Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
A number of studies of emergency department visits, a relatively sensitive outcome for respiratory conditions, have corroborated findings from mortality and hospital admission studies regarding an association of ambient air pollution levels and respiratory health effects.1–4 More refined assessment, including analysis of subgroups defined by specific illness or ages, or of air pollutants not routinely monitored, has been limited by study size and available air quality and health outcome data. Many of the single-city time-series studies have covered a relatively short time-span or involved a moderately low number of daily outcome events, resulting in imprecise effect estimates and often restricting analyses to broad outcome and age groups. Recent multicity time-series studies, although having a relatively large number of daily outcome counts, were limited to routinely available outcome and air-quality datasets.5–7
The present study is part of the Study of Particles and Health in Atlanta (SOPHIA). This collection of studies uses extensive air quality data, including detailed particulate matter (PM) component and size fraction information, from a monitoring station in Atlanta operated by the Aerosol Research and Inhalation Epidemiology Study (ARIES). Emergency department visits for respiratory illness were analyzed in relation to routinely collected criteria pollutant levels for the period 1 January 1993 through 31 August 2000, and in relation to additional air pollutants measured at the ARIES monitoring station for the period 1 August 1998 through 31 August 2000. The results for the cardiovascular visits are presented elsewhere.8 In this work, we took advantage of the large number of respiratory emergency department visits and extensive air quality data to examine multiple pollutants in relation to specific respiratory outcomes.
Ambient Air Quality Data
We selected the pollutants and metrics for this analysis a priori on the basis of current hypotheses regarding potentially causal pollutants and components.9,10 We also included pollutants in the a priori list that may be useful markers for sources or for groups of related pollutants (eg, carbon monoxide as a potential marker for primary traffic-related pollutants).
For the period 1 January 1993 through 31 August 2000, we obtained ambient air quality data for 24-hour average PM10 mass (PM with an average aerodynamic diameter less than 10 micrometers), 8-hour maximum ozone, and 1-hour maximum nitrogen dioxide (NO2), sulfur dioxide (SO2), and carbon monoxide (CO) from several existing monitoring networks, including the Air Quality System (AQS, formerly the Aeorometric Information Retrieval System or AIRS), the Georgia Department of Natural Resources, and Metro Atlanta Index. (See map, with the electronic version of this article.) Ozone levels were not monitored during the winter months when ozone levels in Atlanta are low; the remaining pollutants were measured year-round. The AQS air quality data have been described elsewhere.8
For the final 25 months of the study period (1 August 1998 through 31 August 2000), an extensive suite of pollutants, including PM size fractions and components, was measured at the ARIES monitoring station. We selected the following pollutants and metrics for this analysis a priori: oxygenated hydrocarbons (OHC), PM2.5 mass (PM with an average aerodynamic diameter less than 2.5 micrometers), coarse PM (PM with an average aerodynamic diameter between 2.5 and 10 micrometers), ultrafine PM count (PM with an average aerodynamic diameter between 10 and 100 nanometers [nm]), and the PM2.5 components sulfate, acidity, elemental carbon (EC), organic carbon (OC), and an index of water-soluble transition metals. The metrics for PM size fractions and components and for OHC were 24-hour averages, 8-hour maximum for ozone, and 1-hour maximum for NO2, SO2, and CO. The measurement methods for the ARIES monitoring station have previously been described.8,11
Average temperature and dew point temperature (average of the daily minimum and maximum), as well as additional meteorological data measured at Hartsfield-Atlanta International Airport, were obtained from the National Climatic Data Center network. Speciated pollen counts were obtained from the Atlanta Allergy Clinic.
Emergency Department Data
Of the 41 hospitals in the 20-county Atlanta metropolitan statistical area, 37 agreed to participate and 31 provided usable computerized billing records for at least part of the study period. (The map available with the electronic version of this article shows hospital locations.)
Computerized billing records for all emergency department visits between 1 January 1993 and 31 August 2000 were collected, including primary International Classification of Diseases 9th Revision (ICD-9) diagnostic code, secondary ICD-9 diagnosis codes, age, date of birth, sex, race, and residential zip code. Residents of the Atlanta metropolitan statistical area, determined by residential zip code at the time of the visit, were included in the analyses. Repeat visits within a single day were counted as a single visit.
