Abstract
Rationale: Little is known about the long-term effects of air pollution on pneumonia hospitalization in the elderly.
Objectives: To assess the effect of long-term exposure to ambient nitrogen dioxide, sulfur dioxide, and fine particulate matter with diameter equal to or smaller than 2.5 μm (PM2.5) on hospitalization for community-acquired pneumonia in older adults.
Methods: We used a population-based case–control study in Hamilton, Ontario, Canada. We enrolled 345 hospitalized patients aged 65 years or more for community-acquired pneumonia and 494 control participants, aged 65 years and more, randomly selected from the same community as cases from July 2003 to April 2005. Health data were collected by personal interview. Annual average levels of nitrogen dioxide, sulfur dioxide, and PM2.5 before the study period were estimated at the residential addresses of participants by inverse distance weighting, bicubic splined and land use regression methods and merged with participants' health data.
Measurements and Main Results: Long-term exposure to higher levels of nitrogen dioxide and PM2.5 was significantly associated with hospitalization for community-acquired pneumonia (odds ratio [OR], 2.30; 95% confidence interval [CI], 1.25 to 4.21; P = 0.007 and OR, 2.26; 95% CI, 1.20 to 4.24; P = 0.012, respectively, over the 5th–95th percentile range increase of exposure). Sulfur dioxide did not appear to have any association (OR, 0.97; 95% CI, 0.59 to 1.61; P = 0.918). Results were somewhat sensitive to the choice of methods used to estimate air pollutant levels at residential addresses, although all risks from nitrogen dioxide and PM2.5 exposure were positive and generally significant.
Conclusions: In older adults, exposure to ambient nitrogen dioxide and PM2.5 was associated with hospitalization for community-acquired pneumonia.
There are sparse data on long-term effects of air pollution on pneumonia hospitalization in the elderly. Moreover, previous studies have focused on the effect of short-term increases in air pollution.
We found that long-term ambient levels of NO2 and fine particulate matter with diameter equal to or smaller than 2.5 μm (PM2.5) were independently associated with pneumonia hospitalization in older adults.
The role of air pollution as a risk factor for respiratory diseases such as asthma and chronic pulmonary disease has been recognized (7–14). Less is known, however, about the effect of air pollution on pneumonia hospitalization in older adults, a high-risk population for this illness. Data from animal models support a deleterious effect of air nitrogen dioxide such as its capacity to impair the function of alveolar macrophages and epithelial cells, thereby increasing the risk of lung infections such as influenza, which can predispose to causative bacterial agents of pneumonia such as Pneumococcus (15). Previous epidemiologic reports are, however, limited by use of administrative databases for diagnosis of pneumonia (7–11, 14), lack of individual-level covariates (7–8, 10), and the combination of pneumonia with other respiratory outcomes (7–8). Moreover, these studies have focused on the effect of short-term increases in air pollution. Although one study evaluated the long-term effect of particulate matter of diameter 10 μm or smaller (PM10) on the survival of hospitalized patients discharged alive after chronic pulmonary obstructive disease (16), there are sparse data on the long-term effects of exposures to higher levels of air pollution on hospitalization with community-acquired pneumonia.
We hypothesized that long-term exposures (defined as 1 to 2 yr) to higher levels of air pollution, specifically nitrogen dioxide (NO2), and sulfur dioxide (SO2), as well as fine particulate matter with diameter equal to or smaller than 2.5 μm (PM2.5), would increase the risk of hospitalization with community-acquired pneumonia in persons 65 years and older. To test this hypothesis, we conducted a case–control study in Hamilton, Ontario, Canada from 2003 to 2005, where air pollution exposures levels in 2001–2002 at the residential addresses of case and control subjects were estimated by inverse distance weighting, bicubic splined, and land use regression methods.
