Effects of ambient air quality improvement on mortality from acute air pollution exposure in Beijing

Exposure to air pollutants increase the mortality of population. Developing countries have taken measures to control air pollution. To explore the effects of air quality improvement on mortality, air quality and acute exposure-response coefficients of excess death in Beijing since the 1990’s were analyzed. It was divided into five stages according to the concentration level of pollutants. Coefficients for period 1990 – 2013 were obtained by retrieving literatures published before December 31, 2019. The coefficients for period 2015 – 2017 were obtained by analyzing the daily data of air pollutant concentration, meteorological and human mortality conducting Poisson Generalized Additive Model (GAM). Meta-analysis of random effect model was used to estimate the integrated coefficient of multiple studies at each stages. Comparative analysis was used to analyze the variation of air quality and coefficients in different stages. The results showed that the concentrations of sulfur dioxide (SO 2 ), nitrogen dioxide (NO 2 ), particulate matter with aerodynamic diameter ≤10 μ m (PM 10 ) and ≤2.5 μ m (PM 2.5 ) decreased by up to 50%, 21%, 22% and 15% in different stages. The coefficient of SO 2 on death from respiratory diseases decreased by up to 63.79%. The coefficients of NO 2 on mortality from non-accidental causes, cardiovascular disease, and respiratory disease decreased by up to 0.95%, 1.34% and 0.54%. The coefficients of PM 10 , PM 2.5 on mortality from cardiovascular diseases and respiratory disease were decreased by up to 0.19%, 0.31%, 0.65% and 0.36%. Continued improvements in air quality have reduced the acute impact on the health of the local population.

The impacts of air pollutants on human health have been one of the major problems of public health in the world. Air pollutants caused 4.9 million deaths worldwide in 2017, an increase of 5.8% over 2007 (GBD 2017). Studies have found that air pollution prevention and control measures not only reduce concentration of pollutants, but also reduce pollution-related mortality (Peters et al. 2009), which is of great significance in reducing the mortality caused by air pollution worldwide.
China is one of the developing countries with severe air pollution. In recent years, Chinese have been concerned about the health effects of air pollution, and a lot of measures had been taken to control air pollution. Air quality has continued to improve, but air pollutants remain at high levels. At present, there are few studies on the effect of air quality improvement on human mortality at high level of pollution.
Beijing, as the capital of China, had become a national key development area since the middle of the 20th century and had a rapid social and economic development in 21st century. However the serious air pollution had become a major environmental and public health problem at the same time. In order to prevent the further deterioration of air pollution, since the 1990s, Beijing had adopted a series of measures for air pollution prevention and control and carried out a lot of researches on the health effects of air pollution. In order to provide references for Beijing to adopt more targeted air pollution control measures, as well as pollution prevention and control measures for other polluted areas, we chose Beijing as the research site for a longitudinal study. However, at different stages of development, the major pollutants and research hotspots were also different.
In recent decades, Beijing has gone through a gradual transformation of air pollution from coal-burning pollution, coal-burning and vehicle-exhaust mixture pollution to vehicle-exhaust pollution. In view of the different pollution characteristics, targeted air pollution prevention and control measures were also implemented in five stages. First, before 1998, it was mainly coal-burning pollution. Due to the limitation of monitoring data, the research mainly focused on the health impact of total suspended particulates (TSP) and sulfur dioxide (SO2). Second, since Beijing proposed to bid for the 2008 Olympic Games in 1998, gas had replaced coal as the main domestic fuel. In addition, the increase of number of motor vehicles that use gasoline as fuel increased the concentration of PM2.5 and NO2. The air pollution in Beijing had changed from coalburning pollution to mixture pollution of coal-burning and vehicle-exhaust pollution (Chen et al. 2004;Wang et al. 2013). In order to reduce the concentration of air pollutants, Beijing had implemented control measures for the pollution of industrial, civil coal burning, vehicle and dust. Beijing entered the stage of comprehensive air pollution control. With the increase of monitoring indicators of air pollutants, studies on the health effects of particulate matter with aerodynamic diameter ≤10μm or ≤2.5 μm (PM10, PM2.5) and nitrogen dioxide (NO2) on population were gradually carried out, which were consistent with international research hotspots (Yang et al. 2018).
Third, since 2004, SO2 pollution reached the control target and the annual average concentration was lower than the limit of the national ambient air quality standard in

