Objectives.The purpose of this study was to examine the association between ambient air pollution and hospitalization for respiratory infections among children who were younger than 15 years in Toronto during a 4-year period (1998–2001).
Methods.Exposures averaged during periods that varied from 1 to 7 days were used to assess the effects of air pollutants, including thoracic particulate matter (PM10), fine (PM2.5) and coarse (PM10–2.5) particulate matter, carbon monoxide (CO), sulfur dioxide (SO2), nitrogen dioxide (NO2), and ozone (O3), on hospitalization for respiratory infections. A case-crossover design was used to calculate odds ratios for the hospitalization adjusted for daily weather conditions with an incremented exposure corresponding to the interquartile range in air pollution exposures.
Results.When particulate matter and gaseous pollutants were mutually taken into account, the effect remained pronounced for PM10–2.5 in both boys and girls. The adjusted odds ratio for 6-day average exposure to PM10–2.5 with an increment of 6.5 μg/m3 was 1.15 (95% confidence interval: 1.02–1.30) for boys and 1.18 (95% confidence interval: 1.01–1.36) for girls. The effect also remained for PM10 in boys and for NO2 in girls. PM2.5, CO, SO2, and O3 showed no significant effects on hospitalization for respiratory infection in both genders when other pollutants were taken into consideration.
Conclusion.Our study suggested a detrimental effect of relatively low levels of ambient particulate matter and gaseous pollutants, especially coarse particulate matter and NO2, on hospitalization for respiratory infections in children.
- air pollution
- coarse particulate matter
- gaseous pollutants
- hospitalization for respiratory infection
- case-crossover analysis
- risk assessment
Lower respiratory infection is a main cause of mortality and morbidity in children in developing countries. Morbidity from childhood respiratory infections is also high in developed countries.1 Although respiratory infections in children are usually nonfatal in developed countries, they heavily burden the health care systems.2 Additional progress in preventing the diseases will have a significant impact on children's health.1
Although there is abundant evidence linking outdoor air pollution with respiratory symptoms, reduced lung function, bronchial reactivity, and asthma, the relationship of hospital morbidity for respiratory infections to a relatively low level of exposure to air pollution in children has not been well studied and is considered a knowledge gap in developed countries.3–6 Most previous studies of hospital outcomes associated with air pollution exposures usually either aggregated respiratory infections with other respiratory conditions such as asthma and chronic obstructive pulmonary disease into a single study group or focused on the air pollution effect on respiratory infection in all ages,7 elderly,8 or very young children (aged 0–2 years).9 Children generally breathe more rapidly than adults, they may have more exposure to air pollutants per kilogram of body weight, and respiratory infections are generally more common in boys than in girls.10 So far it has not been clear whether there are gender differences in effects of ambient air pollutants on respiratory infections in children.11,12 This study used a case-crossover design to examine the association between ambient air pollution and respiratory infections in boys, girls, and children as a whole who were younger than 15 years in Toronto, Ontario, Canada.
The present analysis was based on daily air pollution and hospitalization data collected in metropolitan Toronto between 1998 and 2001. Hospitalization data were obtained from the Discharge Abstract Database. We selected hospitalizations for respiratory infections in the sub-age group (0–14 years) as our study population. Hospitalization for respiratory infections was defined as an admission for which respiratory infections were the primary diagnosis that caused the greatest number of hospital days of stay. Respiratory infections in this analysis included the following conditions: laryngitis, tracheitis, bronchitis, bronchiolitis, pneumonia, and influenza (International Classification of Diseases, Ninth Revision codes 464, 466, and 480–487). Admissions were restricted to children who both resided in Toronto and were admitted there. This study included only admissions to hospitals for acute care and active treatment that were considered as emergency or urgent. Both planned admissions and transfers from other institutions were excluded from this analysis. The data were grouped into daily hospital admission counts for both boys and girls.
