Environ Sci Pollut Res (2014) 21:4236–4244 DOI 10.1007/s11356-013-2374-6
RESEARCH ARTICLE
Impact of haze and air pollution-related hazards on hospital admissions in Guangzhou, China Zili Zhang & Jian Wang & Lianghua Chen & Xinyu Chen & Guiyuan Sun & Nanshan Zhong & Haidong Kan & Wenju Lu
Received: 31 July 2013 / Accepted: 13 November 2013 / Published online: 5 December 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Guangzhou is a metropolitan in south China with unique pollutants and geographic location. Unlike those in western countries and the rest of China, the appearance of haze in Guangzhou is often (about 278 days per year on average of 4 years). Little is known about the influence of these hazes on health. In this study, we investigated whether short-term exposures to haze and air pollution are associated with hospital admissions in Guangzhou. The relationships between haze, air pollution, and daily hospital admissions during 2008–2011 were assessed using generalized additive model. Studies were categorized by gender, age, season, lag, and disease category. In haze episodes, an increase in air pollutant emissions corresponded to 3.46 (95 % CI, 1.67, 5.27) increase in excessive risk (ER) of total hospital admissions at lag 1, 11.42 (95 % CI, 4.32, 18.99) and 11.57 (95 % CI, 4.38, 19.26) increases in ERs of cardiovascular illnesses at lags 2 and 4 days, respectively. As to total hospital Responsible editor: Constantini Samara Zili Zhang and Jian Wang contributed equally to this article. Electronic supplementary material The online version of this article (doi:10.1007/s11356-013-2374-6) contains supplementary material, which is available to authorized users. Z. Zhang : J. Wang : L. Chen : X. Chen : G. Sun : N. Zhong : W. Lu (*) Guangzhou Institute of Respiratory Disease, State Key Laboratory of Respiratory Diseases, The First Affiliated Hospital, Guangzhou Medical University, 151 Yanjiang Road, Guangzhou, Guangdong 510120, People’s Republic of China e-mail: [email protected]
admissions, an increase in NO2 was associated with a 0.73 (95 % CI, 0.11, 1.35) and a 0.28 (95 % CI, 0.11, 0.46) increases in ERs at lag 5 and lag 05, respectively. For respiratory illnesses, increases in NO2 was associated with a 1.94 (95 % CI, 0.50, 3.40) increase in ER at lag 0, especially among chronic obstructive pulmonary disease. Haze (at lag1) and air pollution (for NO2 at lag 5 and for SO2 at lag3) both presented more drastic effects on the 19 to 64 years old and in the females. Together, we demonstrated that haze pollution was associated with total and cardiovascular illnesses. NO2 was the sole pollutant with the largest risk of hospital admissions for total and respiratory diseases in both single- and multi-pollutant models. Keywords Air pollutants . Hospital admissions . Haze . Respiratory and cardiovascular diseases Abbreviations COPD URI IHD PM10 SO2 NO2 GAM ER CI ICD10
Chronic obstructive pulmonary disease Upper respiratory infection Ischemic heart disease A diameter measuring less than 10 μm Sulfur dioxide Nitrogen dioxide Generalized additive model Excessive risk Confidence interval The International Classification of Disease 10th Revision
H. Kan (*) School of Public Health, Key Lab of Public Health Safety of the Ministry of Education, Fudan University, Box 249, 130 Dong-An Road, Shanghai 200032, China e-mail: [email protected]
Introduction
W. Lu Department of Laboratory Medicine, The First Affiliated Hospital, Guangzhou Medical University, Guangzhou, China
Many epidemiological studies have shown that short-term increases in outdoor haze or air pollution are associated with
Environ Sci Pollut Res (2014) 21:4236–4244
acute rise of hospital admissions (Dominici et al. 2002, Guo et al. 2009, Teksam et al. 2010). However, majority of such studies were conducted in western countries, where people have different demographic characteristics compared with people from Asian areas. These studies demonstrated that the elevated risk of hospital admissions associated with short-term increases in outdoor haze or air pollution was primarily due to respiratory and cardiovascular admissions (van der Zee et al. 2000). Moreover, studies on the relations between haze, ambient air pollution, and hospital admissions of total (no accidental), respiratory, and cardiovascular illnesses are quite scarce in Asia, especially the haze. Also, studies of the adverse effects of air pollution on hospital admissions in China have mostly been conducted in only a few large cities in China, including Shanghai (Kan and Chen 2003a, 2003b), Beijing (Xu et al. 1994), Chongqing (Zheng et al. 2003), as well as Shenyang (Xu et al. 2000). There remain needs for expressing the