There are four emergency departments of hospitals in Hamilton, Canada serving a catchment area of approximately 2.2 million people. Eligible participants aged 65 years or older had to reside in the catchment area community (defined using forward sortation areas) and present to the emergency department with at least two of the following signs and symptoms: temperature greater than 38°C, productive cough, chest pain, shortness of breath, or crackles on auscultation. They also had to have a new opacity on a chest radiograph that was interpreted by a radiologist as being compatible with pneumonia. In the community, the majority of suspected cases of pneumonia do not have radiographs obtained. However there is potential for misclassification of nonpneumonia cases as cases unless radiographs are obtained to confirm the pneumonia. Therefore, to avoid misclassification bias, we selected subjects with radiologically confirmed pneumonia who were admitted to hospitals. Exclusion criteria included residence in a nursing home and infection at another body site in addition to community-acquired pneumonia. We did not consider nursing home residences because these institutionalized elderly usually have an etiology of pneumonia different from that of community-acquired pneumonia. We excluded patients with infections other than pneumonia to reduce their confounding effect on the association between air pollution and pneumonia. For each case, we selected one control subject aged 65 years and older from the same catchment area, using the modified Mitofsky-Waksberg method of random-digit dialing (17). Control subjects were selected on a contemporaneous basis with case subjects so as to minimize potential bias due to differential environmental exposure. That is, similar numbers of case and control subjects were enrolled during each month of the study. Control participants were excluded if they had been diagnosed with pneumonia in the past 12 months or if they had any other active infection.
Participants were enrolled from July 2003 to April 2005. The study was approved by the McMaster University (Hamilton, ON, Canada) Research Ethics Board. Informed consent was obtained from each participant.
Trained interviewers used structured questionnaires to collect data from participants in the emergency department or in the hospital if the patient was admitted. The same questionnaires were administered by telephone to community control subjects. Information on variables that could be potential confounders was collected. These included age, sex, highest level of education obtained (an indicator of socioeconomic status in this elderly population, which might not have active employment and income), smoking history (elicited by asking whether the participant had smoked 100 or more cigarettes in his or her lifetime, using an item from Canada's National Population Health Survey [18]), and history of occupational exposure (elicited by asking whether the participant had been regularly exposed to gases, fumes, or chemicals at work). We also collected information on chronic obstructive lung disease and functional status to examine them as effect modifiers. Functional status was measured by using a modification of the Barthel Index (19), a scale that evaluates independence on 10 daily functions including bowel and bladder continence, toilet use, dressing, and bathing. On this scale, 0 indicates full dependence and 20 indicates complete independence.
In Hamilton, industry is the main source of SO2. NO2 has transportation as the largest source, with a major contribution from industrial emissions. Industry accounts for over half of directly-emitted airborne particles (PM2.5), with the remainder coming from heating and transportation. SO2 can transform into sulphate particles and NO2 into nitrate particles, which also contribute to ambient concentrations of PM2.5. Air pollution data used in this study were collected at six air quality–monitoring stations operated by the Ontario Ministry of Environment throughout the Hamilton metropolitan area for gaseous pollutants (NO2 and SO2) and at four stations for PM2.5. The concentration of gaseous pollutants NO2 and SO2 in air were measured as parts per billion (ppb) and dust particles (PM2.5) were measured as micrograms per cubic meter of air (μg/m3).
Averages were calculated from daily data, which were included only if valid readings were available for at least 18 hours of the day. Estimates of the long-term exposure data of all three air pollutants at the residential addresses of case and control subjects were predicted using two deterministic interpolators, namely, the bicubic splines method and inverse distance weighting functions based on one over distance squared weights (20). Ambient levels of NO2 at the residential addresses of participants were also predicted, using the land use regression method (21, 22). The interpolations using bicubic splined and inverse distance weighting methods were based on the 2-year (2001 and 2002) average for NO2 and SO2, and a 1-year (2002) average for PM2.5. Spatial variations in NO2 levels as a marker of traffic-related air pollution were extensively measured across Hamilton in 2002–2004 (22, 23). NO2 samples were collected over 2 weeks in the early autumn of 2002, using duplicate two-sided Ogawa passive diffusion samplers at 107 locations across the city. Sampling locations were selected with a location–allocation model, summarized elsewhere, that uses pollution variability over space and a residential population density of persons 65 and older to optimize sampling information for subsequent analyses (21). There were valid readings for 100 locations. Another 30 locations in Hamilton were monitored in the spring of 2004 to assess seasonality of the spatial pattern. These results showed that the pattern was stable over time, and thus only one exposure model developed from the autumn 2002 data was sufficient to capture the spatial gradient in NO2 exposure. The 100 readings from 2002 were used as the dependent variable in a land use regression model (22). More than 85 predictor variables representing land use, roads, traffic, physical geography, and population density were screened. The final model included seven predictors that explained 76% of the variation in NO2 (22).
