Content uploaded by Candis Claiborn
Author content
All content in this area was uploaded by Candis Claiborn on Sep 15, 2014
Content may be subject to copyright.
147
Three multivariate receptor algorithms were applied to
seven years of chemical speciation data to apportion
fine particulate matter to various sources in Spokane,
Washington. Source marker compounds were used to
assess the associations between atmospheric concentra-
tion of these compounds and daily cardiac hospital
admissions and/or respiratory emergency department
visits. Total carbon and arsenic had high correlations
with two different vegetative burning sources and were
selected as vegetative burning markers, while zinc and
silicon were selected as markers for the motor vehicle
and airborne soil sources, respectively. The rate of res-
piratory emergency department visits increased 2% for
a 3.0 µg/m
3
interquartile range change in a vegetative
burning source marker (1.023, 95% CI 1.009–1.038) at
a lag of one day. The other source markers studied
were not associated with the health outcomes investi-
gated. Results suggest vegetative burning is associated
with acute respiratory events. Key words: air pollution;
generalized additive models; multivariate receptor
models; source apportionment; source health effects
analysis.
INT J OCCUP ENVIRON HEALTH 2006;12:147–153
A
irborne particulate matter (PM) mass has been
associated with adverse health outcomes.
1–3
However, chemical composition and average
particle size vary by source. Since one of the research
priorities of the National Research Council (NRC,
1998) is the identification of specific constituents
and/or sources of PM as contributors to adverse health
outcomes, we used a long time series of speciated PM
data to learn more about the association between PM
sources and health.
Previous time-series analyses have examined directly
the relationships between estimated source contribu-
tions to fine particle mass and specific health outcomes
in given communities.
1,4,5,6
These studies use time series
of estimated sources from factor-analysis–based algo-
rithms, such as positive matrix factorization (PMF) or
principal-component analysis, in time-series regression
analysis. Health-effect estimates from these studies are
potentially biased because they use estimated sources as
predictors in the regression models rather than using
measured variables with known error. Using these
imputed measures in a regression model results in
regression-attenuation bias.
7
The resulting bias is attrib-
utable both to measurement error in the species and to
rotational uncertainty in the receptor-model solutions.
Due to rotational uncertainty, the bias can result in
either an overestimate or an underestimate of the asso-
ciation; however, statistical methods have not advanced
sufficiently to eliminate this bias.
8
Alternatively, we can
identify measured individual species that are highly cor-
related with the estimated source contributions derived
from the multivariate receptor models. This approach
potentially allows a limited set of independent variables
with known laboratory analytical measurement errors to
be identified and incorporated in the health-effect
analysis. We apply this approach to a seven-year data set
of speciated PM from Spokane, Washington.
Spokane’s location in a semi-arid Eastern Washing-
ton valley makes it subject to frequent dust storms and
pollutant-trapping temperature inversions. This loca-
tion also has low concentrations of secondary aerosol
and potentially confounding gaseous air pollutants such
as SO
2
and O
3
. There is strong seasonal variability of the
primary particulate sources, and several potential par-
ticulate metals sources are present.
10
A previous PMF
analysis of a subset of the measurements in Spokane
identified several major sources of fine particles, includ-
ing vegetative burning, sulfate aerosol, motor vehicle,
nitrate aerosol, airborne soil, chlorine-rich source, and
metal processing.
10
The primary combustion sources
were the dominant contributors to PM
2.5
.
Of these combustion sources, previous studies have
shown that daily variations in fine PM derived from veg-
etative burning are associated with respiratory emer-
Ambient Woodsmoke and Associated
Respiratory Emergency Department Visits
in Spokane, Washington
ASTRID B. SCHREUDER, PHD, TIMOTHY V. LARSON, PHD, LIANNE SHEPPARD, PHD,
CANDIS S. CLAIBORN, PHD
Received from the Departments of Environmental and Occupa-
tional Health Sciences (ABS), Civil and Environmental Engineering
(TVL), and Biostatistics and Environmental and Occupational
Health Sciences (LS), University of Washington, Seattle, Washing-
ton; and the Department of Civil and Environmental Engineering,
Washington State University, Pullman, Washington (CSC). Sup-
ported by a grant from the Mickey Leland National Urban Air Toxic
Research Center and in part by the United States Environmental
Protection Agency through agreement R-82735501-1.
