Journal of Exposure Analysis and Environmental Epidemiology (2000) 10, 461±477
# 2000 Nature America, Inc. All rights reserved 1053-4245/00/$15.00
www.nature.com/jea
Air pollution, aeroallergens and cardiorespiratory emergency department
visits in Saint John, Canada
DAVID M. STIEB,a ROBERT C. BEVERIDGE,b JEFFREY R. BROOK,c MARC SMITH -DOIRON,a
RICHARD T. BURNETT,a ROBERT E. DALES,a SERGE BEAULIEU,b STAN JUDEKa AND
ALEXANDRE MAMEDOVc
a
Environmental Health Directorate, Health Canada, Ottawa, ON, Canada
Atlantic Health Sciences Corporation, Canada
c
Meteorological Service of Canada, Environment Canada, Canada
b
Existing studies of the association between air pollution, aeroallergens and emergency department ( ED ) visits have generally examined the effects of a few
pollutants or aeroallergens on individual conditions such as asthma or chronic obstructive pulmonary disease. In this study, we considered a wide variety of
respiratory and cardiac conditions and an extensive set of pollutants and aeroallergens, and utilized prospectively collected information on possible effect
modifiers which would not normally be available from purely administrative data. The association between air pollution, aeroallergens and cardiorespiratory
ED visits ( n = 19,821 ) was examined for the period 1992 to 1996 using generalized additive models. ED visit, air pollution and aeroallergen time series were
prefiltered using LOESS smoothers to minimize temporal confounding, and a parsimonious model was constructed to control for confounding by weather and
day of week. Multipollutant and multi - aeroallergen models were constructed using stepwise procedures and sensitivity analyses were conducted by season,
diagnosis, and selected individual characteristics or effect modifiers. In single - pollutant models, positive effects of all pollutants but NO2 and COH were
observed on asthma visits, and positive effects on all respiratory diagnosis groups were observed for O3, SO2, PM10, PM2.5, and SO42 . Among cardiac
conditions, only dysrhythmia visits were positively associated with all measures of particulate matter. In the final year - round multipollutant models, a 20.9%
increase in cardiac ED visits was attributed to the combination of O3 ( 16.0%, 95% CI 2.8 ± 30.9 ) and SO2 ( 4.9%, 95%CI 1.7 ± 8.2 ) at the mean concentration
of each pollutant. In the final multipollutant model for respiratory visits, O3 accounted for 3.9% of visits ( 95% CI 0.8 ± 7.2 ) , and SO2 for 3.7% ( 95% CI 1.5 ±
6.0 ) , whereas a weak, negative association was observed with NO2. In multi - aeroallergen models of warm season asthma ED visits, Ascomycetes, Alternaria
and small round fungal spores accounted for 4.5% ( 95% CI 1.8 ± 7.4 ) , 4.7% ( 95% CI 1.0 ± 8.6 ) and 3.0% ( 95% CI 0.8 ± 5.1 ) , respectively, of visits at their
mean concentrations, and these effects were not sensitive to adjustment for air pollution effects. In conclusion, we observed a significant influence of the air
pollution mix on cardiac and respiratory ED visits. Although in single - pollutant models, positive associations were noted between ED visits and some
measures of particulate matter, in multipollutant models, pollutant gases, particularly ozone, exhibited more consistent effects. Aeroallergens were also
significantly associated with warm season asthma ED visits. Journal of Exposure Analysis and Environmental Epidemiology ( 2000 ) 10, 461 ± 477.
Keywords: aeroallergens, air pollution, cardiac, respiratory, emergency department.
Introduction
Although the availability of routinely collected administrative data on endpoints such as mortality and hospital
admissions has led to a relatively large number of daily time
series analyses of the association of air pollution with these
health endpoints, data on other outcomes such as emergency
department ( ED ) visits are not consistently available. In
this study, we used an ED visits information system that had
been developed for quality of care assessment purposes, to
assemble a time series that could be linked with environmental data. Although existing studies of the association
between environmental factors and ED visits have generally
1. Address all correspondence to: Dr. David M. Stieb, Environmental
Health Directorate, Health Canada, Room D - 407, Jeanne Mance Building,
1904B Tunney's Pasture, Ottawa, ON, Canada, K1A 0K9. Tel.: + 1 - 613 957 - 3132. Fax: + 1 - 613 - 954 - 7612. E-mail: dave stieb@hc - sc.gc.ca
Received 4 February 2000; accepted 27 June 2000.
examined the effects of a few air pollutants or aeroallergens
on individual conditions such as asthma or chronic
obstructive pulmonary disease, in this study, we considered
a wide variety of respiratory and cardiac conditions and an
extensive set of pollutants and aeroallergens. We also
utilized prospectively collected information on possible
effect modifiers, including personal characteristics, and
other factors (Stieb et al., 2000 ). This information would
not normally be available from purely administrative data,
which are often relied upon in other studies.
Methods
Saint John is an industrial city on Canada's Atlantic coast.
The population of approximately 130,000 is served by two
hospital EDs. Access to health services is universal under
Canada's publicly funded health care system.
Stieb et al.
Air pollution, allergens and emergency department visits
Table 1. Summary of model specification by diagnosis, including window length for LOESS filter and weather variables ( number of days lagged ) in final
weather model.
Diagnosis
Window
length
( days )
Weather variables
Angina / myocardial infarction
120
FOGa ( 1 ) , MAXTMPb ( 8 ) , MEANRHc ( 10 ) , MINDPd ( 4 ) , PRECIPe ( 2 ) , TPRECIPf ( 9 )
Congestive heart failure
Dysrhythmia / conduction disturbance
91
183
FOG ( 8 ) , MEANRH ( 1 ) , PRECIP ( 6 ) , PRS3HRCg ( 0,6 ) , PRS88Ch ( 10 ) , TPRECIP ( 7,8 )
MAXDPi ( 1 ) , MEANRH ( 1 ) , PRECIP ( 2 ) ,
All cardiac
Asthma
Chronic obstructive pulmonary disease
Respiratory infection
All respiratory
60
FOG ( 1 ) , MAXDP ( 8 ) , MEANRH ( 1,2,10 ) , PRECIP ( 2,7 ) , PRS3HRC ( 6 ) , PRS88CH ( 1 ) ,
TPRECIP ( 5,7 )
MAXDP ( 9 ) , MAXTMP ( 7 ) , MEANRH ( 2,8 ) , PRECIP ( 5 ) , PRS88CH ( 10 ) , TPRECIP ( 2 )
60
MINDP, MINTMPj, PRECIP ( 0,1 ) , PRS88CH ( 2 )
120
45
31
MEANRH ( 5,8 ) , PRECIP ( 0,5 ) , PRS3HRC, TPRECIP ( 8 )
MEANRH ( 2,5,8 ) , PRS3HRC, PRS88CH
a
Number of hours of fog observed per day.
Daily maximum temperature.
c
Daily mean relative humidity ( % ) .
d
Daily minimum dew point temperature ( 8C ) .
e
Number of hours of rain observed per day.
f
Total precipitation per day ( mm ) .
g
Daily maximum 3 - h pressure change ( kPa ) .
h
Pressure change ( kPa ) from 8 am to 8 pm.
i
Daily maximum dew point temperature ( 8C ) .
j
Daily minimum temperature ( 8C ) .
b
March 31, 1996, ED visit data were collected prospectively
for both hospitals. During this period, information on
personal characteristics, coincident exposures (e.g. smoking ) , and effect modifiers (e.g. outdoor activity patterns,
disease severity ) was obtained from the clinical record and
directly from patients during or after their visit as described
in detail elsewhere (Stieb et al., 2000 ) .