uniqueness of Guangzhou regions and replicating the findings in other cities of China. And to our knowledge, no earlier study has examined the relationships between haze, daily outdoor air pollution, and hospital admissions in Guangzhou. With the rapid development of economy over the past three decades, Guangzhou has experienced haze periods that range from slight transient haze episodes to severe haze episodes, with a sharp increase in air pollutant emissions (Streets et al. 2008). Haze is now a worldwide phenomenon and has caught great attention for its great adverse effects on visibility, cloud formation, public health, and even global climate (Menon et al. 2002, So and Wang 2003). Besides haze, the association between air pollution and hospital admissions due to total, respiratory, and cardiovascular illnesses has also not been reported. It has been formerly found that long-term exposure to outdoor air pollution can increase hospital admissions, reduce the function of the organs such as lung function (Ackermann-Liebrich et al. 1997; Brunekreef 2002; Franco Suglia et al. 2008; Gotschi et al. 2008), and develop chronic respiratory illnesses (Schikowski et al. 2005; Strak et al. 2010). So, we hypothesized that there may be short-term effects of haze or air pollution on the number of daily hospital admissions; in addition, the present study attempts to do statistical verification regarding haze and air pollutant-related hazardous effects on human health by using a quantitative analysis. In current study, we collected the information of daily hospital admissions of total (no accidental), respiratory, and cardiovascular illnesses, the daily monitoring data of outdoor air pollution, the daily meteorological data, and days were divided into haze periods or not at the same time in Guangzhou during the period of Jan of 2008 to Dec of 2011, and we applied GAM to perform the time series analysis to investigate the associations between haze pollution, the outdoors air pollutants, and the number of daily hospital admissions.
4237
Materials and methods Study area and population This study looked into daily hospital admissions in relation to haze pollution and outdoors air pollutants in Guangzhou for the 4-year period from Jan 1, 2008 to Dec 31, 2011 (1,461 days). Guangzhou, one of the most metropolitan cities in China, comprises urban/sub-urban districts and counties, having a subtropical climate, with abundant rainfall in warm seasons and an annual average temperature of 22.3 °C. The city’s annual average rainfall has been 1,200 mL in these years, with a total area of 72,361 km2, and has a population of 13.2 million by the end of 2000, representing 1 % of China’s total. Our monitoring stations included ten districts (289 km2): Yuexiu district, Liwan district, Tianhe district, Haizhu district, Huangpu district, Baiyun district, Panyu district, Huadu district, Luogang district, and Nansha district. We excluded two suburban counties (Zengcheng and Conghua) due to inadequate air pollution monitoring stations in those places. The target population included all permanent residents living in this area, around 8.7 millions in 2008. The major air pollution source is automobile exhaust emissions.
Daily hospital admissions data For this analysis, daily hospital admissions data (excluding accidents and injuries) were obtained from the database of Guangzhou Health Insurance Bureau (GHIB), which is in charge of the Guangzhou Health Insurance System and provides compulsory universal health insurance during 2008– 2011. In Guangzhou, all hospitals are contracted with the GHIB. Computerized records of daily hospital admissions are available for each contracted hospital. All hospitals must submit standard claim documents for medical expenses on a computerized form that includes the necessary information for each admission. Both total non-accidental and cause-specific hospital admissions (A00-R99) were assessed in our work. We investigated study characteristics including disease diagnosis codes (e.g., the International Classification of Disease 10th Revision (ICD10), disease categories including respiratory illnesses (J00-J98) for URI (J00-J06, except J00, X02J00, and X04), pneumonia (J12-J18), chronic obstructive pulmonary disease (COPD) (J20: acute bronchitis, J40-J44: bronchitis, not specified as acute or chronic; simple and mucopurulent chronic bronchitis; unspecified chronic bronchitis; emphysema, J47:other chronic obstructive pulmonary disease), and asthma (J45); cardiovascular illnesses (I00-I99) for dysrhythmias (I44-I49), IHD (I20-I25), and peripheral and cerebrovascular disease (I63-I74, I80-I82); and age of subjects (e.g., 0–1; 2–18; 19–64; ≥65 years).