Respiratory case groups of interest were defined using the primary ICD-9 diagnostic codes (all 2-digit extensions were used unless otherwise specified): asthma (493, 786.09), COPD (491, 492, 496), URI (460–466, 477), pneumonia (480–486), and an all-respiratory-disease group that combines the above 4 groups. We assessed the adequacy of the modeling approach using visits for finger wounds (883.0), an outcome group that has comparable temporal variations to the respiratory outcomes of interest and is expected to be unrelated to air pollution.
All analyses were performed using SAS statistical software, version 8.2 (SAS Institute, Inc., Cary, NC) unless otherwise indicated. We defined a priori single-pollutant models to control for long-term temporal trends and meteorological conditions. For the a priori analyses we used Poisson generalized estimating equations,12 with a stationary 4-dependent correlation structure to account for possible autocorrelation in the outcome data (URI, asthma, all respiratory disease) and Poisson generalized linear models13 for outcomes with minimal autocorrelation (pneumonia, COPD). Risk ratios and 95% confidence intervals were calculated for an increase of approximately a standard deviation of pollutant levels. The basic model had the following form:
where Y indicated the count of emergency department visits for a given day for the outcome of interest. The a priori models contained a 3-day moving average of pollution levels lagged 0, 1, and 2 days relative to the visits (levels on the same day as the visit, 1 day previous, and 2 days previous, respectively) (pollutant). Long-term temporal trends were accounted for using cubic splines with monthly knots [g(γ1,...,γN; time)]. Because ozone data were not available from November through March, ozone models used separate time splines for each year. Additional season indicator variables (the 21st day of March, June, September, and December) were added to further control for seasonal trends (season). Cubic splines also were used to control for daily average temperature [g(δ1,...,δN; temp)] and dew point [g(η1,...,ηN; dew point)] with knots at the 25th and 75th percentiles (moving average of lags 0, 1, and 2). Indicator variables for day of week (DOW), federal holidays (holiday), and hospital entry and exit (hospital) also were included in the a priori model (as the hospitals provided data for varying amounts of time). The cubic splines, g(x), were defined as follows:
where wj (x) = (x-τj ) 3 if x ≥ τ j, and wj (x) = 0 otherwise. The cubic splines were defined so that the first and second derivatives were continuous. We evaluated multipollutant models using the same covariates as the single-pollutant models.
We performed several secondary analyses. To assess the lag structure between pollutant levels and emergency department visits, we initially examined separate models for each lag from 0 to 7 days before the visit (up to 2 weeks prior to the visit for asthma). To estimate the overall effect of a unit increase in pollution during the previous 2 weeks, and to investigate whether associations persisted longer than 3 days, we ran unconstrained distributed lag models, including pollution levels from 0 to 13 days before the visit, with additional cubic terms for lags 3–13 for temperature and dew point (in addition to the cubic splines for lags 0–2). For the distributed lag models we presented results only for the pollutants available for the entire study period as the models became unstable for the pollutants available only 25 months.
We examined age-specific case groups (ages 0–1 year, 2–18, 19 years and older, and 65 years and older) as well as season-specific models for warm (April 15 to October 14) and cool (October 15 to April 14) periods. Daily pollen counts (grass, oak, and ragweed) and daily counts of influenza emergency department visits were assessed as confounders. We also assessed general additive models using S-Plus 2000 software (Insightful Corporation, Seattle, WA) with nonparametric LOESS smoothers and nonparametric smoothing splines (10−14 convergence criterion).14,15
In addition to examining the alternate outcome group believed unrelated to air pollution (finger wounds), we performed other analyses to evaluate the adequacy of the modeling approach. We explored negative lags for pollution (pollution levels on days after the visit) as exposure variables, controlling for positive lags, to evaluate the possibility that the modeling choices induced positive associations. We altered the placement (day of the month) and number of knots (degrees of freedom) in the cubic splines for time.