Air pollution data were obtained from the National Air Pollution Surveillance system, and weather data were obtained from Environment Canada's weather archive. Daily air pollution data were available from a minimum of 4 to a maximum of 7 monitoring stations, including carbon monoxide (CO) and sulfur dioxide (SO2) from 5 monitoring stations, nitrogen dioxide (NO2) and ozone (O3) from 7 stations, and particulate matter of median aerometric diameter <10 and 2.5 μm (PM10 and PM2.5, respectively) from 4 stations. The study region consisted of the cities of Toronto, North York, East York, Etobicoke, Scarborough, and York. These monitoring sites span the breath of the region and include major population areas. Figure 1 shows the locations of monitoring stations for each air pollutant. CO, NO2, O3, and SO2 were measured using “reference methods” or “equivalent methods” as designated by the US Environmental Protection Agency. CO was measured using nondispersive infrared spectrometry, NO2 was measured using chemiluminescence, O3 was measured using chemiluminescence/ultraviolet photometry, and SO2 was measured using coulometry/ultraviolet fluorescence. PM2.5 and PM10 were measured using tapered element oscillating microbalance (TEOM) instruments. Although PM10 and PM2.5 TEOM samplers were not necessarily co-located, we computed the average coarse fraction (PM10–2.5) as average PM10 minus average PM2.5 among sites in Toronto. Although in absolute terms this may introduce error, it should accurately reflect relative day-to-day changes in exposure to coarse particles (Tom Dann, Environment Canada, personal communication, 2004). There is a reasonably good correlation between PM10–2.5 from dichotomous monitor and TEOM coarse fraction. The correlation was 0.85 between dichotomous PM10–2.5 and the difference between TEOM PM10 and TEOM PM2.5 at 1 site that had all 3 measures. The correlation was 0.74 when the values were averaged over all sites.
Daily averaging concentrations of each air pollutant were used in this analysis, except daily maximum 1-hour level was used for O3 because of its diurnal profile of exposures. Average values were calculated when data were available from >1 monitoring station.13 The concentrations of O3 are very low at night, peaking in the late afternoon, and then falling off in the evening.14 Daily information on weather conditions was obtained from Pearson international airport and included daily mean temperatures and dew point temperature.
A bidirectional case-crossover design was used in this study. Bateson and Schwartz15 reported that the bidirectional case-crossover design can control for different patterns of time trends in exposures and outcomes. The level of air pollution at the time of hospitalization for each case (the case period) was compared with a level obtained in a specified period before and/or after the health event (the control period). Cases in this analysis included only children who were 0 to 14 years of age and were admitted to a hospital in the study area, with respiratory infections as the principal reason for the hospital stay during the period between 1998 and 2001.
The acute effects of environmental exposure may be immediate or may occur several days after exposure. In this study, we examined the acute effect of 1-day to multiple-day averages of air pollution ending on the admission date. Previous studies have documented that increased hospitalizations are most strongly associated with air pollution on the day of admission or within up to 4 days.4,7,16 A recent study observed that the estimated effect of multiple-day exposures to air pollution could be stabilized on 5 to 6 days.17 In this study, we calculated 1- to 7-day exposure averages ending on the admission date as the exposures in the case period.
Control periods of 2 weeks before and after the admission date were used in the bidirectional scheme to minimize autocorrelation between case and control exposures and to control for seasonal and long-term effects.17,18 To be matched with the case period, exposures in the control period were expressed as 1- to 7-day averages for each pollutant ending on the date 2 weeks before and after the admission date.
This study applied conditional logistic regression models for the case-crossover design by using the SAS 8.2 statistical package's PHREG procedure, a program for fitting the Cox proportional hazards model.19 The conditional likelihood function for logistic regression can be treated as a special case of Cox partial likelihood, which is used to fit the proportional hazards model.20 We estimated odds ratios (ORs) for hospitalization of respiratory infections in relation to various air pollutants during the case period as compared with the control periods after adjustment for daily mean temperature and dew point temperature. On the basis of previous studies,16,21 we added squared terms of each of the weather conditions as additional covariates. The ORs were calculated on the basis of an increment in exposure corresponding to the interquartile range of each pollutant. The effects of particulate matter on hospitalization for respiratory infections were examined further, taking into consideration the effects of gaseous pollutants (CO, SO2, NO2, and O3). Particulate matter was also taken into account when the relationships between gaseous pollution and hospitalization for respiratory infections were examined. Because thoracic particulate matter (PM10) is a function of fine (PM2.5) and coarse (PM10–2.5) particulate matter, only fine and coarse particulate matter were considered in the analyses of gaseous pollutants in relation to hospitalization for respiratory infections.