Emergency Department Visit Data
ED visit data were obtained from the two hospital EDs for
the following cardiorespiratory conditions: angina pectoris,
myocardial infarction, conduction disturbance, dysrhythmia, congestive heart failure, asthma, chronic obstructive
pulmonary disease (COPD ) , and respiratory infections
( including bronchitis, bronchiolitis, croup, and pneumonia ) . Data were obtained in three blocks. For the period July
1, 1992 to June 30, 1994, a database of cardiorespiratory ED
visits from St. Joseph's Hospital was assembled by a health
records coder based on a manual review of all ED visit
records for this period. For the same period, visit data for
Saint John Regional Hospital were extracted from the
hospital's ED visit database. For the period July 1, 1994 to
Environmental Data
Coefficient of haze ( COH ), carbon monoxide (CO ),
hydrogen sulfide (H2S ) , nitrogen dioxide ( NO2 ) , ozone
(O3 ), sulfur dioxide ( SO2 ), and total reduced sulfur (TRS )
were measured hourly throughout the study period by the
New Brunswick Department of the Environment. Daily
Table 2. Summary data on ED visits.
Diagnosis
Number
of visits
Visits
per day
Percent
admitted to
hospital
Age
Mean
( SD )
Number ( % )
< 16 years old
Number ( % )
65 years old
Angina / myocardial infarction
2435
1.8
90.7
65.4
( 13.1 )
0 ( 0.0 )
1300 ( 53.4 )
Congestive heart failure
1312
1.0
74.4
75.6
( 10.7 )
2 ( 0.2 )
1127 ( 85.9 )
Dysrhythmia / conduction disturbance
1096
0.8
41.0
62.8
( 18.5 )
9 ( 0.8 )
595 ( 54.3 )
All cardiac
4843
3.5
76.0
67.6
( 14.8 )
11 ( 0.2 )
3022 ( 62.4 )
Asthma
4771
3.5
14.9
22.8
( 19.0 )
2115 ( 44.4 )
170 ( 3.6 )
Chronic obstructive pulmonary disease
Respiratory infections
1761
8446
1.3
6.2
44.1
17.7
68.0
30.5
( 11.7 )
( 27.1 )
0 ( 0.0 )
3345 ( 39.7 )
1137 ( 64.6 )
1332 ( 15.8 )
All respiratory
14,978
10.9
19.8
32.5
( 26.9 )
5460 ( 36.5 )
2639 ( 17.6 )
Total
19,821
14.4
33.0
41.0
( 28.8 )
5471 ( 27.6 )
5661 ( 28.6 )
462
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Air pollution, allergens and emergency department visits
Stieb et al.
Table 3. Descriptive summary of environmental data.
Variable
Units
April / May to September / Octobera
All year
n
Mean
SD
95th
Percentile
Max
n
Mean
SD
95th
Percentile
Max
COMEANb
ppm
1278
0.5
0.3
1.1
2.1
551
0.6
0.3
1.1
2.1
COMAXc
ppm
1278
1.6
1.1
3.5
8.6
551
1.7
0.9
2.9
7.1
H2SMEAN
H2SMAX
ppb
ppb
1179
1179
1.1
3.4
1.5
3.2
3.7
9.0
21.0
37.0
421
421
0.8
2.8
1.2
2.5
3.0
8.0
8.0
19.0
NO2MEAN
ppb
1370
8.9
5.5
19.0
35.0
551
10.0
5.1
19.0
32.0
NO2MAX
ppb
1370
20.2
10.6
39.0
82.0
551
21.1
10.0
38.0
82.0
O3MEAN
ppb
1370
22.9
7.6
36.0
49.0
551
22.4
8.3
39.0
49.0
O3MAX
ppb
1370
31.4
8.8
48.0
77.0
551
32.8
10.7
54.0
77.0
SO2MEAN
ppb
1370
6.7
5.6
18.0
60.0
551
7.6
5.2
18.0
29.0
SO2MAX
ppb
1370
23.8
21.0
62.0
161.0
551
25.4
17.8
62.0
137.0
TRSMEAN
TRSMAX
ppb
ppb
1308
1308
1.0
4.6
2.8
14.9
7.0
18.0
23.0
247.0
489
489
1.2
5.1
2.7
12.4
8.0
18.0
16.0
108.0
PM10
g / m3
1260
14.0
9.0
31.4
70.3
551
16.2
10.2
35.2
70.3
PM2.5
g / m3
1260
8.5
5.9
20.5
53.2
551
9.5
6.8
22.8
53.2
H+
nmol / m3
1051
25.7
36.8
97.8
284.0
515
38.6
47.0
142.0
284.0
SO42
nmol / m3
1117
31.1
29.7
89.1
208.0
520
38.7
37.5
114.2
208.0
COHMEAN
103 ln ft
1278
0.2
0.2
0.5
1.5
551
0.2
0.2
0.5
0.7
COHMAX
103 ln ft
1278
0.6
0.5
1.7
4.7
551
0.7
0.6
1.8
4.4
Ascomycetes
Basidiomycetes
spores / m3
spores / m3
399
399
249.0
124.4
498.8
339.7
950.0
482.4
5696.1
4262.3
Ganoderma
spores / m3
399
65.5
115.3
285.0
751.6
Deuteromycetes
spores / m3
399
306.5
331.8
1026.1
1927.6
Alternaria
spores / m3
399
1.7
3.4
6.9
28.8
Cladosporium
spores / m3
399
297.4
326.5
1017.6
1883.9
Epicoccum
spores / m3
399
1.4
2.4
6.4
12.8
Small round
spores / m3
399
67.1
183.6
262.5
3214.4
Ferns
Grasses
pollen grains / m3
pollen grains / m3
399
399
0.6
2.3
1.8
5.0
3.2
14.0
21.6
30.9
Trees
pollen grains / m3
399
93.0
331.5
436.4
3369.7
Weeds
pollen grains / m3
399
1.2
3.8
5.4
53.9
NA
a
April to October for aeroallergens, May to September for others.
Daily mean concentration.
c
Daily maximum concentration.
b
average particulate matter concentrations were measured by
Environment Canada (Brook et al., 1995, 1997 ) starting in
1992. One site was in operation throughout the study period
and up to three special study sites operated on a temporary
basis in 1994 and 1995. PM2.5, PM10, and fine fraction
hydrogen ( H + ) and sulfate ion (SO42 ) were measured.
Three alternative schemes were used to classify exposure:
averaging all available monitors, matching postal code
zones ( first three digits of the postal code ) to the nearest
monitor, and matching postal code zones to the nearest
upwind monitor.
Data on temperature, dewpoint temperature, relative
humidity, humidex, precipitation, pressure change, fog
occurrence, and wind direction, were based on measureJournal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
ments at the Saint John airport, located about 10 km east of
the city. These data were obtained from the Environment
Canada archive.
Aeroallergen data (pollen grains and fungal spores )
were collected using rotation impaction sampling equipment operating at 2400 rpm set to collect 1 min from every
10 -min period over a 24- h interval. Data were collected at
two sites from April through October, 1994 and 1995, and
daily values were averaged over the two sites. Individual
pollen grains and fungal spores were collapsed into the
following classes for analysis: Ascomycetes, Basidiomycetes, Deuteromycetes, small round spores (less than 8± 10
m ), ferns, grasses, trees and weeds. In addition to these
groups, individual Basidiomycetes ( Ganoderma) and
463
464
Stieb et al.
Table 4. Pearson correlations among daily mean concentrations of air pollutants and aeroallergens Ð all year.
H2 S
CO
NO2
O3
SO2
TRS
PM2.5 H +
PM10
H2 S
NO2
0.10
0.68#
0.07*
O3
0.05
0.03
0.02
SO2
0.31#
0.01
0.41#
#
#
TRS
0.07*
0.12
0.16
0.05
0.08**
0.28#
0.11#
0.35#
0.18#
0.36#
0.13#
PM2.5
#
#
0.35
#
#
#
0.11#
0.90#
0.25
#
#
0.02
0.42#
0.51#
#
H
COH
Ascomycetes Basidiomycetes Deuteromycetes Small
0.27
#
0.23
0.09**
#
0.26
0.24
0.30
0.10#
0.26
0.31#
0.001
0.18#
0.48
0.30#
0.57#
0.31#
0.85#
0.22#
0.17#
Ascomycetes
0.09
0.03
0.01
0.06
0.08
0.01
0.04
0.01
0.01
0.002
Basidiomycetes
0.04
0.08
0.03
0.16**
0.08
0.09
0.02
0.02
0.06
0.02
0.11*
0.1*
Deuteromycetes
0.02
0.15*
0.04
0.15**
0.07
0.002
0.06
0.07
0.02
0.02
0.01
0.29#
Ferns
Grass
Trees
0.14**
0.08
0.01
0.12*
0.01
0.13*
#
0.02
0.07
0.01
0.1
0.04
0.003
0.01
0.01
0.13**
#
0.66
0.39#
0.23#
0.06
0.11*
0.12*
0.15**
0.03
0.28
0.11#
0
#
0.06
#
Grass
0.07
0.16**
0.13**
0.1*
0.06
0.03
0.28
0.24
0.08
0.15**
0.13*
0.12*
0.04
0.21#
0.11*
0.1
Trees
Weeds
0.06
0.01
0.02
0.1
0.08
0
0.02
0.02
0.15**
0.01
0.13**
0.1*
0.01
0.17#
0.02
0.14#
0.03
0.01
0
0.04
0.03
0.06
0.01
0
0.09
0.14**
0.03
0.24
0.02
0.01
0.08
0.18#
*p < 0.05, **p < 0.01, #p < 0.001.
