Respiratory Medicine (2015) 109, 372e378
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/rmed
A GIS-based spatial correlation analysis for ambient air pollution and AECOPD hospitalizations in Jinan, China Wenqiao Wang a,b, Yangyang Ying a, Quanyuan Wu c, Haiping Zhang c, Dedong Ma a,*, Wei Xiao a,* a Department of Respiratory Medicine, Qilu Hospital, Shandong University, No. 107, Wenhua Xi Road, Jinan, Shandong, 250012, PR China b Department of Respiratory Diseases, China-Japan Friendship Hospital, Peking University, Beijing, PR China c College of Population, Resources and Environment, Shandong Normal University, No. 88, Wenhua Dong Road, Jinan, Shandong, 250012, PR China
Received 26 September 2014; accepted 17 January 2015
Available online 27 January 2015
KEYWORDS AECOPD hospitalization; Ambient air pollution; GIS; Spatial autocorrelation
Summary Background: Acute exacerbations of COPD (AECOPD) are important events during disease procedure. AECOPD have negative effect on patients’ quality of life, symptoms and lung function, and result in high socioeconomic costs. Though previous studies have demonstrated the significant association between outdoor air pollution and AECOPD hospitalizations, little is known about the spatial relationship utilized a spatial analyzing technique- Geographical Information System (GIS). Objective: Using GIS to investigate the spatial association between ambient air pollution and AECOPD hospitalizations in Jinan City, 2009. Methods: 414 AECOPD hospitalization cases in Jinan, 2009 were enrolled in our analysis. Monthly concentrations of five monitored air pollutants (NO2 , SO2, PM10, O3 , CO) during January 2009eDecember 2009 were provided by Environmental Protection Agency of Shandong Province. Each individual was geocoded in ArcGIS10.0 software. The spatial distribution of five pollutants and the temporal-spatial specific air pollutants exposure level for each individual was estimated by ordinary Kriging model. Spatial autocorrelation (Global Moran’s I) was employed to explore the spatial association between ambient air pollutants and AECOPD hospitalizations. A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model.
* Corresponding authors. E-mail addresses: [email protected] (W. Wang), [email protected] (Y. Ying), [email protected] (Q. Wu), gissuifeng@ 163.com (H. Zhang), [email protected] (D. Ma), [email protected] (W. Xiao). http://dx.doi.org/10.1016/j.rmed.2015.01.006 0954-6111/ª 2015 Elsevier Ltd. All rights reserved.
Spatial correlation between air pollution and AECOPD hospitalizations
373
Results: At residence, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistical significant spatially clustered. The Z-score of SO2, PM10, CO, O3, NO2 at residence is 15.88, 13.93, 12.60, 4.02, 2.44 respectively, while at workplace, concentrations of PM10, SO2, O3, CO and AECOPD hospitalization cases showed statistical significant spatially clustered. The Z-score of PM10, SO2, O3, CO at workplace is 11.39, 8.07, 6.10, and 5.08 respectively. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations due to AECOPD. Conclusions: Ambient air pollution is correlated with AECOPD hospitalizations spatially. A 10 mg/ m3 increase of PM10 at workplace was associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations due to AECOPD in Jinan, 2009. As a spatial data processing tool, GIS has novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research. ª 2015 Elsevier Ltd. All rights reserved.
Background Acute exacerbations of Chronic Obstructive Pulmonary Disease (AECOPD) are important events during disease procedure. AECOPD have negative effect on patients’ quality of life [1,2], symptoms and lung function [3], and result in high socioeconomic costs [4]. It is estimated that AECOPD would lead to 110,000 deaths and over 500,000 hospitalizations per year, the direct cost due to AECOPD was over 18 billion dollars [5]. The acute exacerbations accelerate the rate of decline of lung function [6,7], correlated with significant mortality, particularly in those requiring hospitalizations (Figs. 1 and 2). Exacerbations of COPD can be triggered by various factors. Though respiratory tract infections appear to be the most common causes, not all patients suffering AECOPD have the evidence of infection. Growing evidence support that ambient air pollution is an environmental triggering factor for acute exacerbation of COPD [8e10]. Geographical information system (GIS) aims to integrate the digital capture, management, analysis and visualization of geographically referenced data spatially. Better interpretation of the patterns, trends and relationships between disease and demography, environment, space and time could be obtained through GIS. Thus GIS has important application in medical and health area, especially in the application of etiological research [11,12]. To our knowledge, few studies employed GIS technique to assess spatio-temporal specific exposure to air pollutants in the spatial association analysis between admissions for AECOPD and air pollution. There is no similar study in East China. We aim to assess the spatial association between ambient air pollution exposure and AECOPD hospitalizations in Jinan City through GIS.
Methods Study area & study period We set the study in Jinan City, 2009. Jinan is the capital of Shandong Province in Eastern China. It is located in the
north-western part of Shandong province at 36 010 e37 320 northern latitude and 116 110 e117 440 east of Greenwich.
Target population All cases enrolled in our study were from five large-scale hospitals interspersed in Jinan City. COPD hospitalization cases met with the following inclusion criteria were included in our study: (1) Hospitalization due to acute exacerbations of COPD, identified by International Statistical Classification of Diseases, 10th Revision (ICD-10) codes, J40-J44; (2) Resided and worked in study area (Jinan City) during study period (Jan 2009eDec 2009); (3) Adults patients (age > 18 years). Exclusion criteria: (1) Patients who did not reside or work in Jinan City during 2009. (2) In order to avoid the impact of occupational exposure to air pollutants, patients who worked at highpolluting environment were excluded. Written informed consent was obtained prior to data collection. The study and consent procedure was approved by the Ethics Committee of Qilu Hospital of Shandong University (No. KYLL-2014-4).