4238
Air pollution and meteorological data Outdoor air quality data were provided by the Guangzhou Municipal Environmental Protection Monitoring Center, a central governmental agency. Urban haze period is defined as the weather phenomenon which leads to atmospheric daily average visibility less than 10 km and relative humidity of below 90 % (Sun et al. 2006). Daily outdoors air concentrations of PM10 (particulate matter, a diameter measuring less than 10 μm), SO2 (sulfur dioxide), and NO2 (nitrogen dioxide) were obtained as daily mean values measured from ten state-controlled monitoring stations in Guangzhou. The monitoring stations were fully automated and provided daily readings of PM10, SO2, and NO2. For every day, haze pollution and outdoors air pollutants data were extracted from all of the monitoring stations and averaged. In addition, according to the technical guidelines of the Chinese government, the location of these monitoring stations should be far from traffic intersections or major industrial polluters and should also have sufficient distance to any other emitting source. Thus, the monitoring data could reflect the general background outdoor air pollution levels in Guangzhou. To allow adjustment for the effects of weather on hospital admissions, meteorological data (daily mean temperature, relative humidity, and pressure) were obtained from the Guangzhou Meteorological Bureau. The weather data were measured at a fixed-site station located in Baiyun district of Guangzhou. This station belongs to the Guangzhou Meteorological Office, and the monitoring standard is consistent with international World Meteorological Organization standard, and the data were representative, but small variations in parts of the study area due to the micro-climate effect should not be ruled out. According to the annual temperature of Guangzhou, we divided the season into the warm season from April to October and the cool season from November to March. Data analysis The objectives of the data analysis was to quantify the association between daily hospital admissions and haze pollution, outdoors air pollutants concentrations, while adjusting for effects of time trends (i.e., long-term effects caused by factors such as genetic background and short-term effects such as workday and weekday effects) and meteorological factors in the multivariable modeling. Because the daily number of hospital admissions approximately followed a Poisson distribution, and the relations between hospital admissions and explanatory variables were mostly nonlinear (Daniels et al. 2004); the core analysis we utilized was a generalized additive model (GAM) with log link and Poisson error that accounted for fluctuations in daily numbers of hospital admissions. First of all, we constructed the basic models (without a pollutant and within other covariates, such as day of the week), and we incorporated smoothed spline functions of time and weather
Environ Sci Pollut Res (2014) 21:4236–4244
conditions, which could include non-linear and non-monotonic links between hospital admissions and time/weather conditions, offering a flexible modeling tool (Dominici et al. 2002). After establishment of the basic models, we then introduced the pollutant or haze pollution variables and analyzed their effects on all non-accidental causes, respiratory illnesses, and cardiovascular illnesses. To compare the relative quality of the hospital admissions predictions across these non-nested models, Akaka’s Information Criterion (AIC) was used as a measure of how well the model fitted the data. Smaller AIC values indicate the preferred model. Briefly, we fitted the following log-linear GAM to obtain the estimated pollution log-relative rate β in the study district: Log E ðYiÞ ¼ βX i þ DOW þ ns ðtime;df Þ þ ns ðtemperature=humidity=pressure;df Þ Here, E (Y i) indicates the expected number of hospital admissions at day i; β indicates the log-relative rate of hospital admissions related with a unit increase of air pollutants and with haze or not; X i represents the concentrations of pollutants at day i; DOW is the dummy variable for day of the week; ns is the non-parametric spline function of calendar time, temperature, humidity, and pressure. Regarding the basic models, we also conducted some sensitivity analysis following Qian’s method (Qian et al. 2007). We initialized the df as 7 df/year for time, 3 df for temperature, humidity, and pressure. We fitted both single-pollutants models and multi-pollutant models (models with a different combination of four pollutants per model) to assess the stability of pollutants’ effect. Moreover, we examined the effect of air pollutants with different lag (L) structures of single-day lag (from L0 to L5) and multi-day lag (L01 and L05). Here a lag of 0 day (L0) corresponded to the current-day pollution, and a lag of 1 day referred to the previous-day concentration. In multi-day lag models, L05 corresponded to 6-day cumulative effects value of the current and previous 5 days. Moreover, the meteorological factors used in the lag models (single-day lag and multi-day lag models) were the current day data. We analyzed effects of outdoors air pollution separately for the warm season (from April to October) and cool season (from November to March). All statistical analyses were conducted in R 2.10.1 using the MGCV package. The results were expressed as the percent change in daily hospital admissions in haze periods or per 10 μg/m3 increase of pollutant concentrations.
