OPEN Journal of Exposure Science and Environmental Epidemiology (2018) 28, 297–304 www.nature.com/jes ORIGINAL ARTICLE The spatial variation in the effects of air pollution on cardiovascular mortality in Beijing, China Wenjing Li1, Yang Cao2,3, Runkui Li4,5, Xinming Ma1, Jieying Chen1, Zhenglai Wu1 and Qun Xu1 Owing to lack of data from multiple air quality monitoring stations, studies about spatial association between concentrations of ambient pollutants and mortality in China are rare. To investigate the spatial variation of association between concentrations of particulate matter less than 10 μm in aerodynamic diameter (PM10), nitrogen dioxide (NO2) and carbon monoxide (CO) and cardiovascular mortality in Beijing, China, we collected data including daily deaths, concentrations of PM10,NO2 and CO, and meteorological factors from 1 January 2009 to 31 December 2010 in all 16 districts of Beijing. Generalized additive model (GAM) and generalized additive mixed model (GAMM) were used to examine the citywide and district-specific effects of PM10,NO2 and CO on cardiovascular mortality. The citywide effect derived from GAMM was lower than that derived from GAM, and the strongest effects were identified for 2-day moving average lag 0–1. The interquartile increases in concentrations of PM10,NO2 and CO were associated with 2.46 (95% confidence interval (CI), 1.22–3.72), 4.11 (95%CI, 2.82–5.42) and 2.23 (95%CI, 1.14–3.33) percentage increases in daily cardiovascular mortality by GAMM, respectively. The relative risk of each district compared with reference district was generally statistically significant. The death risk associated with air pollutants varies across different geographic districts in Beijing. The data indicate that the risk is high in suburban areas and rural counties. We found significant and spatially varied adverse effects of air pollution on cardiovascular deaths across the rural and urban areas in Beijing. Journal of Exposure Science and Environmental Epidemiology (2018) 28, 297–304; doi:10.1038/jes.2016.21 Keywords: air pollution; cardiovascular mortality; generalized additive mixed model; generalized additive model; spatio-temporal analysis INTRODUCTION monoxide (CO) and nitrogen oxides (NOX) emissions, increased by 15 Association of air pollution with human health has been evaluated 2.5 times, totaling 5.4 million at the end of 2013. Beijing is still in numerous time-series studies by fitting Poisson regression facing a substantial challenge of balancing economic growth and model at community or city level.1–7 Results of these studies environmental protection. suggest that short-term exposure to ambient air pollution is Concerns about effects of air pollution on population health, for 6,16 16–18 associated with increased mortality from cardiovascular diseases. example, cardiovascular diseases, respiratory diseases and 19–22 Meanwhile, extensive clinical and toxicological studies have also emergency hospital visits have been increasing in Beijing. provided evidence for stronger association between air pollution However, most of the existing studies characterized exposures and deaths from cardiovascular diseases.8–11 However, most of using citywide average ambient concentrations of pollutant, these studies were conducted in developed countries and there which ignored the within-city spatial variation in concentrations were limited amount of research done in China. of ambient air pollutants because of lacking data from multiple air From 2003 to 2012, annual average levels of particulate matter quality monitoring (AQM) stations. In fact, exposure levels vary 23 less than 10 μm in aerodynamic diameter (PM10) in the 31 greatly across the whole area of Beijing. A study showed that air provincial capitals in China were significantly reduced (Figure 1); quality in the southern part was worse than that in the northern however, they were still at the higher end of air pollution level in part in Beijing.24 And from the southeast plain to the northwest the world. In 2013, the report on Global Environmental Competi- mountain area in Beijing, pollution levels presented a decreasing tiveness ranked the air quality of China 87th out of 133 trend. Chen et al.25 pointed that the spatial variation in the effects 12 countries. Beijing, the capital of China, experienced rapid urban of PM10 on health outcomes needs to be considered in assessing expansion and transportation development and hence, led to a the health effect of air pollution, particularly for big cities. The 13 26 serious air pollution problem in the past 10 years, during which study by Zhang et al. applied district-specificPM10 concentra- the number of its resident population increased from 14.56 million tions to quantify the impact of PM10 on population health. in 2003 to 21.15 million in 2013.14 Over the same period, the However, the study did not include the suburban and rural number of automobiles, a major source of ambient air carbon districts and examined the effects