Address correspondence and reprint requests to: Timothy V.
Larson, Department of Civil and Environmental Engineering, Box
352700, University of Washington, Seattle, WA 98195-4803, U.S.A.;
telephone: (206) 543-6815; fax: (206) 685-3836; e-mail: <tlarson@
u.washington.edu>.
gencies,
11,12
whereas variations in fine PM from mobile
source emissions are associated with both cardiac
6,9,13–16
and respiratory
6,9,13,17
outcomes. Previous studies in
Spokane have shown that emergency department (ED)
visits for asthma are associated with combustion-
derived PM, including fine particulate Zn, but not with
fine particle soil.
18
In addition, no significant associa-
tions were observed in Spokane between acute mortal-
ity and the coarse fraction of PM,
19,20
although studies
in other locations have observed associations between
acute adverse health outcomes and this PM fraction.
4,21
METHODS
Our data consist of daily measures taken over a seven-
year period from 1 September 1995 to 15 May 2002 in
Spokane, Washington.
Chemical Speciation Data
Daily fine PM filter samples were collected at a residen-
tial monitoring site between 1995 and 2002. Figure 1
shows this site’s (Rockwood) location 8 km north of the
central business district. These filters were analyzed for
nitrate, sulfate, ammonium, elemental (EC) and
organic carbon (OC), and elemental analyses that
included the following trace elements: Al, As, Br, Ca,
Cu, Fe, K, Mn, Na, Pb, S, Si, Ti, and Zn. Details of the
mass estimation and uncertainty methods are given else-
where.
22,23
During the first four years of this study, we
used the thermal manganese oxidation method (TMO)
to quantify particulate OC and EC. An overestimate of
the ratio of EC to OC was expected due to this method’s
failure to correct for pyrolytically formed char.
10,24
Therefore, only total carbon (TC) was available for the
entire study period. TC, the sum of OC and EC, was
obtained from TMO between 1 September 1995 and 20
September 1999, and from thermal optical transmission
(TOT) between 24 September 1999 and 15 May 2002.
Health Data
Hospital admissions for Spokane were aggregated from
the Comprehensive Hospital Admissions Reporting
System (CHARS) as daily counts of cardiac hospital
admissions (International Classification of Disease,
Ninth Revision World Health Organization, Geneva
(ICD9) codes 390-459). Emergency department visit
records were obtained directly from four Spokane-area
hospitals (Deaconess, Valley, Sacred Heart, and Holy
Family). Daily respiratory ED visits (all respiratory
causes, ICD9 codes 460-519) were aggregated. Any
events from subjects who did not have Spokane-area zip
codes as their residences were removed. Readmissions
within two weeks of the first admission were also
excluded.
Meteorologic Data
The Spokane County Air Pollution Control Authority
(SCAPCA) supplied temperature and relative humidity
data aggregated into daily averages.
Source Apportionment
We used 13 species (K, Fe, S, Zn, Ca, Si, Cu, Mn, Br, Na,
As, NO
3
–
, TC) with good data in each of 2,925 daily
samples for the source-apportionment analyses. These
species were chosen because: they were available for
use by all three multivariate receptor-model algo-
rithms; they were consistently above their respective
detection limits; and they did not exhibit any systematic
changes in concentration due to a corresponding
change in analytical method that might have occurred
during the study period. See Table 1 for a summary of
the fine-particle mass and species concentrations. We
had a maximum of 38% of days with data below the
limit of detection (LOD) and 29% missing (Br, Na, As;
see Table 1). These daily values were used in the source
apportionment analysis described below.
The source contributions were estimated using
three different multivariate receptor algorithms:
Unmix,
25,26
positive matrix factorization (PMF),
27
and
table-driven least squares as implemented in the multi-
linear engine version 2 (ME2).
28
The latter two least-
squares algorithms weight the observed concentrations
by their individual uncertainty estimates that vary for
each species, for each observation. Less weight is given
to more uncertain measurements, including those that
are reported as either below the LOD and/or missing.
Following Polissar et al., we replaced values below the
LOD by half of the LOD and set the corresponding
uncertainty to 5/6 of the LOD values.