Materials and Methods
Based on the characteristics of pollutants and measures of pollution prevention and control in Beijing, the study was divided into five stages: stage 1 (1990-1997)

Statistical Analysis
Time -series analysis: GAM was used to analyze the exposure-response coefficients of short-term exposure air pollutants to the mortality. The model is as follows: To better compare with previous studies, we used relative risk (RR) to express the results. The formula for the relationship between RR and β is shown below: The coefficients of exposure-response obtained from the literature also include excess risk (ER, %). ER is also converted to RR. The formula is shown below: Meta -analysis: Using R3.6 software, Meta random effect model was used to quantitatively estimate the multiple exposure -response coefficients of each stage, so as to obtain the coefficients of air pollution on mortality of population at every stage for trend comparison.
Comparison analysis: Compared to the previous stage, the reduction proportions of both concentrations of pollutant and coefficients of health impact were calculated.
Where x is the variable to be compared; j is the stage to be compared, j-1 is the stage before stage j. if the number of coefficient was 2 or more, meta-analysis will be used to integrate the coefficients. For a stage or a pollutant with few papers, such as SO2 at the first stage, this study used the coefficient of one study to represent the coefficient of this stage or this pollutant. Supplementary analysis was performed for the fifth stage where the raw data could be obtained, which will be described in more detail below. The increasing of concentrations of SO2 and PM10 per 10μg/m 3 may significantly increase the relative risk of death from non-accidental in the population (Table 3). In particular, exposure to SO2 may increase the relative risk of death from respiratory disease and exposure to PM10 may increase the relative risk of death from cardiovascular disease.

Meta -Analysis of Multiple Studies
Meta-analysis was performed on quantitative pooled estimates from multiple studies at stage2 -stage5. Except for the effects of SO2 pollution at stage 3, the effects of SO2, PM10, and PM2.5 on the mortality of the population were statistically significant (Table   4.).  Monitoring data for NO2, PM10 and PM2.5 began in the stage 2. The decrease of NO2 and PM10 were not obvious in the third stage with decrease both by 8%, but it was obvious in the fourth stage with decrease by 21% and 22% respectively. PM10 also decreased by 15% in the fifth stage, but NO2 decreased by only 4%. PM2.5 decreased obviously in stage3 and stage5, by 15% and 11% respectively, but in stage 4, by only 7%. At stage 1, the epidemiological study of air pollution on death had just started, and focus on the effects of SO2 exposure. It was found that SO2 had an impact on death, and the influence coefficient was much higher than other stages. At stage 2, the studies found that SO2 still significantly increased the death from cardiovascular and respiratory diseases in the population, and compared with stage 1, the RR decreased by 9.34% and 1.57% respectively. The effect of SO2 on mortality from respiratory diseases was not statistically significant at stage3 when SO2 concentration reached limit of ambient air quality in China. But at stage5, there was statistical significance again.

Change of Air Quality and Coefficients of Health Impact at Different Stages
There was no significant downward trend in the relative risk of death from cardiovascular and respiratory disease due to NO2 exposure at stage2-4. However, compare to stage3, the RR for non-accidental total death increased by 0.71% at stage4.
The relative risk of deaths from non-accidental, cardiovascular and respiratory disease due to NO2 exposure at stage5 decreased by 0.93%, 1.32% and 0.53%, respectively.
The relative risk of death due to PM10 exposure was decreased by 0.19% in the third stage, compared with the second stage. However, the decreasing trend of stage 3-5 was not obvious. The relative risk of death due to PM2.5 exposure had an obvious trend of gradual reduction. Compared with the previous stage, the relative risk of PM2.5 to non-accidental total deaths was reduced by 0.16% and 0.15% respectively in the fourth and fifth stages, to deaths from cardiovascular diseases were reduced by 0.64% and 0.17%. The relative risk of death from respiratory diseases was reduced by 0.35% in stage 4.
on deaths of population were not statistically significant, except for the effects of PM2.5 on deaths from respiratory diseases. However, the effect of SO2 on the mortality from respiratory diseases was statistically significant.
Note: Fig3 is the figure after removing the maximum value in stage1 to show the trend more clearly.

Discussion
The

AUTHOR CONTRIBUTIONS
The corresponding author had full access to all study data and was responsible for the decision to submit for publication. Study concept and design: XU Dongqun, HAN Jingxiu, MENG Congshen, LIU Jingyi. Acquisition of data: HAN Jingxiu, MENG