Lumley and Levy22 first pointed out that a standard conditional logistic regression analysis for bidirectional case-crossover designs is only approximately correct and in some cases estimates would be biased. Some design strategies have been suggested by several studies18,22,23 to eliminate or reduce these biases. Lumley and Levy22 have a concern that short-term autocorrelations in the exposures may introduce bias analogous to overmatching in a case-control study. However, such bias will be largely reduced when the interval between case and control periods is weekly based. A 1-week interval allows exclusion of a short-term autocorrelation and ensures independence among observations.18 A simulation study23 suggested that selection bias in a case-crossover study design could be reduced by choosing a shorter interval period. A 2-week interval between case and control periods was selected in the present study with considerations of control of short-term autocorrelation and time-varying trends and assessment of potential multiple-day exposure effects. Another simulation study conducted by Levy et al18 showed that there is little bias (0.4%) when an interval of 2 weeks is used.
Table 1 provides summary statistics for air pollution, weather conditions, and hospitalizations for respiratory infections. There were a total of 6782 hospitalizations for respiratory infections in children who were 0 to 14 years of age (3998 for boys and 2784 for girls) with a daily average of 4.64 (2.74 for boys and 1.91 for girls) in Toronto during the period from 1998 to 2001. Daily information for gaseous pollutants was available for the whole study period. Daily averages of particulate matter were available for ∼99% of the total 1461 days between 1998 and 2001.
Correlations between air pollutants are shown in Table 2. PM10–2.5 and PM2.5 both were highly correlated with PM10 (r = 0.87 for PM 2.5, r = 0.76 for PM10–2.5). A lesser degree of correlation was found between PM10–2.5 and PM2.5 (r = 0.33). Gaseous pollutants all were positively correlated with each other, with the exception of O3, which was weakly correlated with the other gaseous pollutants. The gaseous pollutants all were positively correlated with ambient particulate matter.
Table 3 provides crude and adjusted ORs and their 95% confidence intervals (CIs) for exposures to each air pollutant in relation to hospitalizations for respiratory infections in boys and girls separately and in children as a whole. Estimates were calculated for 1- to 7-day average levels of each air pollutant. Effects of exposure averaging of 4 and 6 days are presented in Table 3 because most estimated effects of air pollution observed in this study seemed to increase with increasing number of days of exposure and to stabilize at ∼4 to 6 days. For particulate matter, PM10–2.5 in both boys and girls and PM10 only in boys but not in girls showed consistently significant positive associations with hospitalizations for respiratory infections before and after adjustment for gaseous pollutants (Table 3). The adjusted OR for a 6-day average exposure to PM10–2.5 with an increment of 6.5 μg/m3 was 1.15 (95% CI: 1.02–1.30) for boys and 1.18 (95% CI: 1.01–1.36) for girls. The corresponding OR for PM10 with an increment of 12.5 μg/m3 was 1.25 (95% CI: 1.01–1.54) in boys. There was no significant association between fine particulate matter (PM2.5) and hospitalization for respiratory infections in boys, girls, or children as a whole when gaseous pollutants were taken into account.
For gaseous pollutants associated with hospitalization for respiratory infections, only the effect of 5- to 7-day average exposures to NO2 in girls but not in boys was significant, and the adjusted OR for a 6-day exposure to NO2 was 1.31 (95% CI: 1.05–1.63; Table 3). The associations between gaseous pollutants, including CO, SO2, and NO2, and hospitalizations of respiratory infections in boys were significant only before adjustment for particulate matter. O3 showed no significant effect on hospitalizations for respiratory infections in either boys or girls before and after adjustment for particulate matter.