Exposure assessment Subjects were geocoded using home residence and working site respectively in a Geographical Information System. We applied addresses geocoding techniques using ArcGIS 10.0 software. Each subject was shown on a map as a precise mark in correspondence with his/her home residence or working site respectively. For each marking site, concentrations of air pollutants were estimated using ordinary Kriging method to assess spatiotemporal-specific air pollution exposure. The monthly average concentrations of air pollutants from
374
W. Wang et al.
Figure 1 Distribution of SO2 concentration and AECOPD hospitalization (at residence). An example of the estimated monthly air pollution levels using monitoring data and Kriging method, distribution of residence of AECOPD hospitalization. Take SO2 concentration distribution and admission data of April, 2009 in Jinan, China as an example.
January, 2009 to December, 2009 were provided by Environmental Protection Agency of Shandong Province. The air pollution concentrations, i.e. nitrogen dioxide (NO2), ozone (O3), sulfur dioxide (SO2), particulate matter with aerodynamic diameter less than 10 microns (PM10), carbon monoxide (CO), were monitored in 15 monitoring stations interspersed in different sub districts of Jinan City.
Statistical analysis Global Moran’s I statistic was used for spatial autocorrelation analysis in ArcGIS 10.0. Moran’s I, p value and Z score
were calculated to test the spatially clustered tendency between concentrations of air pollutants and AECOPD hospitalization cases. Confidence level of 99% was selected. Values of P < 0.01 were considered statistically significant. A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model. Monthly numbers of AECOPD admissions were selected as dependent variable. Age, gender, season of hospitalization, smoking status and air pollutants concentrations at residence and workplace were selected as dependent variables. After controlling for the confounding
Spatial correlation between air pollution and AECOPD hospitalizations
375
Figure 2 Distribution of PM10 concentration and AECOPD hospitalization (at workplace). An example of the estimated monthly air pollution levels using monitoring data and Kriging method, distribution of residence of AECOPD hospitalization. Take PM10 concentration distribution and admission data of April, 2009 in Jinan, China as an example.
effects of seasonality, age, gender and smoking status, monthly concentrations of PM10, SO2, CO, NO2, O3 at residence and workplace for each individual were added to the core model to determine the increase of COPD admissions for a 10 mg/m3 increase in each of these air pollutants. An interaction effect between gender (male or female) and smoking status (ex-smoker or non-smoker or current smoker) were analyzed in the model. Values of P < 0.05 were considered statistically significant. All the analysis were performed with SPSS 21.0 software.
Results Demographic information Altogether 414 admissions for acute exacerbations of COPD during January 1，2009eDecember 31，2009 were enrolled in our study. The basic demographic information of enrolled patients (such as age, gender, smoking status, allergy history and comorbidities) were shown in Table 1.
376
W. Wang et al.
Table 1 Characteristics of the whole subjects enrolled in the study.
Table 3 Spatial autocorrelation analysis between air pollutants at workplace and AECOPD hospitalization cases.
Variables
Air pollutants
Moran’s I
Z score
P value
SO2 NO2 PM10 CO O3
0.24 0.04 0.45 0.19 0.21
8.07 0.52 11.39 5.08 6.10
0.00a 0.60a 0.00a 0.00a 0.00a
Items
AECOPD hospitalization cases (n Z 414)
Age (mean SD) Gender (n, %) Male Female Smoking status Non-smoker Ex-smoker Current-smoker Allergy history Yes No Comorbidities Chronic cor pulmonale Ischemic heart disease Heart failure Hypertension Atrial fibrillation Lung cancer Diabetes Cerebrovascular disease
73.4 9.6 245 (59.2%) 169 (40.8%) 142 (34.3%) 138 (33.3%) 134 (32.4%) 143 (34.5%) 271 (65.5%) 255 (61.6%) 203 (49.0%) 55 (13.3%) 158 (38.2%) 23 (5.6%) 16 (3.9%) 50 (12.1%) 39 (9.4%)
a
Table 4 Spatial autocorrelation analysis between air pollutants at residence and AECOPD hospitalization cases (non-smokers þ ex-smokers). Air pollutants
Moran’s I
Z code
P value
SO2 NO2 PM10 CO O3
0.43 0.07 0.51 0.41 0.17
9.10 2.42 17.06 11.27 3.16
0.00a 0.07 0.00a 0.00a 0.00a
a
Spatial correlation at residence At residence, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered (Table 2). The Moran’s I of SO2, PM10, CO, O3, and NO2 at residence is 0.62, 0.43, 0.40, 0.21 and 0.14, respectively. The Z-score of SO2, PM10, CO, O3, and NO2 at residence is 15.88, 13.93, 12.60, 4.02 and 2.44, respectively. Of the five air pollutants, SO2 is of greatest spatial correlation with AECOPD hospitalization. We conducted spatial correlation in non-smokers plus ex-smokers, concentrations of SO2, PM10, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered at residence (Table 4).
Spatial correlation at workplace At workplace, concentrations of PM10, SO2, O3, CO and AECOPD hospitalizations cases showed spatially clustered tendency with statistical significance. The Z-score of PM10, SO2, O3, CO at workplace is 11.39, 8.07, 6.10, and 5.08 respectively. Concentrations of NO2 and AECOPD hospitalization cases did not show spatially clustered tendency (p Z 0.60). Of the four air pollutants, PM10 is of the greatest spatial correlation with AECOPD hospitalization (Table 3).