Results The distribution of daily hospital admissions, meteorological factors, and outdoors air pollutants in Guangzhou between Jan 1, 2008 and Dec 31, 2011 (1,461 days in total) were recorded
Environ Sci Pollut Res (2014) 21:4236–4244
in Table 1. During the 4-year study period, the mean daily concentration of NO2 was 56.0 μg/m3, 32.0 μg/m3 for SO2, and 62.0 μg/ m3 for PM10. Meanwhile, the average daily humidity, pressure, and temperature were 68.0, 1012, and 25 °C, reflecting the subtropical climate in Guangzhou. The daily mean PM10 concentration was above the air quality standard (WHO guideline for PM10 is 50 μg/m3 for daily average). And SO2 and NO2 also showed high concentrations which exceeded the standard (the 24 h mean concentration limits of SO2 and NO2 are 150 and 80 μg/m3, respectively). Spearman correlation coefficients between haze pollution, outdoors air pollutants parameters, and meteorological factors were recorded in Table 2. Generally, there was a positive correlation between three pollutants (PM10, NO2, and SO2) and a negative relation between three pollutants and haze pollution. Figure 1 and Supplementary Table 1 summarized results from the single-day lag (L0-L5) and multi-day lag (L01 and L05) for ERs increases in hospital admissions for total, respiratory, and cardiovascular illnesses in haze episodes. It indicates an increase in air pollutant emissions corresponded to a 3.46 (95 % CI, 1.67, 5.27) increase in ER of total hospital admissions at lag 1 and 11.42 (95 % CI, 4.32, 18.99) and a 11.57 (95 % CI, 4.38, 19.26) increases in ERs of cardiovascular hospital admissions at lag 2 and at lag 4 days, separately. Figure 2 and Supplementary Table 1 summarized ERs estimates and 95 % CIs for the percent increase in hospital admissions for total, respiratory, and cardiovascular diseases Table 1 Distribution of daily data on air pollution, hospital admissions, weather parameters in Guangzhou, China (from Jan 1, 2008 to Dec 31, 2011)
Pollutants NO2 (μg/m3) SO2 (μg/m3) PM10 (μg/m3) Meteorological data Humidity (%) Pressure (Pa) Temperature (°C) Daily admissions (N) Total Respiratory Cardiovascular Total daily admission by age (n) 0–1 2–18 19–64 ≥65 Total daily admission by sex (n) Male Female
4239
by the analyzed causes of air pollutants related to a 10 μg/m3 increase in the exposure variables, obtained from singlepollutant (top rows) and multi-pollutant models (bottom rows). Figure 2a showed positive and statistically significant association between total admissions and increases in NO2, SO2, and PM10 in multi-pollutant models; unlike in the singlepollutant models, most risk estimates were reduced. We estimated that increase in NO2 was associated with a 0.73 (95 % CI, 0.11, 1.35) and a 0.28 (95 % CI, 0.11, 0.46) increases in ERs at lag 5 and lag 05, respectively; an increase in SO2 with a 1.07 (95 % CI, 0.54, 1.60) and a 1.16 (95 % CI, 0.61, 1.70) increases in ERs at lag 1 and lag 3, separately; and an increase in PM10 with a 0.19 (95 % CI, 0.07, 0.30) increase in ER at lag 05. The associations between respiratory admissions and NO2 at lag 0 and lag 2 were statistically significant in singlepollutant models (Fig. 2b). ERs increases in NO2 at lag 0 and lag 2 were associated with a 1.47 (95 % CI, 0.62, 2.33) and a 1.4 (95 % CI, 0.33, 2.48) in respiratory admissions, respectively. In multi-pollutant models, we found statistically significant associations, with ER in NO2 at lag 0 with a 1.94 (95 % CI, 0.50, 3.40) increase in respiratory admissions, especially among COPD with a 3.57 (95 % CI, 1.85, 5.33) (Fig. 3 and Supplementary Table 2). In addition, other components were not associated with total or cardiovascular or respiratory hospital admissions in multi-pollutant model. For each statistically significant pollutant at different lag times, analyses were replicated for different age groups, sex, and season (Fig. 4 and Supplementary Table 3). Overall, the
N
Min
P25
Median
P75
Max
1,461 1,461 1,461
17.6 2.0 8.0
43.2 19.0 43.0
56.0 32.0 62.0
78.4 47.0 88.4
179.2 128.0 202.0
1,461 1,461 1,461
10 993 6
59 1,008 18
68 1,012 25
78 1,018 28
96 1,031 34
1,461 1,461 1,461
147 32 18
1,340 296 129
2,088 462 231
2,420 575 289
5,233 1,380 403
1,461 1,461 1,461 1,461
8 15 72 34
64 134 664 386
108 240 1,156 587
125 272 1,327 602
249 397 2,580 1,420
1,461 1,461
79 64
680 664
1,043 1,043
1,215 1,210
2,677 2,601
4240 Table 2 Spearman correlation coefficients between daily air pollutant concentrations, haze, and weather conditions in Guangzhou, China
*P
RESEARCH ARTICLE