Predicted values of air pollution levels using different methods of estimation at the postal codes of the case and control subjects provided detailed assessments of long-term exposures to air pollution near their residences. These air pollutant data were merged with other health data of the participants collected during the study period.
For the statistical analysis, we defined outcome = 1 for cases and outcome = 0 for controls. Air pollution variables NO2, SO2, and PM2.5 and age were analyzed as continuous variables. The variables male sex, chronic lung disease (chronic obstructive pulmonary disease), history of smoking, and occupational exposure (history of regular exposure to gases, fumes, and chemical at work) were analyzed as binary variables. Level of education was dichotomized as less than a high school education (for lower socioeconomic status) versus high school or more. Barthel Index scores were highly skewed, with most participants scoring high values. Therefore, instead of considering it a continuous variable, we dichotomized the Barthel Index by defining the scores 18 to 20 as better functional status (somewhat to no dependency) and scores 17 and lower as lower functional status (somewhat to highly dependent).
We were cognizant of the fact that the traditional biomedical variables could overwhelm the analysis and make it impossible to detect subtle but important effects of air pollution. Therefore, our approach for the analysis was to, a priori, select key demographic, socioeconomic (education as a proxy measure), and occupational variables for adjustment because they are known confounders for exposure to air pollution. For each of the pollutant variables NO2, SO2, and PM2.5 derived by each method of measurement, a separate logistic regression model was specified a priori that included that particular air pollution variable and the covariates age, sex, education, smoking history, and history of exposure to gases, fumes, or chemical at work. Moreover, we decided a priori to adjust for the interaction effect of any of these preselected covariates and chronic lung disease and functional status in the adjusted analysis if that variable had significant interaction with the air pollutant variable in univariate analysis. Unlike other covariates selected a priori for adjustment, we decided not to adjust for other chronic lung diseases, functional status, and other comorbidities as independent risk factors because these factors would account for such large variances in the outcome that it would be difficult to observe small but important effects of air pollution at the population level.
Crude associations of chronic lung disease, functional status, and covariates specified a priori and each of the air pollutant variables with hospitalization for community-acquired pneumonia were obtained using chi-square tests for binary variables and t tests for continuous variables. The interaction effect of each of the other variables with each of the air pollutant variables was tested by using a logistic regression model, including the variable and the air pollutant variable and their interaction term in the models. Ultimately, we did not keep any interaction term of any variables in the final models because none of them appeared to be effect modifiers. In the adjusted analysis, the odds ratio (OR) and corresponding 95% confidence interval (CI) for each of the air pollutants were obtained for the 5th–95th percentile range increment by using the air pollutant variable changed in scale by dividing by corresponding 5th–95th percentile range. We reported results using the data derived by the inverse distance weighting method as our primary analysis and discussion and also reported results using the data derived by bicubic splined and land use regression methods to assess consistency.
Data were analyzed with SAS for Windows version 9.1 (SAS Institute Inc., Cary, NC).