Data with statistical significance.
Data with statistical significance.
We conducted spatial correlation in non-smokers plus ex-smokers, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered at workplace (Table 5).
Predicting model A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model. Results of the core model were shown in Table 6 (Supplemental Materials). After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. Other pollutants showed no statistically significance. As to other potential confounders included in the model, the season of AECOPD hospitalization (cold season) and gender (male) were proved to be independent risk factors for AECOPD hospitalization. Gender and smoking status have interaction effect on AECOPD hospitalization, the interaction showed that in male ex-smokers, the incidence
Table 2 Spatial autocorrelation analysis between air pollutants at residence and AECOPD hospitalization cases.
Table 5 Spatial autocorrelation analysis between air pollutants at workplace and AECOPD hospitalization cases (non-smokers þ ex-smokers).
Air pollutants
Moran’s I
Z score
P value
Air pollutants
Moran’s I
Z code
P value
SO2 NO2 PM10 CO O3
0.62 0.14 0.43 0.40 0.21
15.88 2.44 13.93 12.60 4.02
0.00a 0.00a 0.00a 0.00a 0.000058a
SO2 NO2 PM10 CO O3
0.32 0.11 0.68 0.28 0.09
6.05 2.72 15.62 7.25 3.47
0.00a 0.00a 0.00a 0.00a 0.00a
a
Data with statistical significance.
a
Data with statistical significance.
Spatial correlation between air pollution and AECOPD hospitalizations of AECOPD hospitalizations decreased, while other combinations showed no statistical significance.
Discussion Our study assessed the spatial correlation of ambient air pollution related exposure on AECOPD admissions in Jinan City, China, 2009, and the results indicate that ambient air pollution is correlated with AECOPD hospitalizations spatially. At residence, SO2 is of the greatest spatial correlation with AECOPD hospitalization, while at workplace, PM10 is of the greatest spatial correlation with AECOPD hospitalization. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. Other pollutants showed no statistically significance. In recent years, air pollution has become a worldwide environmental issue, and China is facing the greatest challenge from deteriorating air quality. It is of great importance and urgency to find out the definite impact of air pollution. Jinan is one of the most polluted city in China, our research proved the spatial correlation between ambient air pollution and AECOPD hospitalizations in Jinan, 2009, which provided novel evidence and new direction for triggering factors and prevention of AECOPD. Though previous studies have demonstrated some kind of associations between ambient air pollution and COPD hospitalizations, publishing data were inconsistent and even conflicting, and little knowledge was known on the spatial correlation between ambient air pollution and AECOPD hospitalization cases. More evidence is needed to demonstrate the solid association between ambient air pollution and AECOPD hospitalizations. In the association study between O3 and COPD admissions, several previous studies have demonstrated the association. In APHEA project conducted in Europe, O3 was associated with daily admissions for COPD, the relative risks for a 50 mg/m3 increase in daily mean concentrations of O3 was 1.04 (1.02, 1.07) [13]. In the research conducted in Hong Kong, associations between O3 and COPD hospitalizations were convinced both in single-pollutant model and multipollutant model [14]. But in the study conducted by Sauerzapf V et al. [15], no associations were observed between O3 and admissions for COPD. Similar conditions exist in the association analysis between SO2 and COPD hospitalizations. In the research conducted in Hong Kong, associations between SO2 and COPD hospitalizations were convinced both in single-pollutant model and multi-pollutant model [14]. But in APHEA project, SO2 did not show a significant association with admissions for COPD (RR Z 1.02, 95%CI [0.98, 1.06]) [13]. Current studies support the role of NO2 in COPD hospitalizations. In APHEA project, NO2 was associated with daily admissions for COPD, the relative risks for a 50 mg/m3 increase in daily mean concentrations of NO2 was 1.02 (1.00, 1.05) [13]. In the study conducted in a rural county of England, each 10 mg/m3 increase in NO2 was associated with a 22% increase in the odds of COPD admission [15]. The role of PM10 in COPD hospitalizations was inconclusive. In a research conducted in Hong Kong,
377
PM10 was associated with COPD hospitalizations in single pollutant model, with RR Z 1.024 for admissions per 10 mg/ m3 increase, though this effect became unstatistically significant in multi-pollutant model [14]. Data were relatively limited referring to the role of CO in COPD hospital admissions. In the study conducted by Sauerzapf V et al. [15], 10 mg/m3 increase in CO was associated with a 2% increase in the odds of COPD admission. Utilizing spatial analysis in GIS, our results proved the spatial correlation between ambient air pollution and AECOPD hospitalization, which added new evidence to the association (especially spatial association) between ambient air pollution and AECOPD hospitalization. As a spatial data processing tool, GIS showed novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research. Most previous studies analyzed the relationship between daily admissions and daily air pollutants levels. Since the locations of patients were interspersed, it was much better to use individualized exposure rather than the mean levels of the whole city. Few studies have taken the individualized exposure into consideration in the association analysis between ambient air pollution and AECOPD admissions, especially using GIS for exposure assessment and spatial analysis. To our knowledge, there is no similar study in East China. Our study laid a solid foundation for the research of ambient air pollution and COPD hospitalizations in East China. Exposure to ambient air pollution is unavoidable. Residence exposures as well as workplace exposures are the main source of exposures in the majority. In our study, both residence exposures and workplace exposures were included, which could better illustrate the role between ambient air pollution and AECOPD hospitalization. Another issue is focused on the representativeness of the cases enrolled in our study. The AECOPD subjects admitted by five large hospital in Jinan mainly distributed in the central urban area. According to our preliminary work, most residential quarters lie in the central urban area in Jinan, which implies that then central urban area possess a very high population density and the most population in Jinan, thus the distribution of enrolled patients was rational and our cases could represent the general AECOPD hospitalization in Jinan City. One limitation of the study is daily or weekly concentrations of air pollutants were not gained due to limited monitoring technique in the year of 2009 applied in Jinan City. Lack of information of PM2.5, which was not regularly monitored in the study period, was also a disadvantage in our study. Another limitation is the cross-sectional nature of the study, which could not demonstrate the causal relationship between ambient air pollution and AECOPD hospitalization. As the monitoring technique developing, further study could be done to better illustrate the association between ambient air pollution and admissions due to AECOPD in East China. Prospective cohort study was also needed to illustrate the causal relationship in the future.