Impact of haze and air pollution-related hazards on hospital admissions in Guangzhou, China Zili Zhang & Jian Wang & Lianghua Chen & Xinyu Chen & Guiyuan Sun & Nanshan Zhong & Haidong Kan & Wenju Lu
Received: 31 July 2013 / Accepted: 13 November 2013 / Published online: 5 December 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Guangzhou is a metropolitan in south China with unique pollutants and geographic location. Unlike those in western countries and the rest of China, the appearance of haze in Guangzhou is often (about 278 days per year on average of 4 years). Little is known about the influence of these hazes on health. In this study, we investigated whether short-term exposures to haze and air pollution are associated with hospital admissions in Guangzhou. The relationships between haze, air pollution, and daily hospital admissions during 2008–2011 were assessed using generalized additive model. Studies were categorized by gender, age, season, lag, and disease category. In haze episodes, an increase in air pollutant emissions corresponded to 3.46 (95 % CI, 1.67, 5.27) increase in excessive risk (ER) of total hospital admissions at lag 1, 11.42 (95 % CI, 4.32, 18.99) and 11.57 (95 % CI, 4.38, 19.26) increases in ERs of cardiovascular illnesses at lags 2 and 4 days, respectively. As to total hospital Responsible editor: Constantini Samara Zili Zhang and Jian Wang contributed equally to this article. Electronic supplementary material The online version of this article (doi:10.1007/s11356-013-2374-6) contains supplementary material, which is available to authorized users. Z. Zhang : J. Wang : L. Chen : X. Chen : G. Sun : N. Zhong : W. Lu (*) Guangzhou Institute of Respiratory Disease, State Key Laboratory of Respiratory Diseases, The First Affiliated Hospital, Guangzhou Medical University, 151 Yanjiang Road, Guangzhou, Guangdong 510120, People’s Republic of China e-mail: [email protected]
admissions, an increase in NO2 was associated with a 0.73 (95 % CI, 0.11, 1.35) and a 0.28 (95 % CI, 0.11, 0.46) increases in ERs at lag 5 and lag 05, respectively. For respiratory illnesses, increases in NO2 was associated with a 1.94 (95 % CI, 0.50, 3.40) increase in ER at lag 0, especially among chronic obstructive pulmonary disease. Haze (at lag1) and air pollution (for NO2 at lag 5 and for SO2 at lag3) both presented more drastic effects on the 19 to 64 years old and in the females. Together, we demonstrated that haze pollution was associated with total and cardiovascular illnesses. NO2 was the sole pollutant with the largest risk of hospital admissions for total and respiratory diseases in both single- and multi-pollutant models. Keywords Air pollutants . Hospital admissions . Haze . Respiratory and cardiovascular diseases Abbreviations COPD URI IHD PM10 SO2 NO2 GAM ER CI ICD10
Chronic obstructive pulmonary disease Upper respiratory infection Ischemic heart disease A diameter measuring less than 10 μm Sulfur dioxide Nitrogen dioxide Generalized additive model Excessive risk Confidence interval The International Classification of Disease 10th Revision
H. Kan (*) School of Public Health, Key Lab of Public Health Safety of the Ministry of Education, Fudan University, Box 249, 130 Dong-An Road, Shanghai 200032, China e-mail: [email protected]
Introduction
W. Lu Department of Laboratory Medicine, The First Affiliated Hospital, Guangzhou Medical University, Guangzhou, China
Many epidemiological studies have shown that short-term increases in outdoor haze or air pollution are associated with
Environ Sci Pollut Res (2014) 21:4236–4244
acute rise of hospital admissions (Dominici et al. 2002, Guo et al. 2009, Teksam et al. 2010). However, majority of such studies were conducted in western countries, where people have different demographic characteristics compared with people from Asian areas. These studies demonstrated that the elevated risk of hospital admissions associated with short-term increases in outdoor haze or air pollution was primarily due to respiratory and cardiovascular admissions (van der Zee et al. 2000). Moreover, studies on the relations between haze, ambient air pollution, and hospital admissions of total (no accidental), respiratory, and cardiovascular illnesses are quite scarce in Asia, especially the haze. Also, studies of the adverse effects of air pollution on hospital admissions in China have mostly been conducted in only a few large cities in China, including Shanghai (Kan and Chen 2003a, 2003b), Beijing (Xu et al. 1994), Chongqing (Zheng et al. 2003), as well as Shenyang (Xu et al. 2000). There remain needs for expressing the uniqueness of Guangzhou regions and replicating the findings in other cities of China. And