Descriptive statistics for the air quality variables are presented in Table 1; Spearman rank correlation statistics between the daily measures were previously published.8 (Appendix Table 1, available with the electronic version of this article, presents the correlation statistics.) The extent of correlation among the pollutants followed expected patterns. Ultrafine PM count levels were negatively correlated with several pollutants, including ozone, PM, and PM components (sulfate, acidity, and metals). CO, NO2, PM2.5 organic carbon, and PM2.5 elemental carbon were moderately correlated (r = 0.55–0.68). PM10 and PM2.5 mass were moderately correlated with the PM2.5 components (r = 0.56–0.77). Acidity and sulfate were highly correlated with each other (r = 0.85) and moderately correlated with ozone (r = 0.64 and 0.63, respectively) and temperature (r = 0.84 and r = 0.64, respectively). The diurnal patterns of CO and NO2 indicate that mobile source emissions contributed substantially to these pollutant levels. SO2 levels peaked in both summer and winter, corresponding to peak energy demands. SO2 levels exhibited marked temporal and spatial variability, with occasional mid-afternoon peaks resulting from power plant plume fumigation events. Compared with other U.S. cities, ozone and PM2.5 are relatively high (with sulfate and organic carbon comprising relatively high proportions of PM2.5 mass), and acidity is relatively low.16
The 31 hospitals providing usable data for these analyses receive 80% of the annual emergency department visits in the Atlanta area, and contributed information on 4,407,535 total emergency department visits. Respiratory problems accounted for 11% of all emergency department visits. For the entire study period, average daily outcome counts of the subgroups ranged from 7 for COPD to 103 for URI, and the combined respiratory disease group had an average daily count of 172 (Table 2). For the final 25 months of the study, the 31 hospitals contributed 1,888,973 visits.
Results from the a priori single-pollutant models examining 3-day moving averages (lags 0, 1, and 2) of pollutant levels are shown in Table 3. PM10, ozone, NO2, and CO were individually associated with 1–3% increases of URI visits per standard deviation increase of pollutant; similar results were observed for the combined respiratory disease group (60% of all respiratory visits were for URI). Weak and less stable associations were observed for URI in relation to SO2, PM2.5, and organic carbon. A 20 pbb increase of NO2 and a 1 ppm increase in CO were associated with 3.5% and 2.9% increases of COPD visits, respectively. Additional estimates for COPD were elevated, but COPD was the smallest outcome group and therefore had the widest confidence intervals. A 2.8% increase in pneumonia visits was associated with a 2 μg/m3 increase of organic carbon. Small increases of asthma visits were observed in relation to standard deviation increases of PM10, ozone, NO2, and CO; however, the confidence intervals were too wide to exclude a null association. Weak or no associations were observed for the finger wound group. Including daily pollen counts or daily influenza emergency department visits in the models did not affect the observed estimates. General additive models provided similar estimates to those from the a priori models.
In the exploratory models assessing the lag structure between pollutant levels and emergency visits (separate models for each lag), the risk ratios for asthma visits were generally positive and strongest with a lag of 5 to 8 days (Fig. 1). The association with ozone appeared to have a shorter lag structure, with the strongest positive associations at lags of 1 and 2 days. The estimates for ultrafine PM count were negative for lags of 0 and 1 day, and positive for lags of 2 through 4 days. The estimates for URI visits were generally highest for the shorter lags (Fig. 2). The gaseous pollutants tended to have stronger positive associations with URI at a lag of 1 day, while the same-day associations were typically stronger for several particle measures (PM10, PM2.5, coarse PM, PM2.5 components). Sulfate and acidity exhibited a similar trend in relation to URI visits, with positive same-day estimates and negative estimates for a lag of 2 days. Associations for pneumonia and COPD visits were generally positive and strongest for same-day pollutant levels and for levels lagged by 1 day.
Results from unconstrained distributed lags models (lags of 0–13 days) are presented in Table 4. The risk ratios from models using 3-day moving averages can be interpreted as the risk ratio per unit increase of a uniform 3-day moving average, while risk ratios from the distributed lag models can be interpreted as the risk ratio per unit increase of a weighted 14-day moving average. Estimates from distributed lag models (lags of 0–13 days) tended to be substantially higher than those from models using the 3-day moving average (lags of 0–2 days) for PM10, NO2, CO, and SO2, reflecting an additional contribution of days 3–13 in the distributed lag model.