Conclusion Ambient air pollution is spatially correlated with AECOPD hospitalizations. At residence, SO2 is of greatest spatial
378 correlation with AECOPD hospitalization, while at workplace, PM10 is of greatest spatial correlation with AECOPD hospitalization. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. As a spatial data processing tool, GIS has novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research.
Acknowledgments The authors would like to thank the Environmental Protection Agency of Shandong Province for providing air pollution data and lists of high-polluting industries, Provincial Hospital Affiliated to Shandong University, Shandong Provincial Qianfoshan Hospital, Affiliated Jinan Central Hospital of Shandong University and Shandong Jiao Tong Hospital for providing the cases required.
Abbreviations AECOPD Acute exacerbations of Chronic Obstructive Pulmonary Disease CO Carbon monoxide GIS Geographical information system NO2 Nitrogen dioxide O3 Ozone PM2.5 Particular matter with aerodynamic diameter less than 2.5 mm PM10 Particular matter with aerodynamic diameter less than 10 mm SO2 Sulfur dioxide
Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.rmed.2015.01.006.
Conflict of interest We declare that there were no conflicts of interest in our article.
W. Wang et al.
References [1] Spencer S, Calverley PM, Burge PS, Jones PW. Impact of preventing exacerbations on deterioration of health status in COPD. Eur Respir J 2004;23(5):698e702. [2] Kessler R, Stahl E, Vogelmeier C, Haughney J, Trudeau E, Lofdahl CG, Partridge MR. Patient understanding, detection, and experience of COPD exacerbations: an observational, interview-based study. Chest 2006;130(1):133e42. [3] Seemungal TA, Donaldson GC, Bhowmik A, Jeffries DJ, Wedzicha JA. Time course and recovery of exacerbations in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;161(5):1608e13. [4] Wouters EF. The burden of COPD in The Netherlands: results from the confronting COPD survey. Respir Med 2003;97(Suppl. C):S51e9. [5] Anzueto A, Sethi S, Martinez FJ. Exacerbations of chronic obstructive pulmonary disease. Proc Am Thorac Soc 2007;4(7): 554e64. [6] Patel IS, Seemungal TA, Wilks M, Lloyd-Owen SJ, Donaldson GC, et al. Relationship between bacterial colonisation and the frequency, character, and severity of COPD exacerbations. Thorax 2002;57(9):759e64. [7] Kanner RE, Anthonisen NR, Connett JE. Lower respiratory illnesses promote FEV(1) decline in current smokers but not exsmokers with mild chronic obstructive pulmonary disease: results from the lung health study. Am J Respir Crit Care Med 2001;164(3):358e64. [8] Ling SH, van Eeden SF. Particulate matter air pollution exposure: role in the development and exacerbation of chronic obstructive pulmonary disease. Int J Chron Obstruct Pulmon Dis 2009;4:233e43. [9] Sint T, Donohue JF, Ghio AJ. Ambient air pollution particles and the acute exacerbation of chronic obstructive pulmonary disease. Inhal Toxicol 2008;20(1):25e9. [10] Peacock JL, Anderson HR, Bremner SA, Marston L, Seemungal TA, et al. Outdoor air pollution and respiratory health in patients with COPD. Thorax 2011;66(7):591e6. [11] Rushton G. Public health, GIS, and spatial analytic tools. Annu Rev Public Health 2003;24:43e56. [12] Naves LA, Porto LB, Rosa JW, Casulari LA. Geographical information system (GIS) as a new tool to evaluate epidemiology based on spatial analysis and clinical outcomes in acromegaly. Pituitary 2015 Feb;18(1):8e15. [13] Anderson HR, Spix C, Medina S, Schouten JP, Castellsague J, et al. Air pollution and daily admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project. Eur Respir J 1997;10(5):1064e71. [14] Ko FW, Tam W, Wong TW, Chan DP, Tung AH, et al. Temporal relationship between air pollutants and hospital admissions for chronic obstructive pulmonary disease in Hong Kong. Thorax 2007;62(9):780e5. [15] Sauerzapf V, Jones AP, Cross J. Environmental factors and hospitalisation for chronic obstructive pulmonary disease in a rural county of England. J Epidemiol Community Health 2009; 63(4):324e8.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/rmed
A GIS-based spatial correlation analysis for ambient air pollution and AECOPD hospitalizations in Jinan, China Wenqiao Wang a,b, Yangyang Ying a, Quanyuan Wu c, Haiping Zhang c, Dedong Ma a,*, Wei Xiao a,* a Department of Respiratory Medicine, Qilu Hospital, Shandong University, No. 107, Wenhua Xi Road, Jinan, Shandong, 250012, PR China b Department of Respiratory Diseases, China-Japan Friendship Hospital, Peking University, Beijing, PR China c College of Population, Resources and Environment, Shandong Normal University, No. 88, Wenhua Dong Road, Jinan, Shandong, 250012, PR China
Received 26 September 2014; accepted 17 January 2015
Available online 27 January 2015
KEYWORDS AECOPD hospitalization; Ambient air pollution; GIS; Spatial autocorrelation