Of participants assessed as possible case subjects for the study, reasons for nonenrollment included the following: 1,536 did not meet clinical eligibility criteria, 1,457 were from nursing homes, 519 did not meet radiological eligibility criteria, 856 had a suspected infection other than pneumonia, 888 either died or were discharged before enrollment, 702 were from outside the catchment regions, and 136 refused to participate. For participants to be assessed as potential control subjects for the study, reasons for nonenrollment included the following: there was no one aged 65 years or older in 2,527 households, 1,295 random digit dialings had incorrect or unavailable telephone numbers, 380 elderly refused to participate, 184 had language or hearing difficulties, and 6 were in nursing homes. Finally, 365 case subjects and 494 control subjects were enrolled in the study.
Figures 1 and 2 show the case and control subjects overlaid on the pollution surfaces for PM2.5 and NO2. In Figure 1, PM2.5 displays a smooth exposure surface with higher levels in the northeast, near the industrial zone and below the Niagara escarpment. The exposure assigned to case and control subjects appears to capture much of the range in the exposure surface. The exposure gradient for PM2.5 across Hamilton is quite large compared with other cities, probably because of the presence of a large industrial steel-making complex along the lakeshore at the north end of the city. Figure 2 shows the case and control subjects overlaid on the NO2 surface from the land use regression model. As illustrated, there is a much finer scale spatial pattern, influenced by nearby source emissions from traffic and to a lesser extent the industrial area in the north end.
The distributions (mean, percentiles, and ranges) of ambient air pollutants (NO2, SO2, and PM2.5) measured by various methods are shown in Table 1. The descriptive statistics of age, sex, smoking, education, and history of exposure to gases, fumes, or chemicals at work, and ambient air pollutions, and the results of the univariate analysis are shown in Table 2. In the univariate analysis, it was observed that long-term exposures to higher NO2 levels measured by all three methods, SO2 levels measured by the bicubic spline method, and PM2.5 levels measured by the inverse distance weighting method were associated with hospitalization with community-acquired pneumonia at the 5% significance level. None of the variables—chronic lung diseases, functional status and other covariates (age, sex, educational level, smoking, and regular exposure to gases, fumes, or chemicals at works)—significantly interacted with any air pollutant variable derived by any method on hospitalization with community-acquired pneumonia. We therefore did not include any interaction term of these variables in the logistic regression model for any air pollution variable in the multivariable analysis.
Percentile | 25th−75th percentile (range) | 5th−95th percentile (range) | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Variable† | Mean | Minimum | 5th | 25th | 50th | 75th | 95th | Maximum | ||||||||
| NO2, ppb (SPL) | 18.68 | 12.47 | 13.20 | 15.72 | 18.52 | 21.18 | 24.79 | 27.44 | 5.46 | 11.59 | ||||||
| NO2, ppb (IDW) | 20.14 | 14.93 | 16.11 | 19.21 | 19.87 | 21.35 | 23.27 | 26.87 | 2.14 | 7.16 | ||||||
| NO2, ppb (LUR) | 15.00 | 8.79 | 10.49 | 13.62 | 14.98 | 16.53 | 18.53 | 25.67 | 2.91 | 8.04 | ||||||
| SO2, ppb (SPL) | 4.65 | 2.27 | 2.46 | 3.31 | 4.62 | 5.66 | 7.60 | 10.43 | 2.35 | 5.14 | ||||||
| SO2, ppb (IDW) | 5.80 | 3.33 | 4.61 | 5.29 | 5.67 | 6.32 | 7.17 | 10.64 | 1.03 | 2.56 | ||||||
