空气污染，尤其细颗粒物污染危害人体健康,已被国内外大量环境流行病学研究证实。时间序列数据已被广泛应用于大气污染急性暴露对各种健康结局的研究，而半参数广义可加模型已成为标准统计方法。这些研究采用的健康结局指标和统计学模型主要针对单个污染物对单个健康结局或多个健康结局的并集的评价。而多维健康结局序列相互之间不独立，且各健康结局的混杂因素也可能并不相同，我们提出作为联立方程组模型特例的似乎不相关-广义可加模型，较之于各方程独立求解有更高的估计效率。并且利用半参数广义可加模型可灵活地处理各因素的线性及复杂的非线性影响，建立污染物与健康指标之间的浓度-反应曲线，探查临界效应，这在高浓度复合型污染的太原市尤其有意义。研究结果为全面高效的评价和预测健康效应提供理论基础和科学方法，并为环境空气质量标准修订提供新的科学依据。

Air pollution, especially fine particulate matter pollution harms human body health, which has been confirmed by a large number of environmental epidemiological studies at home and abroad. Time series data have been widely applied in the atmospheric pollution of acute exposure to a variety of health outcomes research, among which semiparametric Possion generalized additive model has become the standard research method for atmospheric environmental epidemiology, and gratifying results were achieved. But the health outcome indicators and statistical models used in these studies are only for evaluating single pollutant and single health outcome or simply combining the results. However multidimensional health outcomes sequence is not independent with each other, and the confounding factors of health outcomes also may not be the same. In order to estimate higher efficiently, the seemingly unrelated-generalized additive models are more suitable as a special case of the simultaneous equations model. Semiparametric generalized additive models can be more flexible in dealing with linear factors and complex nonlinear effects. So we can establish the concentration response curves between pollutants and health indicators and probe the critical effect. It is particularly significant in high concentration compound pollution of taiyuan city. The prospective outcome of this project will provide theoretical basis and scientific methods for the comprehensive and efficient evaluation health effects, and new scientific clues for ambient air quality will alse be given.

时间序列数据已被广泛应用于大气污染急性暴露对各种健康结局的研究，而半参数广义可加模型已成为标准统计方法。本研究收集了石家庄市居民日死因别资料和三所医院（河北省儿童医院、石家庄市第一医院、河北医科大学第一医院）日门诊量数据，以及空气污染物及气象资料的每日浓度资料,并确定了数据清洗原则，采用EM+spline模型，运用R 3.3.2软件中的mtsdi软件包对缺失值进行填补。建立基于泊松分布的广义可加模型，以污染物的影响作为模型中的线性参数部分或参数样条函数引入模型，同时考虑长期趋势、季节性、日历效应、滞后和累积效应的影响等，把气象条件等混杂因素作为非参数部分，并进一步构建似乎不相关半参数模型，定量评价多种污染物对人群健康结局的影响。本研究进行了石家庄市大气污染物浓度对缺血性心脏病死亡、脑血管病死亡等健康效应的危险度评价，以及石家庄市大气污染物浓度与儿童、成人呼吸系统等疾病的门诊量暴露反应关系危险度评价。为全面高效的评价和预测健康效应提供理论基础和科学方法，并利用广义可加模型可以图示复杂的非线性效应的优势，根据最终构建的浓度反应曲线确定临界效应，为环境空气质量标准修订提供新的科学依据。