#
0.05
0.1*
0.01
0.15** 0
Air pollution, allergens and emergency department visits
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
0.33
0.49#
0.06
#
0.31
0.08*
0.17#
0.02
#
0.19
0.27
0.55#
Small Round
#
0.11
SO4
COH
2
Ferns
0.10#
PM10
+
SO4
#
Air pollution, allergens and emergency department visits
Stieb et al.
Table 5. Percent increase in ED visits associated with mean concentration of pollutant, by season and diagnosis group, based on single - pollutant models.
Pollutant
All year
Metric
May to September
Percent increase
ED visits
p - Value
Metric
Percent increase
ED visits
p - Value
Cardiac
CO
max ( 1,3 ) a
4.3
0.079
max ( 1,9 )
7.1
H2 S
max ( 1,9 )
7.5
0.0001
max ( 1,3 )
8.5
0.027
NO2
mean ( 1,2 )
3.9
0.136
mean ( 1,5 )
10.1
0.051
O3
max ( 1,6 )
16.6
0.013
max ( 1,6 )
21.4
0.012
SO2
TRS
max ( 1,8 )
max ( 1,0 )
4.9
0.6
0.002
0.234
max ( 5,6 )
max ( 1,1 )
2.8
2.4
0.067
0.006
PM10
mean ( 1,3 )
8.2
0.003
mean ( 1,3 )
9.7
0.022
PM2.5
mean ( 1,3 )
4.9
0.055
mean ( 1,3 )
4.4
0.235
ln ( mean ( 1,3 ) ) b
5.4
0.038
mean ( 5,4 )
1.8
0.010
mean ( 5,4 )
2.8
0.008b
mean ( 1,4 )
6.0
0.001
mean ( 1,4 )
6.8
0.014
max ( 1,7 )
5.4
0.027
max ( 1,8 )
8.4
0.049
H+
SO4
2
COH
0.169
Respiratory
CO
mean ( 1,7 )
3.4
0.041
max ( 1,0 )
5.7
0.142
H2 S
mean ( 5,3 )
1.0
0.024
max ( 5,5 )
3.4
0.004
NO2
max ( 1,0 )
3.8
0.070
max ( 1,8 )
11.5
0.017
ln ( max ( 1,8 ) ) b
5.1
0.011b
0.011
O3
mean ( 5,1 )
4.0
0.012
mean ( 5,3 )
5.7
SO2
max ( 1,5 )
3.9
0.0005
max ( 5,3 )
3.9
0.003
TRS
PM10
max ( 1,10 )
mean ( 5,3 )
0.7
2.4
0.013
0.013
mean ( 5,5 )
mean ( 5,4 )
1.1
3.0
0.073
0.035
PM2.5
mean ( 5,3 )
1.9
0.028
mean ( 1,7 )
6.3
0.028
H+
mean ( 5,1 )
2.3
0.0005
mean ( 5,1 )
3.5
0.0004
SO42
mean ( 5,6 )
1.6
0.034
mean ( 5,6 )
2.2
0.032
COH
mean ( 1,7 )
3.2
0.030
max ( 5,2 )
4.2
0.009
a
Numbers in parentheses indicate the number of days over which the metric is averaged and the number of days lagged relative to the ED visit.
b
Percent increase in ED visits calculated for a change in concentration equal to the mean, symmetrical about the mean, i.e., ( / 2 + ) / ( / 2 ) = 3.
Deuteromycetes ( Alternaria, Cladosporium, Epiccocum )
species were examined.
Data Analysis
We used generalized additive models ( Hastie and
Tibshirani, 1990 ) in S -Plus ( MathSoft Inc., 1997 ) to
examine the relationship between daily variations in
environmental factors and cardiorespiratory ED visits. In
this type of model, the residual variance is proportional to
the expected response, thus accommodating over ( under )
dispersion relative to Poisson variation. We used a
nonparametric smoothed function of day of study
( LOESS ) ( Cleveland and Devlin, 1988 ) to prefilter the
time series of both ED visits and environmental variables.
Several candidate window lengths ( 19, 31, 45, 60, 91,
120, 183, 365 days ) for the LOESS filter were considered
for each diagnosis group. The optimum window length
was determined by minimizing the value of Akaike's
Information Criterion ( AIC ) and maximizing the Bartlett's test p -value, as well as examining plots of the
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
autocorrelation function and serial periodograms for each
window length. The AIC is the deviance penalized for the
number of parameters being estimated and Bartlett's test
assesses the presence of serial correlation in the residuals
(Bartlett, 1978 ) . In examining the association between a
given pollutant or aeroallergen and diagnosis group, the
optimum window length for that diagnosis group was also
applied to the pollution or aeroallergen variable. Aeroallergens were only examined in association with asthma
ED visits, given the allergic basis of this condition relative
to the other cardiorespiratory diagnoses. Differences in
ED visit rates by day of the week were removed by
including a term in the filter for day of week. Possible
nonlinearity of air pollution effects was examined by
comparing the fit obtained using a linear air pollution
term with that obtained using a spline smoother of the air
pollution variable. A probability of the F -statistic of less
than 0.05 was interpreted as evidence of nonlinearity, and
alternative functional forms were assessed based on their
model AIC.
465
466
Stieb et al.
Air pollution, allergens and emergency department visits
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Figure 1. Percent increase in ED visits at mean pollutant concentration, by specific cardiac diagnosis group: ( a ) pollutant gases, ( b ) particulate matter. ( ) Myocardial infarction / angina;
( 5 ) congestive heart failure; ( ) dysrhythmia / conduction disturbance; ( 4 ) total cardiac.
Air pollution, allergens and emergency department visits
To account for possible confounding effects of weather
on the association between air pollution or aeroallergens and
ED visits, we included LOESS smooth functions (span of
50% ) of selected weather variables in each model. This
approach allows for considerable flexibility in characterizing the shape of the association between weather and ED
visits. We used a forward stepwise regression procedure to
identify the smallest number of weather variables required
to predict ED visits. For each variable, lags of up to 10 days
were considered. The weather model with the lowest AIC
was selected. Final model specifications are summarized in
Table 1 by diagnostic group.
Possible cumulative effects of air pollution and
aeroallergens on ED visits were assessed by considering
5 -day moving averages, in addition to single -day values.
Air pollution and aeroallergens on the same day as the ED
visit, and the 10 preceding days, were considered. Thus,
single -day values up to lag 10 and 5- day averages up to
lag 6 were considered. The averaging time and lag with
the largest T-value (ratio of parameter estimate to its
standard error ) , positive or negative, was considered for
further analysis. Our decision to examine longer lags than
have typically been considered in earlier studies was
based on our observation that over 20% of ED visitors for
COPD and respiratory infections reported experiencing
symptoms for seven or more days before their ED visit
( Stieb et al., 2000 ) .
The overall effect of air pollution and aeroallergens
respectively was examined by constructing multipollutant
and multi -aeroallergen models using a backward stepwise
procedure. In multipollutant models, each measure of
particulate matter ( PM10, PM2.5, H + , SO42 , COH ) was,
in turn, allowed to compete for inclusion with the other
pollutants. The final multivariate model was that model
associated with the lowest AIC. The percent excess ED
visits associated with a given pollutant or aeroallergen
was calculated by taking the antilog of the parameter
estimate multiplied by the mean concentration of that
variable.
