| Over the thirty years after reform and opening, as the rapid economic development and urbanization, China has become one of the few countries with the severest outdoor air pollution problems. In recent years, China’s air pollution has gradually changed from the conventional coal combustion type to the mixed coal combustion/motor vehicle emission type. The complex air pollution is characterized by increasing and regionalized haze and photochemical smog. Outdoor air pollution in China has aroused wide health concerns. However, the evidence concerning the health effects of air pollution is limited in China, especially multi-center epidemiological studies, thus the knowledge about the health risks of complex air pollution are not sufficient. Given the current levels and features of air pollution in China,4questions remain to be addressed:(1) How are the exposure-response relationships of major air pollutants and health effects?(2) Are any subpopulations more sensitive than the others? Are people in any specific seasons were more susceptible?(3) In air pollution epidemiology, how to cope with the criticism towards the traditional exposure assessment method?(4) Is the existing Air Quality Index (AQI) or its variant (Air Pollution Index, API) suitable to communicating the air pollution-related health risks? As the most ideal and convincing approach in air pollution epidemiology, prospective cohort study is very difficult to implement due to enormous consumption. On the other hand, death data is easier to collect and more robust to socioeconomic factors than the morbidity data. Therefore, in order to cope with the above4questions, we examined the short-term health effects of air pollution by analyzing the association between air pollutant concentrations and daily mortality. This dissertation was divided the following4parts.Part I The association of major air pollutants and daily mortalityThe exposure-response relationship was crucial to the revision of ambient air quality standards for World Health Organization (WHO) and major countries. Most of the existing epidemiological evidence was obtained in developed countries. Therefore, doubts remained concerning the applicability of WHO-Air Quality Guidelines (AQG) in China, which were formulated mostly based on the epidemiological findings in developed countries. The interim targets of AQG also failed to consider the exposure-response relationships in developing countries including China. Previous studies have indicated that the coefficients of exposure-relationships curves in China were somewhat lower than those in developed countries, and a threshold of safe levels might exist. Therefore, the previous findings need to be confirmed via massive epidemiological studies in more Chinese cities.We performed an epidemiological study on the acute health effects of air pollution in17typical Chinese cities, named by China Air Pollution and Health Effects Study (CAPES for short). In CAPES, we systematically analyzed the associations between major air pollutants and daily mortality. The CAPES cities incorporated Anshan, Beijing, Fuzhou, Guangzhou, Hangzhou, Hong Kong, Lanzhou, Nanjing, Shanghai, Shenyang, Suzhou, Taiyuan, Tangshan, Tianjin, Urumqi, Wuhan and Xi’an. We collected daily total non-accidental, cardiovascular and respiratory mortality from the Center for Disease Control and Prevention in each city. We obtained the daily average concentrations from Environmental Monitoring Centers in each city. The air pollutants in this study included particulate matter with an aerodynamic diameter less than10μm (PM10), particulate matter with an aerodynamic diameter less than2.5μm (PM2.5), particulate matter with an aerodynamic diameter less than10μm (PM2.5-10), sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO) and ozone (O3).We apply the internationally-accepted two-stage Bayesian hierarchical models to estimate the acute effects of air pollution on daily mortality in each city and national average levels. In the first stage, we used the time-series method to analyze the association between air pollution and daily mortality. Specifically, its core model was the generalized additive model (GAM) linked the Quasi-Poisson distribution; included the natural spline smooth functions of calendar time to exclude the long-term and seasonal trends of daily mortality, and use partial autocorrelation functions to guide the selection of the degrees of freedom in the smoothness of time; included the indicator variable of "day of week" to exclude any variation within a week; added natural smooth functions of mean temperature and relative humidity to account for the nonlinear confounding effects of weather conditions on daily mortality. In the second stage, we applied the Bayesian hierarchical models to combine the city-specified effect estimates, thus we obtained the estimates of the acute effects of air pollutants at the national average level. The effect estimates were presented as the posterior mean and95%posterior intervals (PIs) of the percent increase in daily mortality associated with a10μg/m3increase in air pollutant concentrations. Additionally, we plotted the exposure-response relationship curves in each city, and then pooled these curves at the national average level. Generally, we entered the moving average air pollutant concentrations of the current day and previous day (lag01) in the models.Among the CAPES cities, the levels of PM and SO2were much higher than those in developed countries. The results showed significant heterogeneity in the effects estimates. At the national average levels, a10μg/m3increase of PM10was associated with increases of0.35%(95%PI:0.18%,0.52%),0.44%(95%PI:0.23%,0.64%).0.44%(95%PI:0.23%,0.64%) in total, cardiovascular and respiratory mortality, respectively; a10μg/m3increase of SO2was associated with increases