In age-specific analyses, associations for pediatric asthma visits (ages 2–18) in relation to PM10 (RR = 1.016 per 10 μg/m3; 95% CI = 0.998–1.034), NO2 (1.027 per 20 ppb; 1.005–1.050), and CO (1.019 per ppm; 1.004–1.035) were stronger than those for adult asthma visits. Associations for infant (ages 0–1) and pediatric URI visits were substantially stronger than those for adults. Infant URI visits were associated with PM10, ozone, PM2.5 mass, and PM2.5 organic carbon (RRs s per standard deviation increase = 1.026–1.042), and pediatric URI visits were associated with these pollutants as well as NO2 and CO (RRs per standard deviation increase = 1.025–1.047).
The associations for asthma tended to be stronger for several pollutants in the warm months (15 April to 14 October), especially for ozone and PM2.5 organic carbon. The estimates for pneumonia and COPD tended to be higher in the cold months.
In sensitivity analyses that varied the numbers of knots in the time splines, there was a tendency toward lower point estimates and larger standard errors as the number of knots increased. (Appendix Table 2 presenting these results is available with the electronic version of this article.) Changing the placement of the knots in the cubic splines for time did not substantially alter the results. Estimates from models using negative lags for pollution, controlling for positive lags, were predominantly null. Results from models for the period 1 August 1998 through 31 August 2000 using the 2 sources of air quality data were not substantially different (Table 5).
Selected multipollutant analyses were performed. For URI visits, risk ratios for ozone were not substantially attenuated when PM10, NO2, and CO were included in the model (Fig. 3). For COPD, a much smaller outcome group, the risk ratios for both NO2 and CO were attenuated in a 2-pollutant model (data not shown). As the estimates for asthma visits were somewhat elevated for several pollutants in the a priori models, we examined multipollutant models for asthma including all combinations of PM10, ozone, NO2, and CO. The estimates for NO2 were generally not attenuated in multipollutant models, while the estimates for the other pollutants suggested weaker or no associations in the multipollutant models (data not shown).
This time-series study of respiratory emergency department visits provided a rare opportunity to examine associations of an extensive suite of ambient pollutant measures with specific respiratory conditions. In the a priori single-pollutant models (3-day moving average of lags of 0, 1, and 2 days for pollutant levels), URI visits were positively associated with PM10, ozone, NO2, and CO. The association with ozone persisted in multipollutant models. The associations observed for URI appeared to be specific to infants and children. COPD was positively associated with NO2, and CO, while pneumonia was positively associated with PM2.5 organic carbon. These results were generally robust to analytic method and model specification. We would expect several positive and negative associations by chance based on the number of tests performed. Overall, the a priori analyses yielded an abundance of positive associations and only a few negative associations.
Though few reasonably strong associations were observed with the PM finer size fraction and PM component measures, these data were available for a shorter time period and thus the estimates were less stable. The ultrafine particle count data, in particular, were missing for 44% of the days, often in blocks of time, which resulted in additional instability of the ultrafine particle models. Ultrafine particle levels also likely have considerable spatial and compositional heterogeneity. Additionally, high concentration days are potentially associated with different types of ultrafine nucleation events.17,18 Further discussion of the ultrafine PM measurements can be found elsewhere.17,18
In single-day lag models, estimates for URI, pneumonia and COPD were stronger for shorter pollutant lag structures (0–2 days), whereas associations for asthma were generally stronger at longer pollutant lags (5–8 days) and persisted for more than a week in distributed lag models. Results from the distributed lag models (lags of 0–13 days) suggest that associations for several of the outcomes persist for longer than the a priori 3-day moving average of lags 0, 1, and 2 days. A longer lag structure is plausible for emergency department visits for less severe respiratory conditions for biologic reasons (an underlying distribution of sensitivity or illness severity in the population) and for behavioral reasons (the time it takes for an exacerbation to become serious enough to necessitate a visit), especially compared with outcomes such as an acute cardiac event.
