? ? Although several studies suggest that air pollution exposure may be associated with adverse birth outcomes, it is challenging to assess pollution exposure during pregnancy. There are numerous factors that need to be taken into account when modeling the relationship between air pollution and any health outcome, including model choice, selection of appropriate covariates as well as the method used to measure exposure. This project attempts to better understand the relationship between maternal exposure to air pollution and birth outcomes by using statistical models that incorporate both spatial modeling techniques and methods for evaluating the associated measurement error. We believe that understanding and addressing these environmental health issues on this vulnerable subgroup of pregnant women is of critical importance to the overall health of the nation. The goals of this project include: 1. Quantifying the uncertainty related to measuring local air pollution exposure, based upon monitoring station data, in order to understand the associated measurement error. 2. Determining how ambient exposure measures from monitoring stations connect to pregnancy outcomes in order to understand how to incorporate the different estimates of exposure and determine how these estimates affect the exposure-response relationship. 3. Using hierarchical modeling techniques to model the relationship between adverse pregnancy outcomes and maternal exposure to air pollution while accounting for the effects of measurement error. The statistical analyses presented in this project will utilize spatial techniques and hierarchical modeling from a Bayesian perspective. We propose model fitting within a Bayesian framework as spatial modeling of multilevel specifications are handled naturally through Bayesian modeling. In the Bayesian context we avoid the need to rely on possibly inappropriate asymptotic inference. The first stage of our hierarchical model will focus on the measurement error associated with measuring exposure. The second stage measures long-term air pollution exposure using observed measurements from monitoring stations. The final stage of the model measures the exposure-response relationship while introducing personal covariate information into the model. Project Summary Although several studies suggest that air pollution exposure may be associated with adverse birth outcomes, it is challenging to assess pollution exposure during pregnancy. There are numerous factors that need to be taken into account when modeling the relationship between air pollution and any health outcome, including model choice, selection of appropriate covariates as well as the method used to measure exposure. This project attempts to better understand the relationship between maternal exposure to air pollution and birth outcomes by using statistical models that incorporate both spatial modeling techniques and methods for evaluating the associated measurement error. We believe that understanding and addressing these environmental health issues on this vulnerable subgroup of pregnant women is of critical importance to the overall health of the nation. The goals of this project include: 1. Quantifying the uncertainty related to measuring local air pollution exposure, based upon monitoring station data, in order to understand the associated measurement error. 2. Determining how ambient exposure measures from monitoring stations connect to pregnancy outcomes in order to understand how to incorporate the different estimates of exposure and determine how these estimates affect the exposure-response relationship. 3. Using hierarchical modeling techniques to model the relationship between adverse pregnancy outcomes and maternal exposure to air pollution while accounting for the effects of measurement error. Research Design and Methods The statistical analyses presented in this project will utilize spatial techniques and hierarchical modeling from a Bayesian perspective. We propose model fitting within a Bayesian framework as spatial modeling of multilevel specifications are handled naturally through Bayesian modeling. In the Bayesian context we avoid the need to rely on possibly inappropriate asymptotic inference. The first stage of our hierarchical model will focus on the measurement error associated with measuring exposure. The second stage measures long-term air pollution exposure using observed measurements from monitoring stations. The final stage of the model measures the exposure-response relationship while introducing personal covariate information into the model. ? ? ?
Different exposure measurement and modeling techniques have been used in air pollution health effects studies including proximity models, interpolation models, land use regression models and dispersion models. These approaches have some limitations which can lead to misclassification of exposure measurements. With proper modeling techniques, we hope to reduce some of the uncertainties associated with exposure-response modeling in order for public health regulators to make better policy decisions on vulnerable subpopulations like pregnant women.
| Gray, Simone C; Gelfand, Alan E; Miranda, Marie Lynn (2011) Hierarchical spatial modeling of uncertainty in air pollution and birth weight study. Stat Med 30:2187-98 |