| PM2.5, μg/m3 (SPL) | 9.51 | 6.01 | 7.12 | 8.12 | 9.23 | 10.85 | 12.64 | 13.41 | 2.73 | 5.52 | ||||||
| PM2.5, μg/m3 (IDW) | 10.70 | 8.94 | 9.27 | 10.18 | 10.57 | 11.25 | 12.39 | 12.97 | 1.07 | 3.12 | ||||||
Case Subjects (n = 365) [Count (%) or Mean (SD)]‡ | Control Subjects (n = 494) [Count (%) or Mean (SD)]‡ | Case vs. Control | |||||
|---|---|---|---|---|---|---|---|
| Variable† | OR or MD§ | 95% CI | P Value | ||||
| Age, yr | 79.8 (7.7) | 74.6 (6.8) | 5.2 | 4.3, 6.2 | <0.001 | ||
| Male sex | 231 (63) | 156 (32) | 3.74 | 2.81, 4.97 | <0.001 | ||
| Education (<high school) | 242 (68) | 217 (44) | 2.70 | 2.03, 3.60 | <0.001 | ||
| Smoking (≥100 cigarettes) | 265 (74) | 244 (49) | 2.92 | 2.17, 3.92 | <0.001 | ||
| History of regular exposure to gases, fumes, or chemicals at work | 193 (55) | 85 (17) | 5.78 | 4.22, 7.91 | <0.001 | ||
| NO2, ppb (SPL) | 19.23 (3.6) | 18.26 (3.5) | 0.97 | 0.49, 1.47 | <0.001 | ||
| NO2, ppb (IDW) | 20.40 (2.2) | 19.94 (2.0) | 0.45 | 0.17, 0.74 | 0.002 | ||
| NO2, ppb (LUR) | 15.25 (2.7) | 14.81 (2.5) | 0.44 | 0.09, 0.80 | 0.013 | ||
| SO2, ppb (SPL) | 4.78 (1.6) | 4.55 (1.6) | 0.23 | 0.02, 0.45 | 0.036 | ||
| SO2, ppb (IDW) | 5.84 (0.8) | 5.78 (0.9) | 0.06 | −0.06, 0.18 | 0.300 | ||
| PM2.5, μg/m3 (SPL) | 9.62 (1.8) | 9.43 (1.7) | 0.19 | −0.05, 0.43 | 0.125 | ||
| PM2.5, μg/m3 (IDW) | 10.77 (0.9) | 10.65 (0.8) | 0.13 | 0.01, 0.24 | 0.036 | ||
The results of the multivariable analyses are displayed in Table 3 estimated by fitting a separate model of each of the air pollutant variables. NO2 was significantly associated with pneumonia hospitalization: OR, 2.30 (95% CI, 1.25 to 4.21), P = 0.007. Similarly, PM2.5 (OR, 2.26; 95% CI, 1.20 to 4.24; P = 0.012) was also associated with pneumonia hospitalization, whereas SO2 did not appear to have any association (OR, 0.97; 95% CI, 0.59 to 1.61; P = 0.918). NO2 estimated by the bicubic spline method (OR, 2.19; 95% CI, 1.25 to 3.83; P = 0.006) and the land use regression method (OR, 1.70; 95% CI, 1.00 to 2.89; P = 0.049) also had large, significant associations with hospital admissions. SO2 and PM2.5 estimated by the spline interpolator had ORs of 1.09 (95% CI, 0.63 to 1.89; P = 0.766) and 1.70 (95% CI, 0.99 to 2.92; P = 0.053), respectively.
Air Pollution Variables† | OR‡ | 95% CI‡ | P Value |
|---|---|---|---|
| NO2, ppb (IDW) | 2.30 | 1.25, 4.21 | 0.007 |
| NO2, ppb (SPL) | 2.19 | 1.25, 3.83 | 0.006 |
| NO2, ppb (LUR) | 1.70 | 1.00 - 2.89 | 0.049 |
| SO2, ppb (IDW) | 0.97 | 0.59, 1.61 | 0.918 |
| SO2, ppb (SPL) | 1.09 | 0.63, 1.89 | 0.766 |
| PM2.5, ppb (IDW) | 2.26 | 1.20, 4.24 | 0.012 |
| PM2.5, ppb (SPL) | 1.70 | 0.99, 2.92 | 0.053 |
All covariates selected a priori for adjustment were significantly associated with community-acquired pneumonia hospitalization in each logistic regression model for each of the air pollutant variables. Strength of association of these variables did not change on the basis of the model used for derivation of each type of air pollution. In the NO2 model derived by the inverse distance weighting method, the estimates of associations for age in 5-year increments was OR, 1.84 (95% CI, 1.62 to 2.08), P < 0.001; for male sex it was OR, 1.89 (95% CI, 1.27 to 2.82), P = 0.002; for a less than high school education it was OR, 2.08 (95% CI, 1.47 to 2.94), P < 0.001; for a history of smoking of 100 or more cigarettes it was OR, 2.44 (95% CI, 1.65 to 3.60), P < 0.001; and for a history of regular exposure to gases, fumes, or chemicals at work it was OR, 4.44 (95% CI, 2.95 to 6.70), P < 0.001. We found similar estimates of association for these variables in other models.