to our knowledge, no earlier study has examined the relationships between haze, daily outdoor air pollution, and hospital admissions in Guangzhou. With the rapid development of economy over the past three decades, Guangzhou has experienced haze periods that range from slight transient haze episodes to severe haze episodes, with a sharp increase in air pollutant emissions (Streets et al. 2008). Haze is now a worldwide phenomenon and has caught great attention for its great adverse effects on visibility, cloud formation, public health, and even global climate (Menon et al. 2002, So and Wang 2003). Besides haze, the association between air pollution and hospital admissions due to total, respiratory, and cardiovascular illnesses has also not been reported. It has been formerly found that long-term exposure to outdoor air pollution can increase hospital admissions, reduce the function of the organs such as lung function (Ackermann-Liebrich et al. 1997; Brunekreef 2002; Franco Suglia et al. 2008; Gotschi et al. 2008), and develop chronic respiratory illnesses (Schikowski et al. 2005; Strak et al. 2010). So, we hypothesized that there may be short-term effects of haze or air pollution on the number of daily hospital admissions; in addition, the present study attempts to do statistical verification regarding haze and air pollutant-related hazardous effects on human health by using a quantitative analysis. In current study, we collected the information of daily hospital admissions of total (no accidental), respiratory, and cardiovascular illnesses, the daily monitoring data of outdoor air pollution, the daily meteorological data, and days were divided into haze periods or not at the same time in Guangzhou during the period of Jan of 2008 to Dec of 2011, and we applied GAM to perform the time series analysis to investigate the associations between haze pollution, the outdoors air pollutants, and the number of daily hospital admissions.
4237
Materials and methods Study area and population This study looked into daily hospital admissions in relation to haze pollution and outdoors air pollutants in Guangzhou for the 4-year period from Jan 1, 2008 to Dec 31, 2011 (1,461 days). Guangzhou, one of the most metropolitan cities in China, comprises urban/sub-urban districts and counties, having a subtropical climate, with abundant rainfall in warm seasons and an annual average temperature of 22.3 °C. The city’s annual average rainfall has been 1,200 mL in these years, with a total area of 72,361 km2, and has a population of 13.2 million by the end of 2000, representing 1 % of China’s total. Our monitoring stations included ten districts (289 km2): Yuexiu district, Liwan district, Tianhe district, Haizhu district, Huangpu district, Baiyun district, Panyu district, Huadu district, Luogang district, and Nansha district. We excluded two suburban counties (Zengcheng and Conghua) due to inadequate air pollution monitoring stations in those places. The target population included all permanent residents living in this area, around 8.7 millions in 2008. The major air pollution source is automobile exhaust emissions.
Daily hospital admissions data For this analysis, daily hospital admissions data (excluding accidents and injuries) were obtained from the database of Guangzhou Health Insurance Bureau (GHIB), which is in charge of the Guangzhou Health Insurance System and provides compulsory universal health insurance during 2008– 2011. In Guangzhou, all hospitals are contracted with the GHIB. Computerized records of daily hospital admissions are available for each contracted hospital. All hospitals must submit standard claim documents for medical expenses on a computerized form that includes the necessary information for each admission. Both total non-accidental and cause-specific hospital admissions (A00-R99) were assessed in our work. We investigated study characteristics including disease diagnosis codes (e.g., the International Classification of Disease 10th Revision (ICD10), disease categories including respiratory illnesses (J00-J98) for URI (J00-J06, except J00, X02J00, and X04), pneumonia (J12-J18), chronic obstructive pulmonary disease (COPD) (J20: acute bronchitis, J40-J44: bronchitis, not specified as acute or chronic; simple and mucopurulent chronic bronchitis; unspecified chronic bronchitis; emphysema, J47:other chronic obstructive pulmonary disease), and asthma (J45); cardiovascular illnesses (I00-I99) for dysrhythmias (I44-I49), IHD (I20-I25), and peripheral and cerebrovascular disease (I63-I74, I80-I82); and age of subjects (e.g., 0–1; 2–18; 19–64; ≥65 years).