The primary analysis consisted of examining the yearround association between air pollution and ED visits
grouped as all respiratory and all cardiac. Sensitivity
analyses were conducted by season (all year vs. May to
September ) , specific diagnosis groups, exposure classification scheme, selected individual characteristics or coincident exposures that might act as effect modifiers, and time
of onset of symptoms. Sensitivity analysis based on
alternative exposure classification schemes was carried out
by generating parameter estimates for each postal code zone
and pooling these using a random effects model ( Dersimonian and Laird, 1986 ) . We used the expanded dataset from
1994 to 1996 (Stieb et al., 2000) to define subgroups based
on possible effect modifiers. These subgroups included
visits resulting in admission to hospital, patients who
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Stieb et al.
smoked, patients whose household included a smoker,
patients who spent less than 2 h or more than 4 h outdoors
per day during the week preceding their visit, mild versus
moderate or severe asthmatics, asthma patients who did not
report use of asthma medications, and patients whose
presenting complaint ( provided to the registration clerk
upon entry to the ED ) was asthma. The latter was selected
because an earlier study found a significant association
between ozone and asthma ED visits among those with a
presenting complaint of asthma ( Stieb et al., 1996 ) . We also
used these expanded data to construct two alternative time
series based on the date of onset of earliest and most
responsible symptoms. Earliest symptoms were defined as
those that appeared first, and most responsible symptoms as
those that prompted the ED visit. We then examined the
association between air pollution and onset of earliest and
most responsible symptoms.
Results
Descriptive information on ED visits is presented in
Table 2. Respiratory visits accounted for about 3 /4 of all
cardiorespiratory visits and respiratory infections accounted for nearly one half. Visits for cardiac conditions
and COPD were most likely to result in admission to
hospital and cardiac patients and those with COPD were
on average considerably older. Descriptive information
on air pollutants and aeroallergens is presented in Table
3 and correlations among unfiltered environmental
variables are presented in Table 4. Considerably stronger
correlations were observed between ozone and NO2
(0.31) , SO2 ( 0.28) , PM10 (0.34) , PM2.5 (0.42 ), and
SO42
(0.43 ), during the May to September period.
Correlations among filtered and unfiltered environmental
variables were similar.
The strongest associations for each pollutant with all
cardiac and all respiratory ED visits based on single pollutant models are summarized in Table 5 by season.
H2S, O3, SO2 and PM10 exhibited strong (p0.01)
positive associations with cardiac visits in year-round
models, whereas H + and SO42 exhibited strong negative
associations. In warm season models, O3 and TRS
exhibited strong positive associations, and H + and
SO42
strong negative associations. Strong positive
associations with respiratory visits in year- round models
were observed for O3, SO2, TRS and PM10, whereas a
strong negative association was observed for H + . In warm
season models, O3 and SO2 exhibited strong positive
associations, and H + , H2S and COH, strong negative
associations. Significant (p < 0.05) nonlinearity of the
effect of PM2.5 on year-round cardiac visits and of NO2
on summertime respiratory visits was observed, and
logarithmic transformation strengthened these associations.
467
468
Stieb et al.
Air pollution, allergens and emergency department visits
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Figure 2. Percent increase in ED visits at mean pollutant concentration, by specific respiratory diagnosis group: ( a ) pollutant gases, ( b ) particulate matter. ( A ) Asthma; ( & ) COPD;
( o ) respiratory infection; ( ~ ) total respiratory.
Air pollution, allergens and emergency department visits
Stieb et al.
Table 6. Percent increase in ED visits associated with mean concentration of pollutant, by season and diagnosis Ð final multipollutant models.
All year
Pollutant
May to September
Metric
Percent increase p - Value
ED visits
95%CI
Pollutant
Metric
Percent increase p - Value
ED visits
95%CI
Cardiac
O3
max ( 1,6 ) a
16.0
0.017
2.8
30.9
CO
max ( 1,9 )
10.1
0.041
0.4
SO2
max ( 1,8 )
4.9
0.003
1.7
8.2
NO2
mean ( 1,5 )
9.1
0.076
0.9
20.1
O3
max ( 1,6 )
17.2
0.042
0.6
36.6
max ( 1,1 )
2.7
0.002
1.0
4.5
mean ( 1,4 )
7.3
0.007
12.2
2.1
8.6
Total
20.9
TRS
SO4
2
Total
20.7
31.8
Respiratory
NO2
max ( 1,0 )
3.6
0.083
7.5
0.5
ln ( NO2 )
max ( 1,8 )
4.7
0.017
0.8
O3
mean ( 5,1 )
3.9
0.015
0.8
7.2
O3
mean ( 5,3 )
3.9
0.097
0.7
8.7
SO2
max ( 1,5 )
3.7
0.0008
1.5
6.0
SO2
max ( 5,3 )
3.9
0.006
1.1
6.7
COH
max ( 5,2 )
4.9
0.002
7.9
1.8
Total
4.0
Total
7.6
a
Numbers in parentheses following mean or max refer to the number of days over which the metric is averaged and the number of days lagged relative to the
ED visit.
The significant, negative associations of various pollutants
with cardiac and respiratory ED visits in the year-round
analysis were not altered appreciably when alternative
window lengths were used to construct the LOESS filter
applied to ED visit and air pollution data.
Results from year-round analysis by specific diagnosis
are presented in Figures 1 and 2 for cardiac and respiratory
conditions respectively. Results were heterogeneous for the
cardiac diagnoses. Only dysrhythmia visits were positively
associated with all measures of particulate matter. There
were positive effects of all pollutants except NO2 and COH
on asthma visits, of all gaseous pollutants on COPD visits,
and on all respiratory diagnosis groups for O3, SO2, PM10,
PM2.5, and SO42 .
Results from multipollutant models are presented in
Table 6. O3 was present in all models and was associated
with a positive and statistically significant ( p< 0.05 ) or
nearly significant effect in all cases. SO2 was present in
all but one of the final models and was also associated
with a positive and statistically significant effect in all
cases. NO2 exhibited a positive and significant or nearly
significant effect in the summer for both cardiac and
respiratory visits, whereas it was associated with a
negative and nearly significant effect on year-round
respiratory visits. Other pollutants appeared less frequently in the final models. In all -year models for both
cardiac and respiratory ED visits, there was little
difference in the magnitude or statistical significance of
effects between single and multipollutant models. In the
multipollutant model for summertime cardiac visits, the
effects of CO, TRS, and SO4 increased in magnitude and
statistical significance relative to single -pollutant models,
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
whereas those of O3 and NO2 decreased. In the
multipollutant model for summertime respiratory visits,
the effect of COH increased in magnitude and statistical
significance relative to single -pollutant models, whereas
those of NO2, O3 and SO2 decreased. The effect of ozone
in multipollutant models was not reduced appreciably
when individual metrics of particulate matter were forced
in, although in some cases it increased dramatically in the
presence of H + . The effect of SO2 was reduced by 1 /4 to
1 /3 relative to the results shown in Table 6, depending on
which metric of particulate matter was forced in. The
choice of final all - year multipollutant model for cardiac
visits and summertime multipollutant model for respiratory visits was not influenced by the substitution of
ln ( PM2.5 ) or ln ( NO2 ) respectively in place of linear
terms for these pollutants, although the model with log
transformed NO2 had a slightly lower AIC than the model
with a linear term for NO2.
The shape of the concentration response function for
each pollutant from the final year-round multipollutant
models is shown in Figures 3 and 4, based on spline plots of
the association. None of the plots suggests deviation from
linearity, and this was confirmed by statistical tests for
nonlinearity, as discussed earlier.
Associations between aeroallergens and asthma ED visits
in single variable models are summarized in Table 7.
Among fungal spores, a strong positive association was
observed with Ascomycetes whereas among pollens, a
strong positive association was observed with weeds, and
grasses. A strong negative association was observed with
Epicoccum. The final multi - aeroallergen model is also
shown in this table. The effect of Ascomycetes was reduced
469
Stieb et al.