In this case–control study, we found that long-term exposure to higher levels of ambient air pollution (NO2 and PM2.5) were independently associated with pneumonia hospitalization of older adults when the ambient levels of pollution variables were derived by the inverse distance weighting method. Such associations were also observed for NO2 when estimating the air pollutant levels by the bicubic spline interpolator and land use regression methods. It was found that the results for NO2 and PM2.5 were slightly sensitive to the choice of the models for deriving ambient pollution levels. The air pollutants in this study were assessed as exposures over the previous 12 months to test the chronic effect rather than an acute exacerbation of symptoms. The epidemiologic effect we found is supported by data from animal models where long-term exposure to NO2 reduced macrophage ability (15, 24).
Our findings with respect to long-term exposure to higher levels of ambient air pollutants are consistent with the results from Taiwanese studies (11, 14). These studies, where medical claims information was used to obtain diagnostic codes for pneumonia, demonstrated associations of short-term increase in ambient NO2 and PM10 levels with increased risk of hospital admissions for pneumonia in older adults. Our findings are also consistent with studies that assessed short-term exposure to PM10 and found a positive association with hospitalization with pneumonia (6, 9, 25–29). A short-term effect of PM2.5 on pneumonia hospitalization was also observed in a large case-crossover study in the United States (30), whereas another study from Taiwan did not find any such association (31).
We did not find any association of long-term exposure to higher SO2 levels on pneumonia hospitalization. This finding agrees with studies (11, 14, 31, 32) in which short-term exposure to a higher level of SO2 did not have any effect on pneumonia hospitalization. In contrast, a study in Brazil found some association of short-term exposure to SO2 with increased risk of emergency room visits with pneumonia (33).
One explanation for our findings is that prolonged exposure to higher levels of ambient air pollution predisposes individuals to pneumonia, rather than merely exacerbating existing disease as might be expected with shorter term exposures (34). NO2 exposure may lead to epithelial cell damage, reducing mucociliary clearance. Local bronchial deficits such as reductions in bronchial macrophages, natural killer cells, macrophages, and CD4/CD8 cell ratios may increase susceptibility to bacterial pathogens (35, 36).
Given the large population exposure to ambient air pollution, the results of this study highlight the important health impact that long-term exposure to ambient air pollution can have on respiratory infections. It also emphasizes the need to monitor emissions from vehicles, given that ground level NO2 is derived predominantly from traffic.
Strengths of this study include a well-characterized study population with pneumonia confirmed by a radiologist. In addition, covariates that generally confound the relationship between air pollution and respiratory conditions were measured using primary data collection by trained interviewers. The assignment of the level of exposure to ambient air pollution was based on extensive field measurements and well-validated models. PM2.5 estimates, although based on few government monitors, were probably adequate to capture the spatial variation in this pollutant, which varies over larger ranges of 10–25 km (24). We acknowledge that the results apply to severe pneumonia and may not be generalizable to milder disease. Furthermore, because we excluded patients with suspected infection at other sites (in addition to pneumonia), this limits generalizability. Moreover, ambient air pollution may not completely reflect the actual amount of exposure because indoor exposures are not taken into account and because outdoor activity may be limited. Another possible limitation is in estimates of ambient air pollution because they were derived from regression models and were not direct measurements at participants' residences. Despite these limitations, our results provide the first evidence of association between long-term exposures to higher level of NO2 and PM2.5, preventable involuntary exposures, with hospitalization with pneumonia, a respiratory disease with a large burden of illness among older adults.