29
Missing values
were replaced by the geometric mean of all the meas-
ured values and the corresponding uncertainty was set
to four times the geometric mean value. For the PMF
analysis, we used the same criteria and approach
148 • Schreuder et al. www.ijoeh.com • INT J OCCUP ENVIRON HEALTH
Figure 1—Map of the Rockwood sampling site located
in a residential area of Spokane, Washington.
VOL 12/NO 2, APR/JUN 2006 • www.ijoeh.com Woodsmoke and Health • 149
detailed by Kim et al.
10
Robust, individual markers of a
particular source’s total mass contributions were iden-
tified using the correlations between the source contri-
bution estimates from all three algorithms and the con-
centrations of individual species.
Health-effects Analysis
All statistical analyses were conducted using S-Plus 6.1
(Insightful Inc., Seattle, WA). We regressed daily car-
diac hospital admissions and respiratory ED visits
against the measured concentrations of selected source
markers using Poisson regression with a generalized
additive model (GAM)
30
and an exact GAM standard-
error estimate.
31
We controlled for season, tempera-
ture, relative humidity, and day of the week. The gen-
eral form for the regression of event counts Y
t
on daily
levels of exposures X
t
was: E(Y
t
) = exp{
0
+
1
X
t–l
+
S(time
t
,
1
) + S(temp
t
,
2
) + *DOW + rhum
t
}, where
temp is the daily average temperature; DOW are indi-
cator variables for day of week, rhum is daily average
relative humidity and l is the lag of the exposure. Our
hypotheses used l = 1 day lag in the models. The
smooth function, S, is constructed using smoothing
splines with a parameter, . Six degrees of freedom per
year were used in this investigation (
1
= 6 * 7 = 42 df).
Temperature was smoothed with
2
= 2 df. A sensitivity
analysis of potentially influential points was conducted
by repeating the analyses with these points removed,
which did not change the results.
RESULTS
Source Apportionment and Source Marker Compounds
Eight sources were identified using three different
multivariate receptor algorithms (PMF, ME2, Unmix).
Their estimated average source contributions to fine
PM mass are shown in Table 2. Both the source contri-
butions and the source compositional features are con-
sistent with the previous source-apportionment analy-
sis at this site.
10
The three algorithms (PMF, ME2, and
Unmix) could not account for averages of 2.0, 1.4, and
0.3 µg/m
3
, respectively, of the overall mean PM
2.5
mass
concentration at this site. The magnitudes of the top-
ranking source contribution, vegetative burning, are
similar across the three receptor-modeling methods.
Due to the larger number of samples in this analysis,
we were able to discern an additional As-rich source
that was not evident in the earlier apportionment.
10
TABLE 1 Summary of fine Particle Mass (ng/m
3
) and 13 Species Concentrations Used for Unmix/PMF/ME2
Arithmetic Geometric BDL† Missing
Species Mean Mean* Minimum Maximum No. (%) No. (%)
PM
2.5
10,580 8,790 930 43,230 — 698 (23.9)
K78612 698 127 (4.3) 98 (3.4)
Fe 125 89 2 1,970 31 (1.1) 98 (3.4)
S 288 244 2 1,270 125 (4.3) 98 (3.4)
Zn 12 8 1 280 269 (9.2) 98 (3.4)
Ca 49 31 3 957 686 (23.5) 98 (3.4)
Si 312 189 6 8,616 752 (25.7) 98 (3.4)
Cu 15 7 0.6 501 833 (28.5) 99 (3.4)
Mn 3.5 2.6 0.5 69 1,146 (38.2) 98 (3.4)
Br 1.1 0.9 0.01 13 0 (0) 848 (29.0)
Na 63 49 0.01 6,995 4 (0.1) 848 (29.0)
As 0.4 0.3 0.05 9.3 0 (0) 848 (29.0)
NO
3
–
579 455 20 5,923 450 (15.4) 224 (7.7)
TC 4,540 3,900 130 23,600 — 220 (7.5)
*Data below the limit of detection were replaced by half of reported detection limit values for the geometric mean calculations.
†Below detection limit.