Air pollution, allergens and emergency department visits
Figure 3. Spline plots of association between air pollution and cardiac ED visits, based on the best - fitting multipollutant model.
slightly in magnitude, but increased in statistical significance, whereas those of Alternaria and small round spores
increased both in magnitude and significance. All of those
aeroallergens that exhibited negative effects in single
variable models exhibited weaker effects in the multi aeroallergen model, with the exception of Ganoderma.
Neither weed nor grass pollen was selected for inclusion in
the final multi -aeroallergen model. The effects of pollens
and fungal spores were not affected by adjustment for air
pollutants, added to the model either singly or in
combination, with the exception of the negative effects of
470
Epicoccum, Ganoderma and ferns, the strength of which
was reduced considerably in the presence of H2S.
The use of alternative air pollution exposure
classification schemes did not result in more precise
estimates of effect, nor did we observe statistically
significant heterogeneity among results for individual
postal code zones.
Few findings from subgroup analyses were remarkable.
Of note, however, was an increase in size of the PM10 effect
on respiratory visits, from 2.4% to 3.9% ( p= 0.094 ) and
4.3% ( p= 0.022 ) in smokers and those with a smoker in the
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Stieb et al.
Figure 4. Spline plots of association between air pollution and respiratory ED visits, based on the best - fitting multipollutant model.
Air pollution, allergens and emergency department visits
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
471
Stieb et al.
Air pollution, allergens and emergency department visits
Table 7. Percent increase in asthma ED visits associated with mean concentration of aeroallergens ( April to October, 1994 ± 1995 ) .
Variable
Metric
Single aeroallergen model
Percent increase ED visits
Multi - aeroallergen model
p - Value
Percent increase ED visits
p - Value
95%CI
Ascomycetes
( 5,6 ) a
5.1
0.003
4.5
0.001
1.8
7.4
Basidiomycetes
Ganoderma
( 1,10 )
( 5,5 )
4.9
3.2
0.071
0.040
4.7
0.002
7.6
1.8
Deuteromycetes
( 1,9 )
6.3
0.220
4.7
0.013
1.0
8.6
Alternaria
( 1,1 )
4.5
0.021
Cladosporium
( 1,9 )
6.1
0.224
( 1,10 )
6.6
0.009
5.2
0.046
10.1
0.1
Small round spores
Epicoccum
( 5,3 )
2.7
0.024
3.0
0.006
0.8
5.1
Ferns
( 1,10 )
1.9
0.090
2.1
0.116
4.6
0.5
Grass
Trees
( 5,0 )
( 1,6 )
3.6
3.7
0.010
0.027
3.6
0.035
6.8
0.3
Weeds
( 5,4 )
2.4
0.002
a
Numbers in parentheses refer to the number of days over which the metric is averaged and the number of days lagged relative to the ED visit.
household, respectively. There was also a near doubling of
the effect of PM10 on excess asthma visits among those with
a presenting complaint of asthma, respectively from 6.4%
( p= 0.025 ) to 12.7% (p =0.007) .
Using the expanded data we collected for the period
1994 ±1996, we determined that 62.3% of patients
whose ED visit was attributed to asthma reported
``asthma'' as their presenting complaint ( i.e., identified
Figure 5. ( A-D ) Percent increase in cardiac and respiratory ED visits at mean pollutant concentration, by days lagged, for ozone and sulfur
dioxide.
472
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Air pollution, allergens and emergency department visits
``asthma'' as the reason for their ED visit when
registering in the ED ) . Those with a presenting
complaint of ``asthma'' were more likely to have been
previously diagnosed (93.6% vs. 59.2%, p= 0.001 ), and
to be classified based on their past history as having
moderate or severe asthma ( 32.5% vs. 17%, p= 0.001 ).
Health care utilization in the previous year and previous
medication use were also more prevalent among those
with a presenting complaint of asthma: 33.8% vs.
25.9% ( p =0.005 ) with more than four physician office
visits, 39.2% vs. 21.6% with more than two ED visits
( p= 0.001 ), 64.1% vs. 50.9% (p =0.001 ) ever admitted
to hospital, 81.8% vs. 71.7% (p =0.001 ) had previously
used inhaled steroids, and 64.9% vs. 45.2% ( p= 0.001 )
had previously used oral steroids. The former group was
also less likely to smoke or to have a smoker in the
household ( 33.2% vs. 39.5%, p= 0.044 and 50.4% vs.
55.4%, p =0.036 respectively ).
A smaller dataset was available for our analysis of
the association of air pollution with onset of symptoms,
because these data were only collected between 1994
and 1996. When we constructed models of year-round
respiratory ED visits that simultaneously contained
single -day air pollution values lagged 0 to 10 days,
we found that the strongest single- day effects for the
series based on date of onset of earliest and most
responsible symptom for PM2.5 and PM10 were for a
lag of 0 days, whereas for the series based on date of
visit, the strongest single- day effect was for a lag of 4
days. There was a gradient in size of effect and
strength of association, with the strongest association
for both PM2.5 and PM10 being for onset of most
responsible symptom (9.2%, p= 0.002 and 10.5%,
p =0.001, respectively ), followed by earliest symptom
( 7.5%, p =0.007 and 8.5%, p= 0.004 ) and date of visit
( 5.1%, p= 0.01 and 6.0%, p =0.008) . No discernible
pattern of this kind was observed for other pollutants or
for cardiac visits.
Discussion
We observed a significant influence of the air pollution mix
on both cardiac and respiratory ED visits. Gaseous
pollutants, especially ozone, exhibited a more consistent
association with ED visits than particulate matter, in that
ozone was always selected and SO2 was selected in most
instances for inclusion in the final multipollutant models of
all cardiac and all respiratory ED visits, whereas the various
particulate matter metrics were not. Because multipollutant
models were constructed based on days when data were
available for all candidate variables, the smaller number of
observations for the various particulate matter metrics could
not account for their failure to persist consistently through
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Stieb et al.
the stepwise selection procedure. The magnitude of effect of
individual pollutants on year- round ED visits was not
altered appreciably in multipollutant models compared to
the effects observed in single -pollutant models.
Other investigators have also examined air pollution
effects on ED visits using multipollutant models. In a recent
study in London, United Kingdom, Atkinson et al. (1999 )
found that although both SO2 and PM10 were significantly
associated with respiratory ED visits, neither appeared to
dominate the other in two pollutant models. In the same
study, the effect of CO on respiratory visits among the
elderly was not sensitive to the inclusion of other pollutants
in the model. Examination of asthma visits by children in
this study revealed that NO2, SO2 and PM10 exhibited the
strongest effects, although they were reduced somewhat in
two -pollutant models. Weaker effects were seen for CO and
Black Smoke, whereas no effect was seen for ozone. In
another study in London, Buchdahl et al. (1996 ) found a
significant association between ozone and acute wheezy
episodes in children, which persisted when NO2 and SO2
were included simultaneously in the model. Although a
significant association with SO2 was also noted in this study
in single- pollutant models, the impact of adjustment for the
other pollutants was not reported. The effects of particulate
matter were not addressed in this study.
In a study in New Jersey, ozone was the strongest
predictor of asthma ED visits among air pollutants selected
using a stepwise approach in which SO2 and PM10 were also
considered, neither of which contributed additional explanatory power (Cody et al., 1992 ) . In a two -pollutant
model of respiratory ED visits in Montreal, Quebec, among
those 65 years of age or older, Delfino et al. (1997, 1998 )
found that the significant, positive effect of ozone was
somewhat reduced relative to a single- pollutant model,
whereas that of PM2.5 was reduced by about one half and
became nonsignificant. In the same study, in a two pollutant model of respiratory visits for children under
two, the significant, positive effect of hydrogen ion was
essentially unchanged compared to single- pollutant results,
whereas that of sulfate was reduced considerably and
became nonstatistically significant. Goldsmith et al. (1983 )
utilized path analysis to examine the influence of air
pollution on ED visits for all conditions in southern
California, and found significant effects of sulfate and
oxidants taking into account the effects of other pollutants.
In one of only two other studies we identified that examined
effects of reduced sulfur compounds on ED visits,
Jorgensen et al. ( 1996 ) found that the logarithm of TRS
was significantly associated with respiratory ED visits in
Prince George, British Columbia, whether or not total
suspended particulate (TSP ) was also included in the
model. This study modeled the effects of air pollution as
subacute, using a latent gamma Markov process. In their
study of wintertime asthma ED visits in Santa Clara County,
473
Stieb et al.