4238
Air pollution and meteorological data Outdoor air quality data were provided by the Guangzhou Municipal Environmental Protection Monitoring Center, a central governmental agency. Urban haze period is defined as the weather phenomenon which leads to atmospheric daily average visibility less than 10 km and relative humidity of below 90 % (Sun et al. 2006). Daily outdoors air concentrations of PM10 (particulate matter, a diameter measuring less than 10 μm), SO2 (sulfur dioxide), and NO2 (nitrogen dioxide) were obtained as daily mean values measured from ten state-controlled monitoring stations in Guangzhou. The monitoring stations were fully automated and provided daily readings of PM10, SO2, and NO2. For every day, haze pollution and outdoors air pollutants data were extracted from all of the monitoring stations and averaged. In addition, according to the technical guidelines of the Chinese government, the location of these monitoring stations should be far from traffic intersections or major industrial polluters and should also have sufficient distance to any other emitting source. Thus, the monitoring data could reflect the general background outdoor air pollution levels in Guangzhou. To allow adjustment for the effects of weather on hospital admissions, meteorological data (daily mean temperature, relative humidity, and pressure) were obtained from the Guangzhou Meteorological Bureau. The weather data were measured at a fixed-site station located in Baiyun district of Guangzhou. This station belongs to the Guangzhou Meteorological Office, and the monitoring standard is consistent with international World Meteorological Organization standard, and the data were representative, but small variations in parts of the study area due to the micro-climate effect should not be ruled out. According to the annual temperature of Guangzhou, we divided the season into the warm season from April to October and the cool season from November to March. Data analysis The objectives of the data analysis was to quantify the association between daily hospital admissions and haze pollution, outdoors air pollutants concentrations, while adjusting for effects of time trends (i.e., long-term effects caused by factors such as genetic background and short-term effects such as workday and weekday effects) and meteorological factors in the multivariable modeling. Because the daily number of hospital admissions approximately followed a Poisson distribution, and the relations between hospital admissions and explanatory variables were mostly nonlinear (Daniels et al. 2004); the core analysis we utilized was a generalized additive model (GAM) with log link and Poisson error that accounted for fluctuations in daily numbers of hospital admissions. First of all, we constructed the basic models (without a pollutant and within other covariates, such as day of the week), and we incorporated smoothed spline functions of time and weather
Environ Sci Pollut Res (2014) 21:4236–4244
conditions, which could include non-linear and non-monotonic links between hospital admissions and time/weather conditions, offering a flexible modeling tool (Dominici et al. 2002). After establishment of the basic models, we then introduced the pollutant or haze pollution variables and analyzed their effects on all non-accidental causes, respiratory illnesses, and cardiovascular illnesses. To compare the relative quality of the hospital admissions predictions across these non-nested models, Akaka’s Information Criterion (AIC) was used as a measure of how well the model fitted the data. Smaller AIC values indicate the preferred model. Briefly, we fitted the following log-linear GAM to obtain the estimated pollution log-relative rate β in the study district: Log E ðYiÞ ¼ βX i þ DOW þ ns ðtime;df Þ þ ns ðtemperature=humidity=pressure;df Þ Here, E (Y i) indicates the expected number of hospital admissions at day i; β indicates the log-relative rate of hospital admissions related with a unit increase of air pollutants and with haze or not; X i represents the concentrations of pollutants at day i; DOW is the dummy variable for day of the week; ns is the non-parametric spline function of calendar time, temperature, humidity, and pressure. Regarding the basic models, we also conducted some sensitivity analysis following Qian’s method (Qian et al. 2007). We initialized the df as 7 df/year for time, 3 df for temperature, humidity, and pressure. We fitted both single-pollutants models and multi-pollutant models (models with a different combination of four pollutants per model) to assess the stability of pollutants’ effect. Moreover, we examined the effect of air pollutants with different lag (L) structures of single-day lag (from L0 to L5) and multi-day lag (L01 and L05). Here a lag of 0 day (L0) corresponded to the current-day pollution, and a lag of 1 day referred to the previous-day concentration. In multi-day lag models, L05 corresponded to 6-day cumulative effects value of the current and previous 5 days. Moreover, the meteorological factors used in the lag models (single-day lag and multi-day lag models) were the current day data. We analyzed effects of outdoors air pollution separately for the warm season (from April to October) and cool season (from November to March). All statistical analyses were conducted in R 2.10.1 using the MGCV package. The results were expressed as the percent change in daily hospital admissions in haze periods or per 10 μg/m3 increase of pollutant concentrations.
