【摘要】 [背景]大气污染是一个重要的公共卫生问题。空气质量健康指数（AQHI）是大气污染健康风险预警和沟通的重要工具，但目前的AQHI构建大多基于单污染物模型，存在明显的局限性。[目的]建立基于大气污染物复合暴露的AQHI（J-AQHI），为进行大气污染健康风险预警和风险沟通提供科学工具。[方法]本研究从云南、广东、湖南、浙江和吉林省疾病监测点系统收集2013年1月1日至2018年12月31日的每日非意外死亡数据，包括死亡日期、年龄、性别和死因，同时分别通过中国气象数据共享服务系统和城市空气质量实时发布平台收集同期逐日气象（温度、相对湿度）及大气污染数据(SO₂、NO₂、CO、PM_2.5、PM10和8 h O3最高浓度)。首先使用Lasso回归筛选大气污染物；然后采用时间分层的病例交叉设计，将每个病例死亡日期的同一月份的同一星期几作为对照，为每个病例分配3～4个对照日；随后应用分布滞后非线性模型（DLNM）建立筛选出的大气污染物与死亡的暴露-反应关系，并进一步计算AQHI；最后利用世界卫生组织《全球空气质量指南》（AQG 2021）中的主要大气污染物指导限值，将AQHI分为四个等级，并比较单污染模型构建的AQHI和多污染物模型构建的J-AQHI的超额死亡风险（ER）。[结果]通过Lasso回归筛选出PM_2.5、SO₂、NO₂、O3共4种污染物，建立DLNM模型，发现PM_2.5、NO₂、SO₂、O3每增加1个四分位数间距，ER及其95%CI分别增加0.71%（0.34%～1.09%）、2.46%（1.78%～3.15%）、1.25%（0.9%～1.6%）和0.27%（-0.11%～0.65%）。构建的J-AQHI呈右偏态分布，将其划分为四级，分别是低风险（0～1）、中风险（2～3）、高风险（4～5）、严重风险（≥6），分别占比为11.25%、64.61%、19.33%和4.81%。对于多污染物模型构建的J-AQHI和单污染物模型构建的AQHI，污染物每增加1个四分位数间距浓度，对应的ER（95%CI）分别增加3.61%（2.93%～4.29%）和3.39%（2.68%～4.11%）。[结论]本研究基于多污染物复合暴露模型构建了J-AQHI，展示了人群实际的空气污染的暴露健康风险，为AQHI计算方法的进一步完善提供新的思路。更多还原

【Abstract】 [Background] Air pollution is a major public health concern. Air Quality Health Index（AQHI） is a very important air quality risk communication tool. However, AQHI is usually constructed by single-pollutant model, which has obvious disadvantages.[Objective] To construct an AQHI based on the joint effects of multiple air pollutants（J-AQHI）, and to provide a scientific tool for health risk warning and risk communication of air pollution.[Methods] Data on non-accidental deaths in Yunnan, Guangdong, Hunan, Zhejiang, and Jilin provinces from January 1, 2013 to December31, 2018 were obtained from the corresponding provincial disease surveillance points systems（DSPS）, including date of death, age, gender, and cause of death. Daily meteorological（temperature and relative humidity） and air pollution data(SO₂, NO₂, CO, PM_2.5, PM10, and maximum 8 h O3 concentrations) at the same period were respectively derived from China Meteorological Data Sharing Service System and National Urban Air Quality Real-time Publishing Platform. Lasso regression was first applied to select air pollutants, then a time-stratified case-crossover design was applied. Each case was matched to 3 or 4 control days which were selected on the same days of the week in the same calendar month. Then a distributed lag nonlinear model（DLNM） was used to estimate the exposure-response relationship between selected air pollutants and mortality, which was used to construct the AQHI. Finally, AQHI was classified into four levels according to the air pollutant guidance limit values from World Health Organization Global Air Quality Guidelines（AQG 2021）, and the excess risks（ERs） were calculated to compare the AQHI based on single-pollutant model and the J-AQHI based on multi-pollutant model.[Results] PM_2.5, NO₂, SO₂, and O3 were selected by Lasso regression to establish DLNM model. The ERs for an interquartile range（IQR） increase and 95% confidence intervals（CI） for PM_2.5, NO₂, SO₂ and O3 were 0.71%（0.34%–1.09%）, 2.46%（1.78%–3.15%）, 1.25%（0.9%–1.6%）, and 0.27%（-0.11%–0.65%） respectively. The distribution of J-AQHI was right-skewed, and it was divided into four levels, with ranges of 0-1 for low risk, 2-3 for moderate risk, 4-5 for high health risk, and ≥6 for severe risk, and the corresponding proportions were11.25%, 64.61%, 19.33%, and 4.81%, respectively. The ER（95%CI） of mortality risk increased by 3.61%（2.93–4.29） for each IQR increase of the multi-pollutant based J-AQHI, while it was 3.39%（2.68–4.11） for the single-pollutant based AQHI.[Conclusion] The J-AQHI generated by multi-pollutant model demonstrates the actual exposure health risk of air pollution in the population and provides new ideas for further improvement of AQHI calculation methods.更多还原