California, Lipsett et al. (1997 ) found that the significant
positive effect of PM10 was unchanged when NO2 was
simultaneously included in the model, whereas the effect of
NO2 was reduced and became non significant. No effect of
ozone was observed. Romieu et al. (1995 ) found that
although both ozone and SO2 were significantly associated
with pediatric asthma ED visits in Mexico City, the effect of
ozone was essentially unchanged in models with SO2 and
every sixth day TSP, whereas the effect of SO2 was reduced.
No association was found with NO2 or TSP.
In another study in Mexico city, which also considered
aeroallergens (see below ) , ozone was positively associated
with asthma ED visits by adults in the dry season (January
to April, November, December ) in a model that also
included grass pollen and SO2, whereas SO2 was positively
associated with visits by seniors in the wet season ( May to
October ) in a model that also included PM10 and tree pollen
( Rosas et al., 1998 ) . In a study by Rossi et al. ( 1993 ) in
Oulu Finland, which also examined the effects of both
pollens and air pollution on asthma ED visits, NO2 was the
only significant predictor among pollutants, based on a
stepwise procedure that also considered H2S, SO2 and TSP.
Our earlier study in Saint John revealed that the significant
positive effect of ozone on summertime asthma ED visits
was unchanged in multipollutant models including SO2 and
NO2 in one model, and every sixth day sulfate and TSP in
another ( Stieb et al., 1996 ) . In their study in Barcelona,
Spain, Sunyer et al. (1993 ) found significant positive
associations with ED visits for COPD for both SO2 and
black smoke (BS ), which persisted when the two pollutants
were considered together. No other pollutants were
considered in this study. In a later meta -analysis of asthma
ED visit studies in Barcelona, Helsinki, London and Paris,
Sunyer et al. (1997 ) found positive and significant effects
of NO2 among adults, and of SO2 among children, both of
which remained unchanged when adjusted for BS. Effects
of ozone were heterogeneous among cities for adults, with
no effect in children, and the effect of BS was positive but
not significant for both adults and children. Finally, in their
study in Valencia, Spain, Tenias et al. (1998 ) found
significant positive effects of NO2 and ozone on asthma
visits, which persisted when adjustment was made for the
effects of the other pollutant, as well as BS and SO2. In
single -pollutant models in this study, BS and SO2 were
positively associated with visits, but the associations were
not statistically significant. Thus, in this group of studies as
in the present study, significant positive associations of
respiratory ED visits with ozone appear to be relatively
insensitive to adjustment for the effects of other pollutants,
whereas results have been less consistent for NO2, SO2 and
other pollutants.
Although the magnitude of the association with ozone
was larger for cardiac than respiratory visits in our study,
the effects of both ozone and other pollutants were more
474
Air pollution, allergens and emergency department visits
heterogeneous among the specific cardiac conditions than
among respiratory conditions. This may be attributable to
the much smaller number of cardiac visits, which would
tend to make effect estimates unstable. Although, to our
knowledge, no other studies have examined the association of ozone or other pollutants with cardiac ED visits,
other endpoints such as cardiac hospital admissions have
been examined. Results have been mixed in terms of
effects of ozone. Some investigators have found that
ozone was significantly associated with cardiac hospital
admissions (Burnett et al., 1997a; Ponka and Virtanen,
1996 ), whereas several others have found no association
or a borderline significant association ( Burnett et al.,
1995, 1997b, 1999; Morris et al., 1995; Poloniecki et al.,
1997; Schwartz, 1997; Schwartz and Morris, 1995 ).
Although the respiratory pathophysiology of ozone is
relatively well understood (Crapo et al., 1992; Devlin et
al., 1991; Vincent et al., 1997 ) , effects on the cardiovascular system are not. Cardiovascular effects may occur by
way of a nonspecific insult, perhaps mediated through the
respiratory system ( Bouthillier et al., 1998) . One factor
that could contribute to a stronger association for cardiac
than respiratory visits in our series is greater diagnostic
accuracy for cardiac than respiratory conditions, which we
observed in an earlier analysis (Stieb et al., 1998a ).
Greater noise in assignment of respiratory diagnoses in
the ED could attenuate observed associations in this
group, particularly for respiratory infections, where we
observed the poorest inter-rater agreement on diagnosis.
At the same time, we did not observe a bias in the
assignment of cardiorespiratory diagnoses in the ED in
relation to air pollution ( Stieb et al., 1998a) . An
additional factor could be greater vulnerability to air
pollution and other insults among those with cardiac
conditions, who were on average much older than those
with respiratory conditions.
The strong negative association between hydrogen ion
and ED visits was surprising given that relatively high
peak levels were observed, although results from
epidemiologic studies of the effects of hydrogen ion
have been mixed. Significant positive associations with
various adverse health outcomes have been reported by
some investigators (Burnett et al., 1997a; Delfino et al.,
1997; Dockery et al., 1996a; Lippmann and Ito, 1995;
Neas et al., 1995; Raizenne et al., 1996; Thurston et al.,
1994 ), whereas in other studies, no effect was apparent
(Dockery et al., 1996b; Schwartz et al., 1996) , or the
effect of hydrogen ion was weaker than that of other
measures of particulate matter (Dockery et al., 1993 ).
This could be partially mediated by a very weak
correlation between levels measured at centrally located
monitors and personal exposure, which we have
observed in Saint John (Stieb et al., 1998b) , and
which has been reported elsewhere (Brauer et al., 1989;
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Air pollution, allergens and emergency department visits
Suh et al., 1992 ). Although this would tend to drive the
effect of hydrogen ion to the null because of excessive
random exposure misclassification, it would not explain
the negative association observed here. A possible
additional factor is that elevated levels of acid, probably
combined with weather conditions such as fog, result in
perceptibly poor air quality, which could cause people
to delay their ED visit, seek care elsewhere, and /or
close their windows and avoid going outdoors, thus
reducing their exposure. Of note, however, in our series,
were positive and statistically significant associations in
single -pollutant models, between hydrogen ion and ED
visits for asthma and dysrhythmia / conduction disturbance. Of all ED visitors, asthma patients were youngest, least likely to be admitted, and spent the greatest
proportion of time outdoors in the week preceding their
ED visit ( Stieb et al., 2000) .
Results of our examination of the association between
asthma ED visits and aeroallergens differed in some
respects from earlier reports. Rosas et al. (1998 ) and
Salvaggio et al. (1971 ) found significant associations
between asthma ED visits and both pollens and fungal
spores in Mexico City and New Orleans, respectively. In
studies in Ottawa, Canada, San Diego, California, and
State College, Pennsylvania, asthma ED visits, asthma
severity and peak flow decrements, respectively, were
associated with fungal spores but not pollens (Dales et al.,
2000; Delfino et al., 1996; Neas et al., 1996 ). Rossi et al.
( 1993 ) found no association between asthma ED visits and
pollens in Oulu, Finland, and Jones et al. (1995 ) found no
association between either fungal spores or pollens and
respiratory ED visits in Baton Rouge, Louisiana. Levels of
Basidiomycetes and Deuteromycetes ( for which we found
no significant association with asthma ED visits ) were
significantly lower in our study than those reported by
Dales et al. (2000) as well as (Basidiomycetes only )
( Rosas et al. and Neas et al. ) , whereas levels of
Ascomycetes were similar to those reported by Dales et
al. and Rosas et al., but lower than that of Neas et al. Pollen
concentrations in our study appeared to be generally
comparable with those reported for Oulu and considerably
higher than those in Baton Rouge and San Diego. Effects
of pollens and fungal spores in the Baton Rouge study may
have been attenuated by the inclusion of a number of
nonallergic respiratory conditions in the analysis (Jones et
al., 1995 ) . As reported by Dales et al. and Neas et al., we
found that the effect of fungal spores was not influenced by
adjustment for the effect of air pollution.
