Modeling Health System Infectious Disease Data

及早发现新出现的病原体引起的生物恐怖主义和传染病暴发是非常重要的 对公众健康很重要，以便迅速实施控制措施。通常会收集到， 自动化保健服务数据，包括微生物学实验室检测、门诊护理和紧急情况 科室就诊、住院、诊断测试和处方药数据可能非常重要 对疾病暴发检测很有用。然而，数学、计算和统计模型是 需要在受监测的结果已经存在的情况下实施这种系统 不需要采取行动的基线水平。例如，如果适当的信号检测方法是 可能会通过识别一个 因咳嗽和发烧而寻求治疗的病人数量异常多。 在这个项目中，我们将开发早期检测传染病爆发的模型，并为 在检测到疫情后对其进行监控。这包括(I)描述自然时间的模型 以及使用感兴趣的保健服务的人数的地理差异，以便调整 例如季节性和星期几效应以及(Ii)不同时空像差检测模型 在疫情发生时发出信号。这些模型将在不同的地理位置应用 规模，从单个医院的个别病房到整个国家，以及不同的数据特殊性 从非常普遍的症状，如发烧到特定的微生物疾病菌株和抗菌素耐药性 从一个细菌物种迁移到另一个细菌物种的图谱。我们将开发和测试我们的新方法和 两个医疗计划中的模式(马萨诸塞州的哈佛朝圣者医疗保健计划和北方凯撒永久医疗计划 覆盖400多万人的加州)，一家美国大型转诊医院(布里格姆妇女医院)，一家 全州(马萨诸塞州)MRSA注册，以及由55家医院组成的国家(阿根廷)联盟 监测抗生素耐药性。模型和方法将使用以下两个历史数据进行评估 这些卫生系统和模拟数据基于不同的传染病传播动态 模特们。

Early detection- of bioterrorism and infectious disease outbreaks caused by emerging pathogens is very important for public health, to allow prompt implementation of control measures. Routinely collected, automated health services data, including microbiology laboratory tests, ambulatory care and emergency department visits, hospitalizations, diagnostic tests, and prescription drug data could potentially be very useful for disease outbreak detection. However, mathematical, computational and statistical models are needed to implement such systems whenever the outcomes under surveillance already exist at some baseline level that does not require action. For example, if appropriate signal detection methods were available, identification of an anthrax bioterrorism attack might be accelerated through recognition of an unusual number of patients seeking care for cough and fever.; In this project, we will develop models for the early detection of infectious disease outbreaks and for monitoring an outbreak after it has been detected. This includes (i) models describing the natural temporal and geographical variation in the number of people utilizing the health services of interest, in order to adjust for e.g. seasonal and day-of-week effects and (ii) different space-time aberration detection models that will generate a signal when an outbreak have occurred. These models will be applied at different geographical scales, from individual wards of a single hospital to a whole country, as well as for different data specificity from very general symptoms such as fever to specific microbial disease strains and antimicrobial resistance profiles that migrate from one bacterial species to another. We will develop and test our new methods and models in two health plans (Harvard Pilgrim Health Care in Massachusetts, and Kaiser Permanente Northern California) that cover over 4 million people, a single large US referral hospital (Brigham and Women's), a statewide (Massachusetts) registry of MRSA, and a national (Argentine) consortium of 55 hospitals that monitors antibiotic resistance. The models and methods will be evaluated using both historical data from these health systems and simulated data based on different infectious disease transmission dynamics models.