Spatial-Temporal Modeling of the Association between Air Pollution Exposure and Birth Outcomes.

摘要

Exposure to high levels of air pollution during the pregnancy is associated with increased probability of preterm birth (PTB), a major cause of infant morbidity and mortality. Classical statistical analyses of this association focus on pollution exposures during large intervals of the pregnancy and ignore the present spatial variability. New statistical methodology is required to specifically determine when a particular pollutant impacts the PTB outcome, to determine the role of different pollutants, and to characterize the spatial variability in these results. In Chapter 2 of this dissertation we develop a model for jointly examining the relationship between exposures to PM 2.5 and ozone and the probability of PTB with a focus on identifying the critical windows of the pregnancy in which increased exposure to these pollutants is particularly harmful. Our new Bayesian spatial model for PTB identifies susceptible windows throughout the pregnancy for multiple pollutants while allowing these windows to vary continuously across space and time. We geo-code vital record birth outcome data from Texas (2002--2004) and link them with standard pollution monitoring data and a newly introduced EPA product of calibrated air pollution model output. We apply the fully spatial model to a region of 13 counties in eastern Texas consisting of highly urban as well as rural areas. Our results indicate that while the critical windows are similar over the domain, the resulting uncertainty in the risk estimates is significantly less than when ignoring the spatial correlation. Different pollutants lead to different critical windows associated with increased probability of PTB and a proper inference procedure is introduced to correctly analyze these windows.;In Chapter 3 we introduce a spatial-temporal hierarchical multivariate probit regression model in the Bayesian setting that identifies periods of time during the first trimester of pregnancy which are particularly impactful in terms of cardiac congenital anomaly development. The model is able to simultaneously consider multiple pollutants and a multivariate cardiac anomaly grouping outcome jointly while allowing the windows to vary in a continuous manner across time and space. In the analysis we utilize numerical chemical model output data which contains information regarding multiple species of PM2.5. To our knowledge this is the first time this estimated speciated PM2.5 output has been used in the environmental health setting. Our introduction of a newly developed spatial-temporal nonparametric prior distribution for the pollution risk effect allows for greater flexibility to model the possibly nonstationary behavior exhibited by these effects. The classic stick-breaking prior is extended to the multivariate setting and to include space and time simultaneously in both the locations and the masses through the use of kernel functions, something previously not considered in the literature. Simulation study results suggest that the newly introduced prior distribution has the flexibility to outperform competitor models in a number of various setting. When applied to the Texas birth data, weeks 3, 7 and 8 of the pregnancy are identified as being impactful in terms of cardiac defect development for multiple pollutants across the spatial domain.