We examined the effects of individual pollutants based
on single- day and multiday measures, and multiple
possible lags. Although we chose to conduct the analysis
in this fashion to detect possible cumulative effects, and
effects lagged several days before the date of the ED
visit, repeated hypothesis tests could result in spuriously
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Stieb et al.
positive results at conventional levels of statistical
significance. However, the effects of ozone and SO2 in
single -pollutant models, for all -year cardiac and respiratory visits, consistently achieved a more stringent level of
statistical significance (p0.01 ). When we constructed
models containing all individual day lags simultaneously
for each pollutant, thus capturing the independent effect
of each day's air pollution exposure, positive associations
of ozone with both all- year cardiac and respiratory visits
were not restricted to a single -day lag (Figure 5A and
C ). Thus, in the case of ozone, our results do not appear
to reflect a spurious but strongly positive association for
a single day. This figure also illustrates that for all -year
respiratory ED visits, although the single- day effects at
lags 1 to 5 were not significant on their own, the
cumulative effect over 5 days was highly significant in
both single and multipollutant models ( Tables 5 and 6 ).
In the case of SO2, there was less of a consistent positive
effect over multiple days (Figure 5B and D ) .
Our analysis of alternative time series based on time
of onset of symptoms revealed that the strength of the
association between PM2.5, PM10 and respiratory visits
increased relative to analysis of the time series based on
date of visit. This observation would tend to support the
biological plausibility of the temporal association between particulate matter and respiratory morbidity. To our
knowledge, this has not been examined by other
investigators.
Results of subgroup analyses based on selected personal
characteristics and concurrent risk factors should be
interpreted with caution because the small sample size
available for analysis can lead to spuriously negative results,
and repeated hypothesis tests can lead to spuriously positive
results. However, they raise several hypotheses that could
be tested in future studies. In particular, our observation of
greater effects of certain pollutants on patients with a
presenting complaint of asthma warrants further investigation. The essential features of this group appear to be more
severe disease and a greater awareness of their diagnosis.
We are not aware of other evidence of specific at - risk
subgroups of asthmatics. Further examination of effect
modification by coincident exposure to tobacco smoke is
also recommended. In a recent study examining ozone and
asthma ED visits in New York City, a strong association was
found for heavy smokers (defined as 13 pack years ),
whereas none was found for light or never smokers
(Cassino et al., 1999 ) . A somewhat higher risk of airpollution- related mortality was also observed in a longitudinal study in relation to tobacco smoke exposure
(Dockery et al., 1993 ) .
In conclusion, we observed a significant influence of the
air pollution mix on cardiac and respiratory ED visits.
Although in single -pollutant models, positive associations
were noted between ED visits and some measures of
475
Stieb et al.
particulate matter, in multipollutant models, pollutant gases,
particularly ozone, exhibited more consistent effects.
Aeroallergens were also significantly associated with warm
season asthma ED visits. The availability of detailed
information on possible effect modifiers and other factors
permitted us to conduct additional exploratory analyses that
would not have been possible using administrative data
alone.
Acknowledgments
The authors thank Lori Churchill, Lauren Clark, Kelly
Harquail, Shane Rossignol, Cheryl Rossignol, and Wendy
Sleigh of Atlantic Health Sciences Corporation, for
assistance with collection of emergency department data,
Lee Coates of Aerobiology Research Laboratories for
providing aeroallergen data, and Dr. Sabit Cakmak of
Health Canada for advice on S -Plus programming.
References
Atkinson R.W., Anderson H.R., Strachan D.P., Bland J.M., Bremner S.A.,
and Ponce de Leon, A. Short - term associations between outdoor air
pollution and visits to accident and emergency departments in London
for respiratory complaints. Eur Respir J 1999: 13: 257 ± 265.
Bartlett M.S. Correlation or Spectral Analysis. The Statistician 1978: 27:
147 ± 158.
Bouthillier L., Vincent R., Goegan P., Adamson I.Y.R., Bjarnason S.,
Stewart M., GueÂnette J., Potvin M., and Kumarathasan P. Acute
effects of inhaled urban particles and ozone: lung morphology,
macrophage activity, and plasma endothelin - 1. Am J Pathol 1998:
153: 1873 ± 1884.
Brauer M., Koutrakis P., and Spengler J.D. Personal exposures to acidic
aerosols and gases. Environ Sci Technol 1989: 23: 1408 ± 1412.
Brook J., Wiebe H.A., Stieb D.M., and Burnett R.T., Saint John N.B.
Particle health effects study: determination of ambient exposures to
respirable particles. In: Particulate Matter: Health and Regulatory
Issues ( VIP - 49 ) , Proceedings of an International Specialty Conference, April 4 ± 6, 1995, Pittsburgh, PA. Pittsburgh: Air and Waste
Management Association, 1995: 347 ± 357.
Brook J.R., Wiebe A.W., Woodhouse S.W., Audette C.V., Dann T.F.,
Callaghan S., Piechowski M., Dabek - Zlotorzynska E., and Dloughy
J.F. Temporal and spatial relationships in fine particle strong acidity,
sulfate, PM10, PM2.5 across multiple Canadian locations. Atmos
Environ 1997: 47: 2 ± 19.
Buchdahl R., Parker A., Stebbings T., and Babiker A. Association between
air pollution and acute childhood wheezy episodes: prospective
observational study. BMJ 1996: 312: 661 ± 665.
Burnett R.T., Cakmak S., Brook J.R., and 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 1997a: 105: 614 ± 620.
Burnett R.T., Dales R.E., Brook J.R., Raizenne M.E., and Krewski D.
Association between ambient carbon monoxide and hospitalizations
for congestive heart failure in the elderly in 10 Canadian cities.
Epidemiology 1997b: 8: 162 ± 167.
Burnett R.T., Dales R., Krewski D., Vincent R., Dann T., and Brook J.R.
Associations between ambient particulate sulfate and admissions to
476
Air pollution, allergens and emergency department visits
Ontario hospitals for cardiac and respiratory diseases. Am J Epidemiol
1995: 142: 15 ± 22.
Burnett R., Smith - Doiron M., Stieb D., Cakmak S., and Brook J. Effects
of particulate and gaseous air pollution on cardiorespiratory hospitalizations. Arch Environ Health 1999: 54: 130 ± 139.
Cassino C., Ito K., Bader I., Ciotoli C., Thurston G., and Reibman J.
Cigarette smoking and ozone - associated emergency department use
for asthma by adults in New York City. Am J Respir Crit Care Med
1999: 159: 1773 ± 1779.
Cleveland W.S., and Devlin S.J. Robust locally - weighted regression and
smoothing scatterplots. J Am Stat Assoc 1988: 74: 829 ± 836.
Cody R.P., Weisel C.P., Birnbaum G., and Lioy P.J. The effect of ozone
associated with summertime photochemical smog on the frequency of
asthma visits to hospital emergency departments. Environ Res 1992:
58: 184 ± 194.
Crapo J., Miller F.J., Mossman B., Prior W.A., and Kiley J.P. Relationship
between acute inflammatory responses to air pollutants and chronic
lung disease. Am Rev Respir Dis 1992: 145: 1506 ± 1512.
Dales R.E., Burnett R.T., Judek S., Cakmak S., and Brook J.R. The
influence of ambient fungal spores on emergency visits for asthma to a
regional childrens' hospital. Am J Respir Crit Care Med ( in press ) .
Delfino R.J., Coate B.D., Zeiger R.S., Seltzer J.M., Street D.H., and
Koutrakis P. Daily asthma severity in relation to personal ozone
exposure and outdoor fungal spores. Am J Respir Crit Care Med 1996:
154: 633 ± 641.
Delfino R.J., Murphy - Moulton A.M., and Becklake M.R. Emergency
room visits for respiratory illnesses among the elderly in Montreal:
association with low level ozone exposure. Environ Res 1998: 76:
67 ± 77.
Delfino R.J., Murphy - Moulton A.M., Burnett R.T., Brook J.R., and
Becklake M.R. Effects of ozone and particulate air pollution on
emergency room visits for respiratory illness in Montreal. Am J Respir
Crit Care Med 1997: 155: 568 ± 576.
Dersimonian R., and Laird N. Meta - analysis in clinical trials. Controlled
Clin Trials 1986: 7: 177 ± 188.