Summary Background: Acute exacerbations of COPD (AECOPD) are important events during disease procedure. AECOPD have negative effect on patients’ quality of life, symptoms and lung function, and result in high socioeconomic costs. Though previous studies have demonstrated the significant association between outdoor air pollution and AECOPD hospitalizations, little is known about the spatial relationship utilized a spatial analyzing technique- Geographical Information System (GIS). Objective: Using GIS to investigate the spatial association between ambient air pollution and AECOPD hospitalizations in Jinan City, 2009. Methods: 414 AECOPD hospitalization cases in Jinan, 2009 were enrolled in our analysis. Monthly concentrations of five monitored air pollutants (NO2 , SO2, PM10, O3 , CO) during January 2009eDecember 2009 were provided by Environmental Protection Agency of Shandong Province. Each individual was geocoded in ArcGIS10.0 software. The spatial distribution of five pollutants and the temporal-spatial specific air pollutants exposure level for each individual was estimated by ordinary Kriging model. Spatial autocorrelation (Global Moran’s I) was employed to explore the spatial association between ambient air pollutants and AECOPD hospitalizations. A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model.
* Corresponding authors. E-mail addresses: [email protected] (W. Wang), [email protected] (Y. Ying), [email protected] (Q. Wu), gissuifeng@ 163.com (H. Zhang), [email protected] (D. Ma), [email protected] (W. Xiao). http://dx.doi.org/10.1016/j.rmed.2015.01.006 0954-6111/ª 2015 Elsevier Ltd. All rights reserved.
Spatial correlation between air pollution and AECOPD hospitalizations
373
Results: At residence, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistical significant spatially clustered. The Z-score of SO2, PM10, CO, O3, NO2 at residence is 15.88, 13.93, 12.60, 4.02, 2.44 respectively, while at workplace, concentrations of PM10, SO2, O3, CO and AECOPD hospitalization cases showed statistical significant spatially clustered. The Z-score of PM10, SO2, O3, CO at workplace is 11.39, 8.07, 6.10, and 5.08 respectively. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations due to AECOPD. Conclusions: Ambient air pollution is correlated with AECOPD hospitalizations spatially. A 10 mg/ m3 increase of PM10 at workplace was associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations due to AECOPD in Jinan, 2009. As a spatial data processing tool, GIS has novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research. ª 2015 Elsevier Ltd. All rights reserved.
Background Acute exacerbations of Chronic Obstructive Pulmonary Disease (AECOPD) are important events during disease procedure. AECOPD have negative effect on patients’ quality of life [1,2], symptoms and lung function [3], and result in high socioeconomic costs [4]. It is estimated that AECOPD would lead to 110,000 deaths and over 500,000 hospitalizations per year, the direct cost due to AECOPD was over 18 billion dollars [5]. The acute exacerbations accelerate the rate of decline of lung function [6,7], correlated with significant mortality, particularly in those requiring hospitalizations (Figs. 1 and 2). Exacerbations of COPD can be triggered by various factors. Though respiratory tract infections appear to be the most common causes, not all patients suffering AECOPD have the evidence of infection. Growing evidence support that ambient air pollution is an environmental triggering factor for acute exacerbation of COPD [8e10]. Geographical information system (GIS) aims to integrate the digital capture, management, analysis and visualization of geographically referenced data spatially. Better interpretation of the patterns, trends and relationships between disease and demography, environment, space and time could be obtained through GIS. Thus GIS has important application in medical and health area, especially in the application of etiological research [11,12]. To our knowledge, few studies employed GIS technique to assess spatio-temporal specific exposure to air pollutants in the spatial association analysis between admissions for AECOPD and air pollution. There is no similar study in East China. We aim to assess the spatial association between ambient air pollution exposure and AECOPD hospitalizations in Jinan City through GIS.
Methods Study area & study period We set the study in Jinan City, 2009. Jinan is the capital of Shandong Province in Eastern China. It is located in the
north-western part of Shandong province at 36 010 e37 320 northern latitude and 116 110 e117 440 east of Greenwich.
Target population All cases enrolled in our study were from five large-scale hospitals interspersed in Jinan City. COPD hospitalization cases met with the following inclusion criteria were included in our study: (1) Hospitalization due to acute exacerbations of COPD, identified by International Statistical Classification of Diseases, 10th Revision (ICD-10) codes, J40-J44; (2) Resided and worked in study area (Jinan City) during study period (Jan 2009eDec 2009); (3) Adults patients (age > 18 years). Exclusion criteria: (1) Patients who did not reside or work in Jinan City during 2009. (2) In order to avoid the impact of occupational exposure to air pollutants, patients who worked at highpolluting environment were excluded. Written informed consent was obtained prior to data collection. The study and consent procedure was approved by the Ethics Committee of Qilu Hospital of Shandong University (No. KYLL-2014-4).