of nitrogen dioxide (NO2) and 1Department of Epidemiology and Biostatistics, Institute of Basic Medicine Sciences Chinese Academy of Medical Sciences, School of Basic Medicine Peking Union Medical College, Beijing 100005, China; 2Unit of Biostatistics, Institute of Environmental Medicine, KarolinskaInstitutet, Stockholm 17177, Sweden; 3Clinical Epidemiology and Biostatistics, Faculty of Medicine and Health, Örebro University, Örebro 70185, Sweden; 4College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China and 5State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Science, Beijing 100101, China. Correspondence: Qun Xu, Department of Epidemiology and Biostatistics, Institute of Basic Medicine Sciences Chinese Academy of Medical Sciences, School of Basic Medicine Peking Union Medical College, Beijing 100005, China. Tel.: +86 10 69156403; Fax: +86 10 69156403. E-mail: [email protected] Received 23 September 2015; revised 2 February 2016; accepted 3 February 2016 Air pollution effects on cardiovascular deaths Li et al 298 Figure 1. Annual average ambient levels of PM10 and NO2 in 31 provincial capitals (including Beijing) of China 2003–2012. The gray-dotted line indicates air quality as Class II based on its annual levels by the National Ambient Air Quality Standards (NAAQS) in China. Data source: China Statistic Yearbook 2004–2013. CO. In our study, we aimed to apply the generalized additive main independent variables. Generalized additive model (GAM) was used mixed model (GAMM) to investigate the spatial characteristics of to associate the mortalities and the pollutants,30 and GAMM to account for 31 association between air concentrations of PM10,NO2 and CO and overdispersion and random effect of district to the additive predictor, daily cardiovascular mortality in the whole area of Beijing. under the assumption that the number of daily deaths from cardiovascular diseases follows a Poisson distribution. Natural splines were used to fit the non-linear time trend of the mortality, adjusting for potential confounders, MATERIALS AND METHODS such as day of week (DOW) and meteorological conditions.32 The full GAMM included smoothing function for variables time, temperature and The study was based on Causes of Death Registry (CDR) of the China humidity, categorical variable DOW and random effect of district, which Centers for Disease Control and Prevention (China CDC), but did not access can be expressed as: any individually identifiable patient data. The study was approved by the ð ð ÞÞ ¼ β þ β ð Þþβ ð Þþ ð ; ´ Þ Institutional Review Board of institute of Basic Medical Sciences, Chinese log E Yi;t 0 1 DOWt 2 pollutioni;t s timet 2 years Academy of Medical Sciences. þ ð ; Þþ ð ; Þþμ s temperaturet 5 s humidityt 2 Zi where E(Y ) refers to the expected count in district i on day t,DOWis Study Area and Period i,t dummy variable for day of week, and Zi is a random intercept for districts i. Beijing is located in the North China Plain with an area of approximate s(.) is the smoothing function realized by natural spline with 2 degrees of 16,807 km2 and a population of 21.15 million by the end of 2013.14 The freedom (d.f.) per year for time to adjust for seasonal and long-term trends study area covered 16 administrative districts of Beijing, including six and 5 degrees of freedom for temperature and 2 for relative humidity. urban districts (Dongcheng, Xicheng, Chaoyang, Fengtai, Shijingshan and The most appropriate degrees of freedom of the time, temperature and Haidian), eight suburban districts (Mentougou, Fangshan, Tongzhou, relative humidity were determined by minimizing Akaike Information Shunyi, Changping, Daxing, Huairou and Pinggu) and two rural counties Criterion (AIC).31 (Yanqing and Miyun) (Figure 2). The study was carried out from 1 January We also compared the citywide effects derived from GAMM and those 2009 to 31 December 2010. from GAM. The GAM and GAMM models include the same non-parametric smoothing function for PM10,NO2 and CO and other confounding factors. Data Collection The GAM can be expressed as: ð ð ÞÞ ¼ β þ β ð Þþβ ð Þþ ð ; ´ Þ The data here include the number of daily cardiovascular deaths, air log E Yt 0 1 DOWt 2 pollutiont s timet 2 years quality, road density and meteorological conditions. þ ð ; Þþ ð ; Þ s temperaturet 5 s humidityt 2 Cardiovascular mortality data were obtained from CDR of China Centers for Disease Control and Prevention (CDC). All deaths in CDR were coded Furthermore, we examined the associations with different lag structures, according to the International Classification of Diseases, version 10 including single-day lags (from lag 0 to lag 4) and multiday lags (moving – – (ICD-10), with codes I00-I99 used as deaths from cardiovascular diseases. average lag 0 1 to lag 0 4). Sensitivity analyses were conducted to investigate the robustness of the Air quality data, including the concentrations of PM10,NO2 and CO, were generated by district AQM stations at 12 fixed sites in Beijing (Figure 2).