Air pollution levels are relatively low in Toronto as compared with many other cities of a similar size in the world.7 Air pollution levels during the study period from 1998 to 2001 generally were lower as compared with those in previous years, between 1980 and 1994.7 The maximum daily concentration of each pollutant was largely reduced, with a reduction ranging from 12% for O3 to 65% for SO2. The mean daily average concentration decreased from 17.99 μg/m3 to 9.59 μg/m3 for PM2.5 and from 30.16 μg/m3 to 20.41 μg/m3 for PM10. For O3, however, there was an increase in the mean level of daily maximum 1-hour concentration, and such an increase could be related to warmer summers.24 The level of PM10 and PM2.5 never exceeded the US National Ambient Air Quality Standards,25 with an average level being ∼13% of the standard of 150 μg/m3 for PM10 and 15% of the standard of 65 μg/m3 for PM2.5. For gaseous pollutants, similarly, the means of daily average levels of CO, SO2, NO2, and O3 were below the National Ambient Air Quality Standards,25 whereas NO2 exceeded the standard of 53 ppb on 2 days and O3 exceeded the standard of 120 ppb on 1 day.
Although air pollution levels are relatively low in Toronto, coarse and thoracic particulate matter in this study showed significant effects on hospitalization for respiratory infection. When gaseous pollutants were included in regression models, the effects of coarse PM remained significant for both genders, and that of thoracic PM was significant for boys. Fine particulate matter was not associated with hospitalization for respiratory infections in either gender.
Previous studies showed inconsistent results regarding particulate matter and respiratory infections. PM10 was found to be associated with hospitalizations for pneumonia in all ages in Birmingham, UK, between 1992 and 199426 and in elderly people in Minnesota in the period 1986–1989.27 One study in German cities found that total suspended particles (TSP) was significantly related to pediatrician-reported croup but not to bronchitis.28 Another study in Rome, Italy, found no relationship between TSP and emergency department admissions for acute respiratory infections in children.29
PM10–2.5 tended to have a greater effect on hospitalization for respiratory infections than PM2.5, which is consistent with previous findings for asthma hospitalizations in children.7,17 There is also evidence for other health outcomes, including mortality from all causes, respiratory diseases, and cardiovascular disease and hospitalizations for cardiovascular diseases.30 Coarse particles deposit in the upper airways of the lungs31 and are associated with increased cytotoxicity and proinflammatory cytokines interleukin-6 and interleukin-8.32 An experimental study showed that exposure to coarse particles significantly exacerbated pulmonary infection in mice.33 Particulate matter is likely immunosuppressive and may undermine normal pulmonary antimicrobial defense mechanisms.34 Additional studies are needed to explore the potential mechanisms.
In our study, some gaseous pollutants showed significant effects on hospitalization for respiratory infections, but the influences were no longer significant when particulate matter was taken into consideration. The only exception is the effect of NO2 in girls, which remained significant even after controlling for particulate matter. Several previous studies have linked respiratory infection to exposure to NO2.28,29,35,36 Two of these studies also looked at the effect of TSP,28,29 and only 1 study28 found an association between TSP and croup cases. None of these studies considered the exposure to inhalable particulate matter with aerometric diameter ≤10 μm. Most of these studies did not perform gender-specific analyses. Only 1 case-control study in Stockholm found that wheezing bronchitis was related to outdoor NO2 exposure in girls but not in boys, which was consistent with our finding of NO2 effects on respiratory infections. NO2 exposure was found to be associated with a reduction in peak expiratory flow with virus infection by up to 75%.5
Boys have smaller airways relative to their lung volume than girls.37 Other factors such as smooth muscle and vascular functions and hormonal status may also play a role in the gender-related susceptibility in air pollution effects on respiratory infections. In the present study, boys were more likely to be admitted to the hospital for respiratory infection than girls. A recent study10 suggested that in children who were younger than 15 years, the hospitalization for respiratory infection was more common in boys than in girls, but such a gender difference decreased with increasing age and reversed in children and young adults. There is a lack of consistent results for gender differences in health effects of various air pollutants.