Devlin R.B., McDonnell W.F., Mann R., Becker S., House D.E.,
Schreinemachers D., and Koren H.S. Exposure of humans to ambient
levels of ozone for 6.6 hours causes cellular and biochemical changes
in the lung. Am J Respir Cell Mol Biol 1991: 4: 72 ± 81.
Dockery D.W., Cunningham J., Damokosh A.I., Neas L.M., Spengler J.D.,
Koutrakis P., Ware J.H., Raizenne M., and Speizer F.E. Health effects
of acid aerosols on North American children: respiratory symptoms.
Environ Health Perspect 1996a: 104: 500 ± 505.
Dockery D.W., Hoek G., Schwartz J., and Neas L.M. Specific Air
Pollutants and the Philadelphia mortality associations. In: Lee, J., and
Phalen, R., eds. Proceedings of the Second Colloquium on Particulate
Air Pollution and Human Health, May 1 ± 3, 1996, Park City, Utah.
University of Utah and University of California, Irvine, December
1996b.
Dockery D.W., Pope A., Xu X., Spengler J.D., Ware J.H., Fay M.E.,
Ferris B.G., and Speizer F.E. An association between air pollution
and mortality in six U.S. cities. N Engl J Med 1993: 329: 1753 ±
1759.
Goldsmith J.R., Griffith H.L., Detels R., Beeser S., and Neumann L.
Emergency room admissions, meteorologic variables, and air pollutants: a path analysis. Am J Epidemiol 1983: 118: 759 ± 778.
Hastie T., and Tibshirani R. Generalized Additive Models. Chapman and
Hall: London, UK, 1990.
Jones G.N., Sletten C., Mandry C., and Brantley P.J. Ozone level effect on
respiratory illness: an investigation of emergency department visits.
South Med J 1995: 88: 1049 ± 1056.
Jorgensen B., Lundbye - Christensen S., Song X.K., and Sun L. A
longitudinal study of emergency room visits and air pollution for
Prince George, British Columbia. Stat Med 1996: 823 ± 836.
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Air pollution, allergens and emergency department visits
Lippmann M., and Ito K. Separating the effects of temperature and season
on daily mortality from those of air pollution in London: 1965 ± 1972.
Inhalation Toxicol 1995: 7: 85 ± 97.
Lipsett M., Hurley S., and Ostro B. Air pollution and emergency room
visits for asthma in Santa Clara County, California. Environ Health
Perspect 1997: 105: 216 ± 222.
MathSoft, Inc., Data Analysis Products Division. S - Plus User's Guide.
MathSoft: Seattle, WA, 1997.
Morris R.D., Naumova E.N., and Munasinghe R.L. Ambient air pollution
and hospitalization for congestive heart failure among elderly people in
seven large US cities. Am J Public Health 1995: 85: 1361 ± 1365.
Neas L.M., Dockery D.W., Burge H., Koutrakis P., and Speizer F.E.
Fungus spores, air pollutants and other determinants of peak expiratory
flow rate in children. Am J Epidemiol 1996: 143: 797 ± 807.
Neas L.M., Dockery D.W., Koutrakis P., Tollerud D.J., and Speizer F.E.
The association of ambient air pollution with twice daily peak
expiratory flow rate measurements in children. Am J Epidemiol 1995:
141: 111 ± 122.
Poloniecki J.D., Atkinson R.W., Ponce - de - Leon A., and Anderson H.R.
Daily time series for cardiovascular hospital admissions and previous
day's air pollution in London, UK. Occup Environ Med 1997: 54:
535 ± 540.
Ponka A., and Virtanen M. Low - level air pollution and hospital
admissions for cardiac and cerebrovascular diseases in Helsinki. Am
J Public Health 1996: 86: 1273 ± 1280.
Raizenne M., Neas L.M., Damokosh A.I., Dockery D.W., Spengler J.D.,
Koutrakis P., Ware J.H., and Speizer F.E. Health effects of acid
aerosols on North American children: pulmonary function. Environ
Health Perspect 1996: 104: 506 ± 514.
Romieu I., Meneses F., Sienra - Monge J.J.L., Huerta J., Velasco S.R.,
White M.C., Etzel R.A., and Hernandez - Avila M. Effects of urban air
pollutants on emergency visits for childhood asthma in Mexico City.
Am J Epidemiol 1995: 141: 546 ± 553.
Rosas I., McCartney H.A., Payne R.W., Calderon C., Lacey J., Chapela R.,
and Ruiz - Velazco S. Analysis of the relationships between environmental factors ( aeroallergens, air pollution, and weather ) and asthma
emergency admissions to a hospital in Mexico City. Allergy 1998: 53:
394 ± 401.
Rossi O.V.J., Kinnula V.L., Tienari J., and Huhti E. Association of severe
asthma attacks with weather, pollen, and air pollutants. Thorax 1993:
4: 244 ± 248.
Salvaggio J., Seabury J., and Schoenhardt E.A. New Orleans asthma: V.
Relationship between charity hospital asthma admission rates,
semiquantitative pollen and fungal spore counts, and total particulate
aerometric sampling data. J Allergy Clin Immunol 1971: 48: 96 ± 114.
Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(5)
Stieb et al.
Schwartz J. Air pollution and hospital admissions for cardiovascular
disease in Tucson. Epidemiol 1997: 8: 371 ± 377.
Schwartz J., Dockery D.W., and Neas L.M. Is daily mortality associated
specifically with fine particles? J Air Waste Manage Assoc 1996: 46:
927 ± 939.
Schwartz J., and Morris R. Air pollution and hospital admissions for
cardiovascular disease in Detroit, Michigan. Am J Epidemiol 1995:
142: 23 ± 35.
Stieb D.M., Beveridge R.C., Rowe B.H., Walter S.D., and Judek S.
Assessing diagnostic classification in an emergency department:
implications for daily time - series studies of air pollution. Am J
Epidemiol 1998a: 148: 666 ± 670.
Stieb D.M., Beveridge R.C., Smith - Doiron M., Burnett R.T., Judek S.,
Dales R.E., and Anis A.H. Beyond administrative data: characterizing
cardiorespiratory disease episodes among patients visiting the
emergency department. Can J Public Health 2000: 91: 107 ± 112.
Stieb D.M., Brook J.R., Broder I., Judek S., Burnett R., and Beveridge
R.C. Personal exposure of adults with cardiorespiratory disease to
particulate acid and sulfate in Saint John, New Brunswick, Canada.
Appl Occup Environ Hyg 1998b: 13: 461 ± 468.
Stieb D.M., Burnett R.T., Beveridge R.C., and Brook J.R. Ozone and
asthma emergency department visits in Saint John, New Brunswick,
Canada. Environ Health Perspect 1996: 104: 1354 ± 1360.
Suh H.H., Spengler J.D., and Koutrakis P. Personal exposures to acid
aerosols and ammonia. Environ Sci Technol 1992: 26: 2507 ± 2516.
Sunyer J., Saez M., Murillo C., Castellsague J., Martinez F., and Anto
J.M. Air pollution and emergency room admissions for chronic
obstructive pulmonary disease: a 5 - year study. Am J Epidemiol 1993:
137: 701 ± 705.
Sunyer J., Spix C., Quenel P., Ponce - de - Leon A., Ponka A.,
Barumandzadeh T., Touloumi G., Bacharova L., Wojtyniak B., Vonk
J., Bisanti L., Schwartz J., and Katsouyanni K. Urban air pollution and
emergency admissions for asthma in four European cities: the APHEA
project. Thorax 1997: 52: 760 ± 765.
Tenias J.M., Ballester F., and Rivera M.L. Association between hospital
emergency visits for asthma and air pollution in Valencia, Spain.
Occup Environ Med 1998: 55: 541 ± 547.
Thurston G.D., Ito K., Hayes C.G., Bates D.V., and Lippmann M.
Respiratory hospital admissions and summertime haze air pollution in
Toronto, Ontario: consideration of the role of acid aerosols. Environ
Res 1994: 65: 271 ± 290.
Vincent R., Bjarnason S.G., Adamson I.Y.R., Hedgecock C., Kumarathasan P., Guenette M., Goegan P., and Bouthillier L. Acute pulmonary
toxicity of urban particulate matter and ozone. Am J Pathol 1997: 151:
1564 ± 1570.
477