The results from this study are generally consistent with previously reported associations of ambient air pollution and respiratory morbidity.1–4 (A brief description and supplemental references are provided in the electronic version of this article.) ED visits for respiratory outcomes have been relatively consistently associated with ozone and PM10, and to a lesser extent with NO2, SO2, and CO.
In previous studies in Atlanta, which examined only asthma exacerbations, investigators reported associations of PM10 and ozone levels with pediatric asthma emergency department visits and hospital admissions in the summer.19–21 In the present study, a 25 ppb increase in ozone was associated with a 2.6% increase in asthma visits in the warm months. Associations for pediatric asthma visits were somewhat stronger than those for adults for PM10, NO2 and CO.
Most previous studies that included PM component data (primarily PM2.5 sulfate and acidity) have been in the northeastern United States and southeastern Canada.22–29 Delfino et al22 observed associations of PM2.5 mass and sulfate, as well as of PM10 and ozone, with respiratory emergency department visits. Stieb et al23 also reported positive associations for PM2.5 mass and sulfate, as well as for ozone, SO2, and PM10, with asthma emergency department visits. Associations of acidity and sulfate with respiratory hospital admissions have been observed by several investigators.24–29 We did not observe any associations for sulfate or acidity in the a priori analyses; however, given the width of the estimated confidence intervals, the study results are not inconsistent with even reasonably strong positive associations of respiratory outcomes with these and other pollutants. Additionally, acidity levels in the previous studies reporting associations with acidity were generally higher than the levels observed in Atlanta for this study.
Our understanding of the biologic mechanisms underlying associations between ambient air pollution and respiratory morbidity is evolving. Inhaled air pollutants may exacerbate existing respiratory disease, resulting in increased reactivity, decreased lung function, and increased respiratory symptoms.30,31 In addition, inhaled pollutants may enhance the allergic response to an allergen.32,33
Many of the pollutant measurements at the ARIES monitoring site appeared to be spatially representative of Atlanta area. Measurements of criteria pollutants were available from both the ARIES and AQS monitoring sites; concentrations measured at the 2 sites were highly correlated and not substantially different in magnitude. Analyses of the ARIES criteria pollutant measurements yielded results comparable to those from analyses of the AQS measurement for the same pollutants. The spatial distribution of ambient PM2.5 mass and several of its constituents, including sulfate, organic carbon, and elemental carbon, appeared to be relatively uniform across available monitoring stations; measurements from the ARIES monitoring site were similar to those from other monitoring sites in Atlanta. No information was available to assess the spatial variability for ultrafine particle count or oxygenated hydrocarbons.
Several issues need to be considered in interpreting the single- and multipollutant results. The single-pollutant results are likely confounded, at least in part, by correlated pollutants. Multipollutant models are typically used to address confounding by correlated pollutants, but results from multipollutant models may also be misleading. Pollutants are measured with differing levels of error (including instrument error as well as other sources of error), whereas some potentially important pollutants may not be measured. A pollutant that exhibits a relatively strong association in a multipollutant model may be acting as a surrogate for an unmeasured or poorly measured pollutant.
The goal of this study was to assess the association between ambient pollution levels and respiratory morbidity. Ambient pollution levels are of interest for the assessment of population-level health effects of air pollution as well as for regulatory purposes. The measurement error that results from using centrally located monitors is likely to attenuate associations, but would not likely induce spurious associations. Additionally, personal behavior such as air conditioning use or time spent outdoors may affect personal exposure levels. This could affect the magnitude of the observed associations when compared with other locations with different behavior profiles. Eighty-three percent of households in Atlanta have central air conditioning,34 which could weaken associations observed in Atlanta during the warm season relative to those observed in other areas.35 However, in season-specific analyses, associations were often stronger or of similar magnitude in the warm season compared with the cool season or to the year-round analyses, especially for ozone.