Exposure assessment Subjects were geocoded using home residence and working site respectively in a Geographical Information System. We applied addresses geocoding techniques using ArcGIS 10.0 software. Each subject was shown on a map as a precise mark in correspondence with his/her home residence or working site respectively. For each marking site, concentrations of air pollutants were estimated using ordinary Kriging method to assess spatiotemporal-specific air pollution exposure. The monthly average concentrations of air pollutants from
374
W. Wang et al.
Figure 1 Distribution of SO2 concentration and AECOPD hospitalization (at residence). An example of the estimated monthly air pollution levels using monitoring data and Kriging method, distribution of residence of AECOPD hospitalization. Take SO2 concentration distribution and admission data of April, 2009 in Jinan, China as an example.
January, 2009 to December, 2009 were provided by Environmental Protection Agency of Shandong Province. The air pollution concentrations, i.e. nitrogen dioxide (NO2), ozone (O3), sulfur dioxide (SO2), particulate matter with aerodynamic diameter less than 10 microns (PM10), carbon monoxide (CO), were monitored in 15 monitoring stations interspersed in different sub districts of Jinan City.
Statistical analysis Global Moran’s I statistic was used for spatial autocorrelation analysis in ArcGIS 10.0. Moran’s I, p value and Z score
were calculated to test the spatially clustered tendency between concentrations of air pollutants and AECOPD hospitalization cases. Confidence level of 99% was selected. Values of P < 0.01 were considered statistically significant. A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model. Monthly numbers of AECOPD admissions were selected as dependent variable. Age, gender, season of hospitalization, smoking status and air pollutants concentrations at residence and workplace were selected as dependent variables. After controlling for the confounding
Spatial correlation between air pollution and AECOPD hospitalizations
375
Figure 2 Distribution of PM10 concentration and AECOPD hospitalization (at workplace). An example of the estimated monthly air pollution levels using monitoring data and Kriging method, distribution of residence of AECOPD hospitalization. Take PM10 concentration distribution and admission data of April, 2009 in Jinan, China as an example.
effects of seasonality, age, gender and smoking status, monthly concentrations of PM10, SO2, CO, NO2, O3 at residence and workplace for each individual were added to the core model to determine the increase of COPD admissions for a 10 mg/m3 increase in each of these air pollutants. An interaction effect between gender (male or female) and smoking status (ex-smoker or non-smoker or current smoker) were analyzed in the model. Values of P < 0.05 were considered statistically significant. All the analysis were performed with SPSS 21.0 software.
Results Demographic information Altogether 414 admissions for acute exacerbations of COPD during January 1，2009eDecember 31，2009 were enrolled in our study. The basic demographic information of enrolled patients (such as age, gender, smoking status, allergy history and comorbidities) were shown in Table 1.
376
W. Wang et al.
Table 1 Characteristics of the whole subjects enrolled in the study.
Table 3 Spatial autocorrelation analysis between air pollutants at workplace and AECOPD hospitalization cases.
Variables
Air pollutants
Moran’s I
Z score
P value
SO2 NO2 PM10 CO O3
0.24 0.04 0.45 0.19 0.21
8.07 0.52 11.39 5.08 6.10
0.00a 0.60a 0.00a 0.00a 0.00a
Items
AECOPD hospitalization cases (n Z 414)
Age (mean SD) Gender (n, %) Male Female Smoking status Non-smoker Ex-smoker Current-smoker Allergy history Yes No Comorbidities Chronic cor pulmonale Ischemic heart disease Heart failure Hypertension Atrial fibrillation Lung cancer Diabetes Cerebrovascular disease
73.4 9.6 245 (59.2%) 169 (40.8%) 142 (34.3%) 138 (33.3%) 134 (32.4%) 143 (34.5%) 271 (65.5%) 255 (61.6%) 203 (49.0%) 55 (13.3%) 158 (38.2%) 23 (5.6%) 16 (3.9%) 50 (12.1%) 39 (9.4%)
a
Table 4 Spatial autocorrelation analysis between air pollutants at residence and AECOPD hospitalization cases (non-smokers þ ex-smokers). Air pollutants
Moran’s I
Z code
P value
SO2 NO2 PM10 CO O3
0.43 0.07 0.51 0.41 0.17
9.10 2.42 17.06 11.27 3.16
0.00a 0.07 0.00a 0.00a 0.00a
a
Spatial correlation at residence At residence, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered (Table 2). The Moran’s I of SO2, PM10, CO, O3, and NO2 at residence is 0.62, 0.43, 0.40, 0.21 and 0.14, respectively. The Z-score of SO2, PM10, CO, O3, and NO2 at residence is 15.88, 13.93, 12.60, 4.02 and 2.44, respectively. Of the five air pollutants, SO2 is of greatest spatial correlation with AECOPD hospitalization. We conducted spatial correlation in non-smokers plus ex-smokers, concentrations of SO2, PM10, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered at residence (Table 4).
Spatial correlation at workplace At workplace, concentrations of PM10, SO2, O3, CO and AECOPD hospitalizations cases showed spatially clustered tendency with statistical significance. The Z-score of PM10, SO2, O3, CO at workplace is 11.39, 8.07, 6.10, and 5.08 respectively. Concentrations of NO2 and AECOPD hospitalization cases did not show spatially clustered tendency (p Z 0.60). Of the four air pollutants, PM10 is of the greatest spatial correlation with AECOPD hospitalization (Table 3).
Data with statistical significance.
Data with statistical significance.