Results The distribution of daily hospital admissions, meteorological factors, and outdoors air pollutants in Guangzhou between Jan 1, 2008 and Dec 31, 2011 (1,461 days in total) were recorded
Environ Sci Pollut Res (2014) 21:4236–4244
in Table 1. During the 4-year study period, the mean daily concentration of NO2 was 56.0 μg/m3, 32.0 μg/m3 for SO2, and 62.0 μg/ m3 for PM10. Meanwhile, the average daily humidity, pressure, and temperature were 68.0, 1012, and 25 °C, reflecting the subtropical climate in Guangzhou. The daily mean PM10 concentration was above the air quality standard (WHO guideline for PM10 is 50 μg/m3 for daily average). And SO2 and NO2 also showed high concentrations which exceeded the standard (the 24 h mean concentration limits of SO2 and NO2 are 150 and 80 μg/m3, respectively). Spearman correlation coefficients between haze pollution, outdoors air pollutants parameters, and meteorological factors were recorded in Table 2. Generally, there was a positive correlation between three pollutants (PM10, NO2, and SO2) and a negative relation between three pollutants and haze pollution. Figure 1 and Supplementary Table 1 summarized results from the single-day lag (L0-L5) and multi-day lag (L01 and L05) for ERs increases in hospital admissions for total, respiratory, and cardiovascular illnesses in haze episodes. It indicates an increase in air pollutant emissions corresponded to a 3.46 (95 % CI, 1.67, 5.27) increase in ER of total hospital admissions at lag 1 and 11.42 (95 % CI, 4.32, 18.99) and a 11.57 (95 % CI, 4.38, 19.26) increases in ERs of cardiovascular hospital admissions at lag 2 and at lag 4 days, separately. Figure 2 and Supplementary Table 1 summarized ERs estimates and 95 % CIs for the percent increase in hospital admissions for total, respiratory, and cardiovascular diseases Table 1 Distribution of daily data on air pollution, hospital admissions, weather parameters in Guangzhou, China (from Jan 1, 2008 to Dec 31, 2011)
Pollutants NO2 (μg/m3) SO2 (μg/m3) PM10 (μg/m3) Meteorological data Humidity (%) Pressure (Pa) Temperature (°C) Daily admissions (N) Total Respiratory Cardiovascular Total daily admission by age (n) 0–1 2–18 19–64 ≥65 Total daily admission by sex (n) Male Female
4239
by the analyzed causes of air pollutants related to a 10 μg/m3 increase in the exposure variables, obtained from singlepollutant (top rows) and multi-pollutant models (bottom rows). Figure 2a showed positive and statistically significant association between total admissions and increases in NO2, SO2, and PM10 in multi-pollutant models; unlike in the singlepollutant models, most risk estimates were reduced. We estimated that increase in NO2 was associated with a 0.73 (95 % CI, 0.11, 1.35) and a 0.28 (95 % CI, 0.11, 0.46) increases in ERs at lag 5 and lag 05, respectively; an increase in SO2 with a 1.07 (95 % CI, 0.54, 1.60) and a 1.16 (95 % CI, 0.61, 1.70) increases in ERs at lag 1 and lag 3, separately; and an increase in PM10 with a 0.19 (95 % CI, 0.07, 0.30) increase in ER at lag 05. The associations between respiratory admissions and NO2 at lag 0 and lag 2 were statistically significant in singlepollutant models (Fig. 2b). ERs increases in NO2 at lag 0 and lag 2 were associated with a 1.47 (95 % CI, 0.62, 2.33) and a 1.4 (95 % CI, 0.33, 2.48) in respiratory admissions, respectively. In multi-pollutant models, we found statistically significant associations, with ER in NO2 at lag 0 with a 1.94 (95 % CI, 0.50, 3.40) increase in respiratory admissions, especially among COPD with a 3.57 (95 % CI, 1.85, 5.33) (Fig. 3 and Supplementary Table 2). In addition, other components were not associated with total or cardiovascular or respiratory hospital admissions in multi-pollutant model. For each statistically significant pollutant at different lag times, analyses were replicated for different age groups, sex, and season (Fig. 4 and Supplementary Table 3). Overall, the
N
Min
P25
Median
P75
Max
1,461 1,461 1,461
17.6 2.0 8.0
43.2 19.0 43.0
56.0 32.0 62.0
78.4 47.0 88.4
179.2 128.0 202.0
1,461 1,461 1,461
10 993 6
59 1,008 18
68 1,012 25
78 1,018 28
96 1,031 34
1,461 1,461 1,461
147 32 18
1,340 296 129
2,088 462 231
2,420 575 289
5,233 1,380 403
1,461 1,461 1,461 1,461
8 15 72 34
64 134 664 386
108 240 1,156 587
125 272 1,327 602
249 397 2,580 1,420
1,461 1,461
79 64
680 664
1,043 1,043
1,215 1,210
2,677 2,601
4240 Table 2 Spearman correlation coefficients between daily air pollutant concentrations, haze, and weather conditions in Guangzhou, China
*P