TABLE 2 Estimated Average Source Contributions to Fine-particle Mass Concentrations in µg/m
3
PMF ME2 Unmix PMF†
Vegetative burning 3.6 (2.3)* 4.1 (2.6) 4.3 (3.8) 5.3 (4.3)
As-rich 1.2 (1.3) 0.9 (1.0) 0.5 (0.7) —
Motor vehicle 0.9 (0.9) 1.0 (1.0) 0.6 (0.8) 1.3 (0.9)
Sulfate aerosol 1.1 (1.0) 1.5 (1.2) 2.3 (1.7) 2.3 (1.2)
Nitrate aerosol 0.7 (0.8) 0.6 (0.7) 1.4 (1.3) 1.1 (1.5)
Airborne soil 1.0 (1.2) 1.0 (1.2) 0.6 (0.8) 1.0 (1.2)
Cu-rich 0.08 (0.15) 0.06 (0.12) 0.3 (0.6) 0.3 (0.3)
Marine 0.06 (0.12) 0.07 (0.14) 0.3 (0.3) 0.7 (0.9)
*( ) = standard deviation of daily estimates.
†Previous analysis by Kim et al. (2003) of a subset of 945 samples taken between January 1995 and December 1997.
150 • Schreuder et al. www.ijoeh.com • INT J OCCUP ENVIRON HEALTH
The As-rich source profile is similar to our derived veg-
etative-burning profile (with the exception of more
As) and therefore may be due to the burning of As-
treated wood.
32
Burning treated wood was prohibited
by the Spokane County Air Pollution Agency in late
1994, which is consistent with the fact that the contri-
bution from this study’s As-rich source is relatively
small compared with the more general vegetative-
burning source, and that it also declined over the time
period of this study. Table 3 summarizes the average
percentage of each species mass concentration associ-
ated with each source. As expected, the nitrate, sulfate,
As-rich, Cu-rich, and airborne soil sources contributed
most of the NO
3
–
, S, As, Cu, and Si, respectively (our
“Cu-rich” feature was previously named “metals”
10
). In
addition, the motor vehicle source contributed most
of the Zn and the vegetative-burning source con-
tributed most of the TC.
Table 4 shows correlations for potential source-
marker compounds with the source-contribution esti-
mates. To facilitate interpretation, only correlations > 0.3
are shown. From Table 4, TC, As, and Zn are most
highly correlated with the vegetative-burning source,
As-rich source, and motor-vehicle source, respectively.
Finally, as expected, Si is highly correlated with the air-
borne-soil source. Based on these results, we treat Si,
TC, and Zn as markers of the fine particle mass con-
tributed by three different source-related features: air-
borne soil, vegetative burning, and motor-vehicle
exhaust, respectively. For completeness, we also exam-
ined the association between As and respiratory ED
visits, hypothesizing that As is a tracer for an additional,
smaller source of wood combustion.
Health Effects Analysis
Table 5 gives summary statistics for the variables used in
the analyses. This data set is a slightly smaller subset of
that used in the source-apportionment analysis and
therefore the species means differ somewhat in Tables 1
and 5. Table 6 gives overall estimated relative risks
(RRs) for an interquartile (IQR: 75th–25th percentile)
range increase in the identified source-marker com-
pounds. There was no association of Si, As, or Zn with
either of the health outcomes. In contrast, we estimated
a 2% increase in respiratory ED visits for an interquar-
TABLE 3 Estimated Percentages of Species Contributed by Each Source Feature, Unmix/PMF/ME2
Sources/ Vegetative Motor Sulfate Nitrate Airborne
Species Burning As-rich Vehicle Aerosol Aerosol Soil Cu-rich Marine
K 37/13/19 1/14/—* 6/19/22 11/5/9 3/2/2 37/46/46 — 4/—/—
Fe 15/4/4 — 7/12/12 3/—/— — 68/80/80 — 7/3/4
S 14/27/31 —/1/— 2/—/— 65/34/38 14/19/17 1/18/15 2/—/— 2/—/—
Zn 7/3/2 2/3/— 74/90/90 4/—/4 5/3/3 5/—/— 2/1/1 —
Si 3/—/— — 3/3/3 — 1/—/— 91/97/97 — 3/—/—
Cu —/—/2 — ————99/96/97 —
Mn —3/— —/6/— 12/18/16 20/—/— —/3/2 68/74/75 — —/2/2
Na 8/—/— — 4/—/— 6/—/— — 14/5/— — 67/95/100
As 1/—/— 97/97/88 — —/—/7 ————
NO
3
–
————99/99/99 — — —
TC 55/96/85 5/—/11 7/2/4 12/—/— 14/—/— — 4/—/— 3/1/—
*— indicates <1%.