It remains unclear how ambient outdoor air pollutants interact with respiratory infectious agents. Experimental evidence suggests that exposures to ambient air pollution may adversely affect lung defense functions such as aerodynamic filtration, mucociliary clearance, particle transport, and detoxification by alveolar macrophages. Macrophages can inhibit viral replication and also limit viral infections by removing the debris of destroyed cells and by presenting viral antigens to T lymphocytes.5
The present study applied averaged air pollution exposures over centrally sited outdoor monitors to be a surrogate of personal exposures. Misclassification of individual exposures would be a concern when exposure measurement is at the population level. Such an exposure error generally does not lead to substantial bias in the risk estimates while the variance of the estimate is increased.38 Although the case-crossover study has the advantage of incorporating measurements of exposure or potential effect modifiers into the analysis when this information is available on an individual level, the hospitalization data for the present study were available only at the aggregated level. Future studies applying geospatial modeling would be powerful to estimate environmental exposure concentrations at the neighborhood or even the individual level and therefore take into account variations in environmental exposures across a study region.39
Our study suggested a detrimental effect of relatively low levels of ambient particulate matter and gaseous pollutants on hospitalization for respiratory infections in children. After adjusting for other pollutants, hospitalization for respiratory infections was still significantly associated with coarse particulate matter in both genders, with thoracic particulate matter in boys, and with NO2 in girls. Effects of fine particulate matter CO, SO2, and O3 were not significant when other pollutants were included in the models.
Dr Chen is a Canadian Institutes of Health Research Investigator Award recipient. We thank Marc Smith-Doiron for preparing the data.
- Accepted February 14, 2005.
- Reprint requests to (Y.C.) Department of Epidemiology and Community Medicine, University of Ottawa, 451 Smyth Rd, Ottawa, ON, Canada K1H 8M5. E-mail:
No conflict of interest declared.
- ↵West JV. Acute upper airway infections. Br Med Bull.2002;61 :215– 230
- ↵Chauhan AJ, Johnston SL. Air pollution and infection in respiratory illness. Br Med Bull.2003;68 :95– 112
- ↵Graham NM. The epidemiology of acute respiratory infections in children and adults: a global perspective. Epidemiol Rev.1990;12 :149– 178
- ↵Schwartz J. Air pollution and hospital admissions for the elderly in Birmingham, Alabama. Am J Epidemiol.1994;139 :589– 598
- ↵Burnett RT, Smith-Doiron M, Stieb D, et al. Association between ozone and hospitalization for acute respiratory diseases in children less than 2 years of age. Am J Epidemiol.2001;153 :444– 452
- ↵Peters JM, Avol E, Gauderman WJ, et al. A study of twelve Southern California communities with differing levels and types of air pollution. II. Effects on pulmonary function. Am J Respir Crit Care Med.1999;159 :768– 775
- ↵SAS Institute Inc. SAS/ETS Software: Changes and Enhancements, Release 8.02. Cary, NC: SAS Institute Inc; 2001
- ↵Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. 3rd ed. Pacific Grove, CA: Duxbury Press; 1998
- ↵Schwartz J, Spix C, Touloumi G, et al. Methodological issues in studies of air pollution and daily counts of deaths or hospital admissions. J Epidemiol Community Health.1996;50(suppl 1) :S3– S11
- ↵Macfarlane R, Campbell M, Basrur SV. Toronto's Air: Let's Make It Healthy. Toronto, ON, Canada: Health Promotion and Environmental Protection Office, Toronto Public Health; 2000
- ↵US Environmental Protection Agency. National Ambient Air Quality Standards (NAAQS). Washington, DC: U.S. EPA Office of Air Quality Planning and Standards; 2004
- ↵Wordley J, Walters S, Ayres JG. Short term variations in hospital admissions and mortality and particulate air pollution. Occup Environ Med.1997;54 :108– 116
- ↵Fusco D, Forastiere F, Michelozzi P, et al. Air pollution and hospital admissions for respiratory conditions in Rome, Italy. Eur Respir J.2001;17 :1143– 1150
- ↵US Environmental Protection Agency. Air Quality Criteria for Particulate Matter (Report No. EPA/600/AP-95/001b). Research Triangle Park, NC: US Environmental Protection Agency; 1995
- ↵Zelikoff JT, Schermerhorn KR, Fang K, et al. A role for associated transition metals in the immunotoxicity of inhaled ambient particulate matter. Environ Health Perspect.2002;110(suppl 5) :871– 875
- ↵Pershagen G, Rylander E, Norberg S, et al. Air pollution involving nitrogen dioxide exposure and wheezing bronchitis in children. Int J Epidemiol.1995;24 :1147– 1153
- Copyright © 2005 by the American Academy of Pediatrics