How robust are the estimated effects of air pollution on health? Accounting for model uncertainty using Bayesian model averaging.
How robust are the estimated effects of air pollution on health? Accounting for model uncertainty using Bayesian model averaging.
复制标题
DOI：
10.1016/j.sste.2016.04.001
复制
发表时间：
2016-08
影响因子：
3.4
通讯作者：
Leyland AH
中科院分区：
文献类型：
--
作者：
Pannullo F;Lee D;Waclawski E;Leyland AH
我们探讨了污染健康效应对三个因素的敏感性。NO2的估计，剥夺的选择和空间自相关模型的选择。这些因素的选择导致污染健康影响的广泛差异。BMA用于估计整体影响，同时考虑模型的不确定性。总的来说，污染对健康的影响是积极的，但处于边缘。空气污染对人类健康的长期影响可以通过小面积生态研究来估计，在这些研究中，健康结果与空气污染浓度和其他协变量（如社会经济贫困）进行回归。社会经济贫困是多因素的，难以衡量，包括收入、教育和住房等方面。然而，这些变量可能是高度相关的，这意味着人们可以创建一个整体的剥夺指数，或使用个人特征，这可能会导致各种污染健康影响。模型选择的其他方面可能会影响污染健康估计，如污染估计，空间自相关模型。因此，我们提出了一个贝叶斯模型平均方法，联合收割机从多个统计模型的结果，以产生一个更强大的代表性的整体污染健康影响。我们调查了2006年至2012年期间苏格兰中西部二氧化氮浓度与心肺死亡率之间的关系。
We explored the sensitivity of the pollution-health effect to three factors. Estimation of NO2, choice of deprivation and choice of spatial autocorrelation model. Choice of these factors leads to a wide variation in pollution-health effects. BMA is utilised to estimate an overall effect while accounting for model uncertainty. Overall, a positive but borderline pollution-health effect was obtained. The long-term impact of air pollution on human health can be estimated from small-area ecological studies in which the health outcome is regressed against air pollution concentrations and other covariates, such as socio-economic deprivation. Socio-economic deprivation is multi-factorial and difficult to measure, and includes aspects of income, education, and housing as well as others. However, these variables are potentially highly correlated, meaning one can either create an overall deprivation index, or use the individual characteristics, which can result in a variety of pollution-health effects. Other aspects of model choice may affect the pollution-health estimate, such as the estimation of pollution, and spatial autocorrelation model. Therefore, we propose a Bayesian model averaging approach to combine the results from multiple statistical models to produce a more robust representation of the overall pollution-health effect. We investigate the relationship between nitrogen dioxide concentrations and cardio-respiratory mortality in West Central Scotland between 2006 and 2012.