| 1. | Mokdad AH, Marks JS, Stroup DF, Gerberding JL. Actual causes of death in the United States, 2000. JAMA 2004;291:1238–1245. |
| 2. | Marrie TJ. Community-acquired pneumonia in the elderly. Clin Infect Dis 2000;31:1066–1078. |
| 3. | Fry A, Shay DK, Holman RC, Curns AT, Anderson LJ. Trends in hospitalizations for pneumonia among persons aged 65 years or older in the United States, 1988–2002. JAMA 2005;294:2712–2719. |
| 4. | Koivula I, Sten M, Mäkelä PH. Risk factors for pneumonia in the elderly. Am J Med 1994;96:313–320. |
| 5. | Farr BM, Bartlett CLR, Miller DL. Risk factors for community-acquired pneumonia diagnosed upon hospital admission. Respir Med 2000;94:954–963. |
| 6. | Jackson ML, Neuzil KM, Thompson WW, Shay DK, Yu O, Hanson CA, Jackson LA. The burden of community-acquired pneumonia in seniors: results of a population-based study. Clin Infect Dis 2004;39:1642–1650. |
| 7. | Zanobetti A, Schwartz J, Dockery DW. Airborne particles are a risk factor for hospital admissions for heart and lung disease. Environ Health Perspect 2000;108:1071–1077. |
| 8. | Burnett RT, Brook JR, Yung WT, Dales RE, Krewski D. Association between ozone and hospitalization for respiratory diseases in 16 Canadian cities. Environ Res 1997;72:24–31. |
| 9. | Spix C, Anderson HR, Schwartz J, Vigotti MA, LeTertre A, Vonk JM, Touloumi G, Balducci F, Piekarski T, Bacharova L, et al. Short-term effects of air pollution on hospital admissions of respiratory diseases in Europe: a quantitative summary of APHEA study results. Air pollution and health: a European approach. Arch Environ Health 1998;53:54–64. |
| 10. | Medina-Ramon M, Zanobetti A, Schwartz J. The effect of ozone and PM10 on hospital admissions for pneumonia and chronic obstructive pulmonary disease: a national multicity study. Am J Epidemiol 2006;163:579–588. |
| 11. | Cheng MF, Tsai SS, Chen PS, Yang CY. Air pollution and hospital admissions for pneumonia in a tropical city: Kasohsiung, Taiwan. J Toxicol Environ Health A 2007;70:2021–2026. |
| 12. | Atkinson RW, Anderson HR, Sunyer J, Ayres J, Baccini M, Vonk JM, Boumghar A, Forastiere F, Forsberg B, Touloumi G, et al. Acute effects of particulate air pollution on respiratory admissions: results from APHEA 2 project. Air pollution and health: a European approach. Am J Respir Crit Care Med 2001;164:1860–1866. |
| 13. | Chauhan AJ, Krishna MT, Frew AJ, Holgate ST. Exposure to nitrogen dioxide and respiratory disease risk. Rev Environ Health 1998;13:73–90. |
| 14. | Chiu HF, Cheng MH, Yang CY. Air pollution and hospital admissions for pneumonia in a subtropical city: Taipai, Taiwan. Inhal Toxicol 2009;21:32–37. |
| 15. | Frampton MW, Smeglin AM, Roberts NJ Jr, Finkelstein JN, Morrow PE, Utell MJ. Nitrogen dioxide exposure in vivo and human alveolar macrophage inactivation of influenza virus in vitro. Environ Res 1989;48:179–192. |
| 16. | Zanobetti A, Bind MA, Schwartz J. Particulate air pollution and survival in a COPD cohort. Environ Health 2008;7:48. |
| 17. | Hartge P, Brinton LA, Rosenthal JF, Cahill JI, Hoover RN, Waksberg J. Random digit dialing in selecting a population-based control group. Am J Epidemiol 1984;120:825–833. |