of0.75%(95%PI:0.47%,1.02%),0.83%(95%PI:0.47%,1.19%),1.25%(95%PI:0.78%,1.73%) in total, cardiovascular and respiratory mortality, respectively; a10μg/m3increase of NO2was associated with increases of1.63%(95%PI:1.09%,2.17%),1.80%(95%PI:1.00%,2.59%),2.52%(95%PI:1.44%,3.59%) in total, cardiovascular and respiratory mortality, respectively; a10μg/m3increase of PM2.5was associated with increases of0.40%(95%PI:0.18%,0.61%),0.47%(95%PI:0.23%,0.72%),0.46%(95%PI:0.19%,0.74%) in total, cardiovascular and respiratory mortality, respectively; a10μg/m3increase of PM2.5-10was associated with increases of0.43%(95%PI:0.07%,0.79%),0.33%(95%PI:-0.01%,0.67%),0.58%(95%PI:-0.35%,1.51%) in total, cardiovascular and respiratory mortality, respectively; a1mg/m3increase of CO was associated with increases of2.89%(95%PI:1.68%,4.11%),4.17%(95%PI:2.66%,5.68%) and2.03%(95%PI:0.49%,4.55%) in total, cardiovascular and respiratory mortality, respectively; a10μg/m3increase of O3was associated with increases of0.32%(95%PI:0.18%,0.46%),0.52%(95%PI:0.28%,0.76%),-0.04%(95%PI:-0.41%,0.34%) in total, cardiovascular and respiratory mortality, respectively. Except for PM25, the coefficients of exposure-response relationships were comparable to those obtained in developed countries. The exposure-response relationship coefficient for PM2.5and daily mortality was almost one half lower than that in developed countries. The exposure-response relationship curves were almost linear with no observable thresholds for various pollutants.Sensitivity analyses suggested that the effect estimates were generally strongest on lag01days and could persist for3-4days; in the two-pollutant models, PM10, PM2.5, NO2, CO and O3had the independent health effects, while the effects of SO2 and PM2.5-10lost statistical significance; the effect estimates tended to decreased a little as more aggressive control of time and adjustment for more lags of temperatures, but remain stable to some extent. The sensitivity analyses revealed that the estimates of the acute mortality effects of air pollution were generally insensitive to the selection of model parameters.Part Ⅱ The analyses on the susceptible factors of the acute health effects of air pollutionWhen exposed to the same levels of air pollutants, some subjects in a population might have severer reactions to the health hazards. The identification of these susceptible subgroups has important public health significance. Air pollution is a mixture with very complex components that may vary from season to season. In addition, the exposure pattern for local residents may also differ by season. Therefore, it is plausible that the acute health risks associated with air pollution might have seasonal patterns, i.e., residents might be more susceptible to air pollution in some seasons than in other seasons.We compared the effects of air pollutants on daily total mortality classified by age groups, sex, and education attainment. The results showed that the elders of65years above were significantly more susceptible than the persons aged5-64years; females were a litter more susceptible than the males; people with low education attainment were more susceptible than those with high attainment of education.In order to understand the seasonal patterns of air pollutant’s health effects, we divided CAPES cities according to the geographic latitudes into3categories:northern cities (Anshan, Shenyang, Beijing, Tangshan, Tianjin, Taiyuan, Xi’an, Urumqi, Lanzhou), middle cities (Nanjing, Hangzhou, Suzhou, Shanghai, Wuhan), and southern cities (Fuzhou, Guangzhou, Hong Kong). We selected PM10as the indicator of air pollution to explore the seasonality in effect estimates. We applied two-stage Bayesian hierarchical models to estimate the national-and regional-average estimates of air pollution’s effects. At the first stage, we constructed "main effect" models,"seasonal effect", and "smooth effects" models, respectively when estimating the city-specified effects. At the second stage, we used the Bayesian hierarchical models to combine the city-specified estimates. The "main effect" models assumed the PM10effects hold the same across the year; the "seasonal effect" allowed differences of PM10effects in various seasons; and the "smooth effects" models allowed the PM10effects change smoothly within a year.The daily mortality and air pollution levels differed appreciably across seasons. Generally, winter and summer days have higher levels of air pollution than summer and fall days; the northern cities has severer air pollution than the middle and southern cities. According to the "main effects" models, a10μg/m3increase of PM10was associated with0.35%(95%PI:0.13%,0.56%) increase of daily total mortality. In the "seasonal effects" models, the PM10effects in winter and summer were strongest, but were statistically insignificant in spring and fall. A10μg/m3increase of PM10was associated with increases of0.45%(95%PI:0.15%,0.76%),0.17%(95%PI:-0.09%,0.43%),0.55%(95%PI:0.15%,0.96%) and0.25%(95%PI:-0.05%,0.56%) in winter, spring, summer and fall. Southern cities had the biggest effect estimates, followed by middle and northern cities. We plotted the time-varying effects of PMio within a year using the "smooth effects" models. It clearly suggested two peaks of PM10effects in winter and summer, consistent with the results of the "seasonal effects" models. This seasonal pattern was consistently observed in northern, middle and southern cities, and was relative robust to the alternative adjustment of gaseous pollutants and temperatures on more lag days. Similar seasonality patterns still remained when we used interquartile range as alternative effect measure scale. At last, we also analyze the seasonality of SO2and NO2effects using the same models, and found similar "two peak" patterns in effect estimates.Part Ⅲ An improvement to air pollution epidemiology:Indoors and outdoors