引起的暴发的病例分布的时空模式的知识，以较低的误警率更早地检测到。2.在检测到暴发时，将其部分定性为生物制剂的室外气雾剂释放。3.对生物制剂释放的地点、数量和时间的最佳估计。一旦响应人员意识到事件，该信息对于管理额外的监视和响应是必不可少的。我们建议进一步开发算法，并衡量我们可以实现这些好处的程度。研究的具体目的是：1.确定贝叶斯气溶胶释放探测器(BARD)区分生物毒剂风传引起的疫情和其他传播方式引起的疫情的能力。2.将BARD在检测模拟风媒疫情方面的性能与 不是专门为检测风媒疫情而设计的算法的性能。 3.确定在以多种方式扩展其风媒暴发模型后，与BARD的推断在计算上是否仍然容易处理。4.评估在特定目标#3中进行的扩展，以相对于在特定目标#1和#2中使用的BARD基线版本在检测性能方面的改进。

The proposed research addresses the problem of early detection of bioterrorism attacks where bioterrorist release a biological agent into the atmosphere as an aerosol. Such attacks have the potential to kill tens to hundreds of thousands of individuals. Early detection of a windborne attack enables an early response and the earlier the response, the greater the number of lives saved. Over the past three years, we have developed an algorithm designed specifically to detect outbreaks caused by windborne attacks. The potential advantages of this algorithm include: 1. Earlier detection at a lower false-alarm rate as a result of utilizing weather data and knowledge of the spatial and temporal patterns in case distributions in an outbreak resulting from an aerosol release. 2. Partial characterization of the outbreak at the time of detection as an outdoor, aerosol release of a biological agent. 3. A best estimate of location, quantity, and timing of the release of the biological agent. Once responders are aware of an event, this information is essential for the management of additional surveillance and response. We propose to further develop the algorithm and measure the degree to which we can achieve these benefits.The specific aims of the research are to: 1. Determine the ability of the Bayesian Aerosol Release Detector (BARD) to discriminate between outbreaks caused by windborne transmission of biological agents and outbreaks due to other modes of transmission. 2. Compare the performance of BARD for the detection of simulated windborne outbreaks with the performance of algorithms not specifically designed to detect windborne outbreaks. 3. Determine whether inference with BARD remains computationally tractable after extending in a number of ways its models of windborne outbreaks. 4. Evaluate the extensions made in Specific Aim #3 for improvements in detection performance relative to the baseline versions of BARD used in Specific Aims #1 and #2.