We conducted spatial correlation in non-smokers plus ex-smokers, concentrations of SO2, PM10, NO2, CO, O3 and AECOPD hospitalization cases showed statistically significant spatially clustered at workplace (Table 5).
Predicting model A generalized linear model (GLM) using a Poisson distribution with log-link function was used to construct a core model. Results of the core model were shown in Table 6 (Supplemental Materials). After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. Other pollutants showed no statistically significance. As to other potential confounders included in the model, the season of AECOPD hospitalization (cold season) and gender (male) were proved to be independent risk factors for AECOPD hospitalization. Gender and smoking status have interaction effect on AECOPD hospitalization, the interaction showed that in male ex-smokers, the incidence
Table 2 Spatial autocorrelation analysis between air pollutants at residence and AECOPD hospitalization cases.
Table 5 Spatial autocorrelation analysis between air pollutants at workplace and AECOPD hospitalization cases (non-smokers þ ex-smokers).
Air pollutants
Moran’s I
Z score
P value
Air pollutants
Moran’s I
Z code
P value
SO2 NO2 PM10 CO O3
0.62 0.14 0.43 0.40 0.21
15.88 2.44 13.93 12.60 4.02
0.00a 0.00a 0.00a 0.00a 0.000058a
SO2 NO2 PM10 CO O3
0.32 0.11 0.68 0.28 0.09
6.05 2.72 15.62 7.25 3.47
0.00a 0.00a 0.00a 0.00a 0.00a
a
Data with statistical significance.
a
Data with statistical significance.
Spatial correlation between air pollution and AECOPD hospitalizations of AECOPD hospitalizations decreased, while other combinations showed no statistical significance.
Discussion Our study assessed the spatial correlation of ambient air pollution related exposure on AECOPD admissions in Jinan City, China, 2009, and the results indicate that ambient air pollution is correlated with AECOPD hospitalizations spatially. At residence, SO2 is of the greatest spatial correlation with AECOPD hospitalization, while at workplace, PM10 is of the greatest spatial correlation with AECOPD hospitalization. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. Other pollutants showed no statistically significance. In recent years, air pollution has become a worldwide environmental issue, and China is facing the greatest challenge from deteriorating air quality. It is of great importance and urgency to find out the definite impact of air pollution. Jinan is one of the most polluted city in China, our research proved the spatial correlation between ambient air pollution and AECOPD hospitalizations in Jinan, 2009, which provided novel evidence and new direction for triggering factors and prevention of AECOPD. Though previous studies have demonstrated some kind of associations between ambient air pollution and COPD hospitalizations, publishing data were inconsistent and even conflicting, and little knowledge was known on the spatial correlation between ambient air pollution and AECOPD hospitalization cases. More evidence is needed to demonstrate the solid association between ambient air pollution and AECOPD hospitalizations. In the association study between O3 and COPD admissions, several previous studies have demonstrated the association. In APHEA project conducted in Europe, O3 was associated with daily admissions for COPD, the relative risks for a 50 mg/m3 increase in daily mean concentrations of O3 was 1.04 (1.02, 1.07) [13]. In the research conducted in Hong Kong, associations between O3 and COPD hospitalizations were convinced both in single-pollutant model and multipollutant model [14]. But in the study conducted by Sauerzapf V et al. [15], no associations were observed between O3 and admissions for COPD. Similar conditions exist in the association analysis between SO2 and COPD hospitalizations. In the research conducted in Hong Kong, associations between SO2 and COPD hospitalizations were convinced both in single-pollutant model and multi-pollutant model [14]. But in APHEA project, SO2 did not show a significant association with admissions for COPD (RR Z 1.02, 95%CI [0.98, 1.06]) [13]. Current studies support the role of NO2 in COPD hospitalizations. In APHEA project, NO2 was associated with daily admissions for COPD, the relative risks for a 50 mg/m3 increase in daily mean concentrations of NO2 was 1.02 (1.00, 1.05) [13]. In the study conducted in a rural county of England, each 10 mg/m3 increase in NO2 was associated with a 22% increase in the odds of COPD admission [15]. The role of PM10 in COPD hospitalizations was inconclusive. In a research conducted in Hong Kong,
377
PM10 was associated with COPD hospitalizations in single pollutant model, with RR Z 1.024 for admissions per 10 mg/ m3 increase, though this effect became unstatistically significant in multi-pollutant model [14]. Data were relatively limited referring to the role of CO in COPD hospital admissions. In the study conducted by Sauerzapf V et al. [15], 10 mg/m3 increase in CO was associated with a 2% increase in the odds of COPD admission. Utilizing spatial analysis in GIS, our results proved the spatial correlation between ambient air pollution and AECOPD hospitalization, which added new evidence to the association (especially spatial association) between ambient air pollution and AECOPD hospitalization. As a spatial data processing tool, GIS showed novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research. Most previous studies analyzed the relationship between daily admissions and daily air pollutants levels. Since the locations of patients were interspersed, it was much better to use individualized exposure rather than the mean levels of the whole city. Few studies have taken the individualized exposure into consideration in the association analysis between ambient air pollution and AECOPD admissions, especially using GIS for exposure assessment and spatial analysis. To our knowledge, there is no similar study in East China. Our study laid a solid foundation for the research of ambient air pollution and COPD hospitalizations in East China. Exposure to ambient air pollution is unavoidable. Residence exposures as well as workplace exposures are the main source of exposures in the majority. In our study, both residence exposures and workplace exposures were included, which could better illustrate the role between ambient air pollution and AECOPD hospitalization. Another issue is focused on the representativeness of the cases enrolled in our study. The AECOPD subjects admitted by five large hospital in Jinan mainly distributed in the central urban area. According to our preliminary work, most residential quarters lie in the central urban area in Jinan, which implies that then central urban area possess a very high population density and the most population in Jinan, thus the distribution of enrolled patients was rational and our cases could represent the general AECOPD hospitalization in Jinan City. One limitation of the study is daily or weekly concentrations of air pollutants were not gained due to limited monitoring technique in the year of 2009 applied in Jinan City. Lack of information of PM2.5, which was not regularly monitored in the study period, was also a disadvantage in our study. Another limitation is the cross-sectional nature of the study, which could not demonstrate the causal relationship between ambient air pollution and AECOPD hospitalization. As the monitoring technique developing, further study could be done to better illustrate the association between ambient air pollution and admissions due to AECOPD in East China. Prospective cohort study was also needed to illustrate the causal relationship in the future.
