Statistical Methods for Complex Enivronmental Health Data

环境颗粒物（PM）空气污染是对公众健康的主要威胁，但目前制定空气质量标准的方法并没有反映PM化学混合物复杂的多污染物性质。最近的工作表明，通过针对产生最有害化学成分的PM来源，可能存在减少环境PM的公共卫生负担的机会。目前，制定新的多污染物空气质量干预策略的科学基础不足，现有的统计方法也不能充分解决数据带来的挑战。调查人员开发了广泛使用的统计方法，用于开展国家环境空气污染与健康的流行病学研究，并确定迫切需要一套新的统计方法来评估复杂空气污染物混合物对健康的影响。第一个目标是开发一个时空贝叶斯分层多变量受体模型，用于识别空气污染化学混合物的来源并估计其对人口健康结果的影响。创新的重点是(a)对污染源的健康影响进行国家综合评估；(b)使用时空模型进行源分配；(c)引入关于源概况和排放的国家数据库，为模型开发和参数估计提供信息。第二个目标是开发新的多变量时空模型，用于估计环境暴露对社区健康的影响，考虑到空间偏差和测量误差。第三个目标是将新开发的统计方法应用于一项关于空气污染和健康结果的国家研究（医疗保险队列空气污染研究）的数据，以(a)在国家、区域和地方范围内估计PM来源对人口健康的短期影响；(b)估计PM成分对健康的短期和长期影响，并确定有毒成分的来源。第四个目标是开发模块化和可扩展的开源软件，实现新的统计方法。通过在国家研究中提供关于PM成分和来源的相对毒性的关键证据，并通过开发新的统计方法来克服当前的方法挑战，该应用程序的目的将为有针对性的干预和空气质量控制战略奠定基础，这些战略将对广大人群产生重大的公共卫生影响。

DESCRIPTION (provided by applicant): Statistical Methods for Complex Environmental Health Data Project Summary Ambient particulate matter (PM) air pollution is a major threat to public health, but current approaches to setting air quality standards do not reflect the complex multi-pollutant nature of the PM chemical mixture. Recent work indicates that opportunities may exist to reduce the public health burden of ambient PM by targeting the sources of PM that produce the most harmful chemical constituents. Currently, the scientific basis for developing new multi-pollutant air quality intervention strategies is insufficient and available statistical methods do not adequately address the challenges presented by the data. The investigators have developed widely-used statistical methodology for conducting national epidemiological studies of ambient air pollution and health and have identified the critical need for a new set of statistical methods for assessing the health effects of complex air pollutant mixtures. The first aim will develop a spatial-temporal Bayesian hierarchical multivariate receptor model for identifying sources of air pollution chemical mixtures and estimating their effect on population health outcomes. Innovation focuses on (a) conducting an integrated national assessment of the health effects of pollution sources; (b) the use of spatial-temporal models for source apportionment; and (c) the introduction of national databases on source profiles and emissions to inform model development and parameter estimation. The second aim will develop novel multivariate spatial-temporal models for estimating community-level health effects of ambient environmental exposures, accounting for spatial misalignment and measurement error. The third aim will apply the newly developed statistical methods to data from a national study of air pollution and health outcomes, the Medicare Cohort Air Pollution Study, to (a) estimate short-term population health effects of PM sources on a national, regional, and local scale; (b) estimate short- and long-term health effects of PM constituents and identify the sources of toxic constituents. The fourth aim will develop modular and extensible open source software implementing new statistical methods. By providing critical evidence about the relative toxicities of PM constituents and sources in a national study and by developing novel statistical approaches to overcome current methodological challenges, the aims of this application will lay the foundation for targeted interventions and air quality control strategies that will have a substantial public health impact across broad populations. PUBLIC HEALTH RELEVANCE: Relevance Ambient particle air pollution is a major public health problem and current approaches to regulating pollutant levels are sub-optimal. This project will develop novel statistical methods to be applied to national databases for estimating the health effects of ambient particle air pollution chemical constituents and sources. The evidence generated by this work will serve as the foundation for more targeted air quality control strategies.