TABLE 4 Pearson Correlation Coefficients for Source Features vs Species, Unmix/PMF/ME2
Sources/
Possible Vegetative Motor Sulfate Nitrate Airborne
Markers Burning As-rich Vehicle Aerosol Aerosol Soil Cu-rich Marine
K 0.57/0.42/0.43 —/0.37/— —/0.44/0.45 —/—/0.36 — 0.71/0.73/0.73 — —
Fe — — — — — 0.94/0.94/0.95 — —
S —/0.35/0.36 —/0.35/— — 0.87/0.50/0.52 —/0.46/0.47 — — —
Zn 0.35/0.46/0.44 — 0.92/0.94/0.93 — 0.32/—/— — — —
Ca — — — — — 0.85/0.93/0.94 — —
Si — — — — — 0.97/0.92/0.93 — —
Cu — — — — — — 0.99/0.98/0.99 —
Mn — — — — — 0.91/0.80/0.81 — —
Br 0.50/0.45/0.42 0.31/0.41/0.35 —/0.38/0.37 —/0.91/0.94 — — — —
Na — — — — — — — —/0.73/0.73
As 0.35/0.52/0.38 0.96/0.95/0.85 —/0.33/0.38 —/0.31/0.40 — — — —
NO
3
–
—/0.34/0.34 — —/0.30/0.30 — 0.99/0.94/0.95 — — —
TC 0.86/0.92/0.91 —/0.50/0.35 0.41/0.54/0.56 — 0.41/0.33/0.33 — — —
*— indicates a correlation <0.30.
VOL 12/NO 2, APR/JUN 2006 • www.ijoeh.com Woodsmoke and Health • 151
tile range change in TC (vegetative-burning marker)
(IQR 3.0 µg/m
3
, 1.023, 95% CI 1.009–1.038) lagged
one day. Table 7 shows the RRs lagged one day by heat-
ing (October–February) versus non-heating
(March–September) season. The non-heating season
diluted the association, while the relationship was larger
(1.051, 95% CI 1.010–1.094) in the heating season.
DISCUSSION
Our source apportionment results are consistent with
those of a previous source apportionment in Spokane.
10
They used PMF to deduce the sources of PM
2.5
at the
same residential site in Spokane for 1995–1997 using a
subset of our final dataset. Using the same species with
the exception of S rather than SO
4
=
and Cl rather than
Na, they found seven of the eight source features
reported here. We found not only source-contribution
estimates (see Table 2), but also source compositional
features similar to those reported by Kim et al.
10
Our
substantially larger data set over the entire study period
(2,925 vs 945 samples) allowed us to also resolve an
additional, albeit minor, As-rich feature that they could
not discern. It is noteworthy that in this airshed total
particulate carbon and zinc were excellent individual
tracers of motor-vehicle source and vegetative-burning
contributions, respectively. The motor-vehicle source is
enriched in Zn, presumably because of the use of the
anti-wear additive zinc dialkyldithiophosplate.
33
The
total carbon association is due to the fact that vegeta-
tive-burning particles are enriched in total carbon rela-
tive to other species and that this vegetative burning
source is a dominant contributor to PM
2.5
in this air-
shed, especially in the winter heating season.