| 18. | Anonymous. 1998 National Population Health Survey: content for June 1998. January 21, 2000 [accessed 14 October 2009]. Available from: http://www.statcan.ca/english/concepts/nphs/quest98e.pdf. |
| 19. | Collin C, Wade DT, Davies S, Horne V. The Barthel ADL Index: a reliability study. Int Disabil Stud 1988;10:61–63. |
| 20. | Burrough P, McDonnell R. Principles of geographical information systems. New York: Oxford University Press; 1998. |
| 21. | Kanaroglou PS, Jerrett M, Morrison J, Beckerman B, Arain MA, Gilbert NL, Brook JR. Establishing an air pollution monitoring network for intra-urban population exposure assessment: a location–allocation approach. Atmos Environ 2005;39:2399–2409. |
| 22. | Sahsuvaroglou T, Arain A, Kanaroglou P, Finkelstein N, Newbold B, Jerrett M, Beckerman B, Finkelstein M, Brook J, Gilbert N. A land use regression model for predicting traffic pollution concentrations in Hamilton, Ontario. J Air Waste Manag Assoc 2006;56:1059–1069. |
| 23. | Arain MA, Blair R, Finkelstein N, Brook JR, Sahsuvaroglu T, Beckerman B, Zhang L, Jerrett M. The use of wind fields in a land use regression model to predict air pollution concentrations for health exposure studies. Atmos Environ 2007;41:3453–3464. |
| 24. | Kumae T, Arakawa H. Comparison of effects of in vivo nitrogen dioxide exposure starting from different periods on alveolar macrophage activity, assessed by a chemiluminescence technique in Brown-Norway rats. Luminescence 2006;21:226–232. |
| 25. | Moore DK, Jerrett M, Mack W. A land use regression model for predicting ambient fine particulate matter across Los Angeles, CA. J Environ Monit 2007;9:246–252. |
| 26. | Fusco D, Forastiere F, Michelozzi P, Spadea T, Ostro B, Arca M, Perucca CA. Air pollution and hospital admissions for respiratory conditions in Rome, Italy. Eur J Respir Dis 2001;17:1143–1150. |
| 27. | Schwartz J. Air pollution and hospital admissions for the elderly in Birmingham, Alabama. Am J Epidemiol 1994;139:589–598. |
| 28. | Schwartz J. PM10, ozone, and hospital admissions for the elderly in Minneapolis-St. Paul, Minnesota. Arch Environ Health 1994;49:366–374. |
| 29. | Schwartz J. Is there harvesting in the association of airborne particles with daily deaths and hospital admissions? Epidemiology 2001;12:55–61. |
| 30. | Zanobetti A, Schwartz J. Air pollution and emergency admissions in Boston, MA. J Epidemiol Community Health 2006;60:890–895. |
| 31. | Bell ML, Levy JK, Lin Z. The effect of sandstorms and air pollution on cause-specific hospital admissions in Taipei, Taiwan. Occup Environ Med 2008;65:104–111. |
| 32. | Moolgavkar SH, Luebeck EG, Anderson EL. Air pollution and hospital admissions for respiratory causes in Minneapolis-St. Paul and Birmingham. Epidemiology 1997;8:364–370. |
| 33. | Martins LC, Latorre Mdo R, Cardoso MR, Goncalves FL, Saldiva PH, Braga AL. Air pollution and emergency room visits due to pneumonia and influenza in Sao Paulo, Brazil. Rev Saude Publica 2002;36:88–94. |
| 34. | Nel A. Atmosphere: air pollution–related illness: effects of particles. Science 2005;308:804–806. |
| 35. | Olivieri D, Scoditti E. Impact of environmental factors on lung defenses. Eur Respir Rev 2005;14:51–56. |
| 36. | Chauhan AJ, Johnston SL. Air pollution and infection in respiratory illness. Br Med Bull 2003;68:95–112. |