time-weighted exposure assessmentIn traditional time-series studies, the exposure assessment was always criticized in that the outdoor measurements at monitors were directly used as the surrogate of personal exposure levels, but one person typically spent most of his or her time indoors. Therefore, this traditional exposure assessment approach could not reflect the true exposure levels of ambient air pollution.Therefore, we applied mathematical statistical techniques to account for the indoors/outdoors air exchange rate, PM surface removal rates, and PM penetration factors, in order to obtain the simulated indoor exposure to outdoor-originated PM. Then, we calculated the total person exposure as the time-weighted PM10concentrations (PM10wtd for short) according to the indoor/outdoor time ratio of activity. Considering some of the above parameters were not available in some CAPES cities, we set a base scenario and multiple sensitivity scenarios. We obtained the associations between PM10wtd and daily mortality using the similar methodology as illustrated in the first part of this thesis.The results showed that the simulated PM10wtd is30%-60%lower than the original ambient measurements. At the national average levels, a10μg/m3increase of PM10wtd in the base scenario was associated with0.65%(95%PI:0.29%,1.00%) increase of total mortality; in various sensitivity scenarios, the magnitude of association between PM10wta and mortality was generally2times higher than the original estimates that directly used the outdoor measurements. Furthermore, the t values in regression models for all scenarios were higher than that in the original models, suggesting an improvement of precision of effect estimates. According to the Cochran’s Q heterogeneity tests, the Q statistics obtained in all scenarios were smaller than that gained in the original model, revealing some decrease in the between-city heterogeneity of PM10effects. In summary, the proposed method could improve the exposure assessment and further improve the magnitude and precision of estimates of PM10acute effects in a time-series study.Part Ⅳ The construction and validation of China’s air quality health indexThe sound evaluation of ambient air quality could provide necessary evidence for the authorities to prevent air pollution and improve the air quality management. Currently, major countries including China generally adopted AQI/API to evaluate the daily air quality. AQI/API only used the pollutant with the highest exceeding ratio to a regulatory limit to reflect the air quality status. Thus they ignored the impact of other co-pollutants to air quality, and failed to reflect the well-known linear with no thresholds association between air pollution and health effects. Furthermore, the cut points used for calculating AQI/API in China were made partly according to the findings of long-term cohort study in the US. Therefore, the air quality daily reporting system in China might have important shortcomings.We directly applied the exposure-response relationship obtained in the CAPES to construct a newly Air Quality Health Index (AQHI), and selected PM10+NO2, or PM2.5+NO2as the indicators of air quality. The formula of AQHI were PM10AQHI=10/17*100*[exp(0.000154*PM10)-1+exp(0.000664*NO2-1], and PM2.5AQHI=10/15*100*[exp(0.000172*PM2.5)-1+exp(0.000664*NO2-1]. The results of validation studies showed that at the national average levels, the association magnitude between AQHI and mortality were larger than that between API/AQI and mortality. In Shanghai, the power of AQHI to predict daily morbidity (outpatient visits, emergency room visits and hospitalizations) was much stronger than API. Therefore, our studies indicated AQHI has stronger ability of predicting health effects than API/AQI. On the other hand, theoretically, AQHI has the following4strengths:(1) AQHI directly adopted the exposure-response relationship in China, thus could reflect the characteristic of air pollution-related health effects;(2) AQHI directly applied the findings from time-series models, thus could reflect sensitively the acute health effects associated with a short-term change in air quality;(3) AQHI might be a relatively comprehensive indicator of air quality and its health effects;(4) AQHI could reflect the well-known linear non-threshold association with health effects. Therefore, the proposed AQHI provided a more effective tool to communicate the air pollution-related health risks to the public, and might be beneficial to the existing air quality daily reporting system in China.In summary, the main findings included:1) Air pollution in Chinese cities was severe, especially for particles and sulfur dioxide;2) The evidence on the significant associations between major air pollutants (PM10/PM2.5/SO2/NO2/CO/O3)and adverse health effects were generally consistent with those obtained from developed countries;3) SO2and PM2.5-10might not have independent associations with health outcomes;4) The exposure-response relationship curves were almost linear without any thresholds;5) The size of PM2.5effects per unit increase was almost one half lower than developed countries;6) Elders, females and people with low education attainment were more susceptible to the acute effects of air pollution;7) The health risks of PM10in winter and summer days were stronger than in spring and fall days, and PM10in southern cities had stronger health effects than in northern cities;8) Integrating the indoor exposure to PM10infiltrated from outdoors in the exposure assessment could improve the effect estimates in air pollution epidemiology.; and9) AQHI based on time-series estimates might be better than the existing AQI and API, and might be a good tool for communicating the air pollution-related health risks to the public. These findings might have important policy implications. |
PDF Full Text Request