We used an a priori approach to reduce possible biases associated with multiple testing and selective reporting of effect estimates. The pollutant metrics, outcome groups of interest, temporal relationship of the pollutant and outcome, and control for temporal trend were chosen prior to examining the data. We then performed secondary analyses to explore the associations further. Although there was some variability when we changed the number of knots to control for time, the overall conclusions would not have been substantially altered had we chosen a model with different knot frequency as the a priori model. We considered over-controlling for time a more conservative alternative to undercontrolling.
In this study, a large sample size and extensive air quality measurements allowed us to examine specific respiratory outcome groups in relation to air pollutants not routinely available for epidemiologic studies. The results contribute to the evidence of an association of several correlated gaseous and particulate pollutants (including ozone, NO2, CO, PM, and organic carbon) with specific respiratory conditions.
This research used air quality data from a monitoring station operated by ARIES and managed by Ron Wyzga and Alan Hansen of EPRI. Principal air quality collaborators on the ARIES study include: Eric Edgerton and Ben Hartsell at Atmospheric Research & Analysis, Inc; Peter McMurry and Keung Shan Woo at the University of Minnesota; Rei Rassmussen at the Oregon Graduate Institute; Barbara Zielinska at the Desert Research Institute; and Harriet Burge, Christine Rogers, Helen Suh, and Petros Koutrakis at the Harvard School of Public Health. We thank the Atlanta Allergy Clinic for providing pollen data. We acknowledge the helpful advice given by the ARIES advisory committee: Tina Bahadori at the American Chemistry Council; Rick Burnett at Health Canada; Isabelle Romieu at Instituto Nacional de Salud Publica; Barbara Turpin at Rutgers University; John Vandenberg at the U.S. Environmental Protection Agency; and Warren White at University of California at Davis. We thank Keely Cheslack-Postava, Jacqueline Tate, and Marlena Wald for their assistance. We are also grateful to the participating hospitals, whose staff members devoted many hours of time as a public service.
1. Environmental Protection Agency. Air quality criteria for particulate matter. Washington, DC: Office of Research and Development, National Center For Environmental Assessment, Research Triangle Park Office, Research Triangle Park, NC EPA/600/P-99/002bB, 2001.
2. Dockery DW, Pope CA. Acute respiratory effects of particulate air pollution. Annu Rev Public Health
3. Bascom R, Bromberg PA, Costa DA, et al. Health effects of outdoor air pollution. Part 1. Am J Respir Crit Care Med
4. Bascom R, Bromberg PA, Costa DA, et al. Health effects of outdoor air pollution. Part 2. Am J Respir Crit Care Med
5. Samet JM, Zeger SL, Dominici F, et al. The National Morbidity, Mortality, and Air Pollution Study Part II: Morbidity, Mortality, and Air Pollution in the United States.
Research Report 94. Cambridge MA: Health Effects Institute; 2000.
6. Schwartz J, Zanobetti A, Bateston T. Morbidity and mortality among elderly residents of cities with daily PM measurements. In: Revised Analyses of Time-Series Studies of Air Pollution and Health.
Special Report. Boston MA: Health Effects Institute; 2003:25–58.
7. Atkinson RW, Anderson HR, Sunyer J, et al. Acute effects of particulate air pollution on respiratory admissions: results from APHEA 2 project. Am J Respir Crit Care Med
8. Metzger KB, Tolbert PE, Klein M, et al. Ambient air pollution and cardiovascular emergency department visits. Epidemiology
9. Albritton DL, Greenbaum DS. Atmospheric observations: helping build the scientific basis for decisions related to airborne particulate matter. Report of the PM Measurements Research Workshop
, Chapel Hill, NC; 1998.
10. Schlesinger RB. Properties of ambient PM responsible for human health effects: coherence between epidemiology and toxicology. Inhal Toxicol
. 2000;12(suppl 1):23–25.
11. Van Loy M, Bahadori T, Wyzga R, Hartsell B, Edgerton E. Aerosol Research and Inhalation Epidemiology Study (ARIES): PM2.5
mass and aerosol component concentrations and sampler intercomparisons. J Air Waste Manage Assoc
12. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics
13. McCullagh P, Nelder JA. Generalized Linear Models.
ed. New York: Chapman and Hall; 1989.
14. Hastie T, Tibshirani R. Generalized Additive Models
. London: Chapman and Hall; 1990.