Conclusion Ambient air pollution is spatially correlated with AECOPD hospitalizations. At residence, SO2 is of greatest spatial
378 correlation with AECOPD hospitalization, while at workplace, PM10 is of greatest spatial correlation with AECOPD hospitalization. After adjusting for potential confounders in the model, only the PM10 concentrations at workplace showed statistical significance, with a 10 mg/m3 increase of PM10 at workplace associated with a 7% (95%CI: [3.3%, 10%]) increase of hospitalizations for acute exacerbations of COPD. As a spatial data processing tool, GIS has novel and great potential on air pollutants exposure assessment and spatial analysis in AECOPD research.
Acknowledgments The authors would like to thank the Environmental Protection Agency of Shandong Province for providing air pollution data and lists of high-polluting industries, Provincial Hospital Affiliated to Shandong University, Shandong Provincial Qianfoshan Hospital, Affiliated Jinan Central Hospital of Shandong University and Shandong Jiao Tong Hospital for providing the cases required.
Abbreviations AECOPD Acute exacerbations of Chronic Obstructive Pulmonary Disease CO Carbon monoxide GIS Geographical information system NO2 Nitrogen dioxide O3 Ozone PM2.5 Particular matter with aerodynamic diameter less than 2.5 mm PM10 Particular matter with aerodynamic diameter less than 10 mm SO2 Sulfur dioxide
Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.rmed.2015.01.006.
Conflict of interest We declare that there were no conflicts of interest in our article.
W. Wang et al.
References [1] Spencer S, Calverley PM, Burge PS, Jones PW. Impact of preventing exacerbations on deterioration of health status in COPD. Eur Respir J 2004;23(5):698e702. [2] Kessler R, Stahl E, Vogelmeier C, Haughney J, Trudeau E, Lofdahl CG, Partridge MR. Patient understanding, detection, and experience of COPD exacerbations: an observational, interview-based study. Chest 2006;130(1):133e42. [3] Seemungal TA, Donaldson GC, Bhowmik A, Jeffries DJ, Wedzicha JA. Time course and recovery of exacerbations in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;161(5):1608e13. [4] Wouters EF. The burden of COPD in The Netherlands: results from the confronting COPD survey. Respir Med 2003;97(Suppl. C):S51e9. [5] Anzueto A, Sethi S, Martinez FJ. Exacerbations of chronic obstructive pulmonary disease. Proc Am Thorac Soc 2007;4(7): 554e64. [6] Patel IS, Seemungal TA, Wilks M, Lloyd-Owen SJ, Donaldson GC, et al. Relationship between bacterial colonisation and the frequency, character, and severity of COPD exacerbations. Thorax 2002;57(9):759e64. [7] Kanner RE, Anthonisen NR, Connett JE. Lower respiratory illnesses promote FEV(1) decline in current smokers but not exsmokers with mild chronic obstructive pulmonary disease: results from the lung health study. Am J Respir Crit Care Med 2001;164(3):358e64. [8] Ling SH, van Eeden SF. Particulate matter air pollution exposure: role in the development and exacerbation of chronic obstructive pulmonary disease. Int J Chron Obstruct Pulmon Dis 2009;4:233e43. [9] Sint T, Donohue JF, Ghio AJ. Ambient air pollution particles and the acute exacerbation of chronic obstructive pulmonary disease. Inhal Toxicol 2008;20(1):25e9. [10] Peacock JL, Anderson HR, Bremner SA, Marston L, Seemungal TA, et al. Outdoor air pollution and respiratory health in patients with COPD. Thorax 2011;66(7):591e6. [11] Rushton G. Public health, GIS, and spatial analytic tools. Annu Rev Public Health 2003;24:43e56. [12] Naves LA, Porto LB, Rosa JW, Casulari LA. Geographical information system (GIS) as a new tool to evaluate epidemiology based on spatial analysis and clinical outcomes in acromegaly. Pituitary 2015 Feb;18(1):8e15. [13] Anderson HR, Spix C, Medina S, Schouten JP, Castellsague J, et al. Air pollution and daily admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project. Eur Respir J 1997;10(5):1064e71. [14] Ko FW, Tam W, Wong TW, Chan DP, Tung AH, et al. Temporal relationship between air pollutants and hospital admissions for chronic obstructive pulmonary disease in Hong Kong. Thorax 2007;62(9):780e5. [15] Sauerzapf V, Jones AP, Cross J. Environmental factors and hospitalisation for chronic obstructive pulmonary disease in a rural county of England. J Epidemiol Community Health 2009; 63(4):324e8.