15. Dominici F, McDermott A, Zeger SL, Samet JM. On the use of generalized additive models in time-series studies of air pollution and health. Am J Epidemiol
16. Butler AJ, Andrew MS, Russell AG. Daily sampling of PM2.5
in Atlanta: results of the first year of the Assessment of Spatial Aerosol Composition in Atlanta study. J Geophys Res.
17. Woo K S, Chen DR, Pui DYH, McMurry PH. Measurement of Atlanta aerosol size distributions: observations of ultrafine particle events. Aerosol Sci Technol
18. McMurry PH, Woo KS. Size distributions of 3-100-nm urban Atlanta aerosols: measurement and observations. J Aerosol Med
19. White MC, Etzel RA, Wilcox WD, Lloyd C. Exacerbations of childhood asthma and ozone pollution in Atlanta. Environ Res
20. Tolbert PE, Mulholland JA, MacIntosh DL, et al. Air quality and pediatric emergency room visits for asthma in Atlanta. Am J Epidemiol
21. Friedman MS, Powell KE, Hutwagner L, Graham LM, Teague WG. Impact of changes in transportation and commuting behaviors during the 1996 Summer Olympic Games in Atlanta on Air Quality and Childhood Asthma. J Am Med Assoc
22. Delfino RJ, Murphy-Moulton AM, Burnett RT, Brook JR, Becklake MR. Effects of air pollution on emergency room visits for respiratory illnesses in Montreal, Quebec. Am J Respir Crit Care Med
23. Stieb DM, Beveridge RC, Brook JR, et al. Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint John, Canada. J Exp Anal Environ Epidemol
24. Thurston GD, Ito K, Kinney PL, Lippman M. A multi-year study of air pollution and respiratory hospital admissions in three New York state metropolitan areas: results for 1988 and 1989 summers. J Expo Anal Environ Epidemiol
25. Burnett RT, Dales RE, Raizenne ME, et al. Effects of low ambient levels of ozone and sulfate on the frequency of respiratory admissions to Ontario hospitals. Environ Res
26. Burnett RT, Dales R, Krewski D, Vincent R, Dann T, Brook JR. Associations between ambient particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am J Epidemiol
27. Burnett RT, Cakmak S, Brook JR, Krewski D. The role of particulate size and chemistry in the association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases. Environ Health Perspect
28. Gwynn RC, Burnett RT, Thurston GD. A time-series analysis of acidic particulate matter and daily mortality and morbidity in the Buffalo, New York, region. Environ Health Perspect
29. Lippmann M, Ito K, Nadas A, Burnett RT. Associations of Particulate Matter Components With Daily Mortality and Morbidity in Urban Populations.
Research Report 95. Cambridge MA: Health Effects Institute; 2000.
30. Pope CA. Epidemiology of fine particulate air pollution and human health: biologic mechanisms and who's at risk. Environ Health Perspect
. 2000;108(suppl 4):713–723.
31. Goldsmith CA, Kobzik L. Particulate air pollution and asthma: a review of epidemiological and biological studies. Rev Environ Health
32. Rusznak C, Devalia JL, Davies RJ. Airway response of asthmatic subjects to inhaled allergen after exposure to pollutants. Thorax
33. Ormstad H, Johansen BV, Gaarder PI. Airborne house dust particles and diesel exhaust particles are allergen carriers. Clin Exp Allergy
34. U.S. Census Bureau, Current Housing Reports, Series H170/96–21. American Housing Survey for the Atlanta Metropolitan Area in 1996. U.S. Government Printing Office, Washington, DC: 1997. Available at: www.census.gov/hhes/www/ahs.html
; Internet; accessed December 14, 2004.
35. Janssen NAH, Schwartz J, Zanobetti A, Suh HH. Air conditioning and source-specific particles as modifiers of the effect of PM10
on hospital admissions for heart and lung disease. Environ Health Perspect
“It is the mark of an educated person to look for precision in each class of things only so far as the nature of the subject permits.”
Supplemental Digital Content
© 2005 Lippincott Williams & Wilkins, Inc.