We found evidence for adverse health outcomes asso-
ciated with TC, our source marker for general vegeta-
tive-burning. The vegetative burning feature derived
from the receptor model predicts that wood-smoke par-
TABLE 5 Daily Summary Statistics for Model Variables
Percentile
______________________________________________________________
Outcome Mean SD 5th 25th 50th 75th 95th
Daily cardiac hospital
admissions 5.3 1.9 24579
Daily respiratory ED visits 16.3 6.9 7 11 15 20 29
TC 4.6 µg/m
3
2.6 µg/m
3
1.4 µg/m
3
2.8 µg/m
3
4.0 µg/m
3
5.8 µg/m
3
9.4 µg/m
3
As 0.45 ng/m
3
0.47 ng/m
3
0.08 ng/m
3
0.18 ng/m
3
0.31 ng/m
3
0.57 ng/m
3
1.1 ng/m
3
Zn 12.0 ng/m
3
11.5 ng/m
3
2 ng/m
3
4 ng/m
3
8 ng/m
3
16 ng/m
3
34 ng/m
3
Si 310 ng/m
3
410 ng/m
3
4 ng/m
3
85 ng/m
3
190 ng/m
3
410 ng/m
3
930 ng/m
3
Relative humidity 69% 19% 38% 53% 70% 86% 95%
Temperature 47.4° F 15.5° F 25.3° F 35.4° F 45.4° F 60.0° F 73.4° F
PM
2.5
10.6 µg/m
3
6.8 µg/m
3
2.9 µg/m
3
5.8 µg/m
3
8.9 µg/m
3
13.5 µg/m
3
25.1 µg/m
3
TABLE 6 Relative Risks (RRs) and 95% CIs for the Entire Study Period of Hypothesized Health Outcomes for a Given
IQR Increase
Cardiac Hospital All Respiratory ED
IQR Increase RR (95% CI) RR (95% CI)
TC (vegetative burning) 3.0 µg/m
3
lag 0 1.008 (0.983–1.032) 1.012 (0.997–1.027)
lag 1 0.999 (0.975–1.024) 1.023 (1.009–1.088)
*
As (As-rich) 0.39 ng/m
3
lag 0 0.994 (0.976–1.012) 1.008 (0.998–1.019)
lag 1 0.997 (0.980–1.015) 1.002 (0.992–1.013)
Zn (motor vehicle) 11.6 ng/m
3
lag 0 1.007 (0.987,1.028) 1.012 (0.999–1.024)
lag 1 1.003 (0.983,1.024) 1.006 (0.994–1.019)
Si (airborne soil) 324 ng/m
3
lag 0 1.011 (0.994–1.028) 0.999 (0.990–1.009)
lag 1 1.005 (0.989–1.021) 0.999 (0.989–1.009)
PM
2.5
7.7 µg/m3
lag 0 1.008 (0.985–1.032) 1.011 (0.997–1.025)
lag 1 1.000 (0.978–1.023) 1.013 (0.999–1.027)
*
p = 0.002.
152 • Schreuder et al. www.ijoeh.com • INT J OCCUP ENVIRON HEALTH
ticles (not total PM
2.5
) contain about 80% TC by mass.
Therefore, this epidemiologic association translates to a
2.3/3.0 * 0.8 = 0.6% increase in respiratory ED visits per
µg/m
3
of woodsmoke particle mass. This is reasonably
consistent with the results of McGowan et al.,
11
who
found a 0.2% (3.37/14.8) increase in all respiratory
admissions per µg/m
3
increase in PM
10
in Christchurch,
New Zealand, a region that was significantly impacted
by woodsmoke during the heating season. This result is
also consistent with the reported associations between
outdoor fine PM and measures of acute respiratory
response in woodsmoke-impacted neighborhoods,
34–37
and the fact that ambient wood-smoke particles readily
penetrate indoors.
38
In this study, the other combustion-source marker
compounds, As and Zn, did not show associations with
the health outcomes studied. In contrast, previous
health analyses in Spokane found an increased risk of
ED visits for asthma with higher meteorologic stagna-
tion index and an increased risk of all respiratory and
asthma-only ED visits for increases in both fine particle
Zn and CO, the latter being a general indicator of stag-
nant air conditions and increased levels of surface com-
bustion sources.
23,18,39
In addition, Si, a marker of fine-
particle soil, also did not show associations with any of
the health outcomes, consistent with previous observa-
tions in Spokane.
18
However the inability to detect asso-
ciation with the other source markers is not surprising,
given their relatively low PM
2.5
source contributions
compared with vegetative burning at this location, as
summarized in Table 2.
Our results suggest vegetative burning is associated
with acute respiratory events. This conclusion is quali-
fied by the relatively small community population and
the relatively low pollution levels in Spokane. While we
compensated for these features by using a seven-year
time series, the low pollution level and our need to use
TC rather than the more traditional temperature-
resolved carbon fractions results in less-than-optimal
resolving power in the source-apportionment analysis.
Another important limitation is potential misclassifica-
tion of exposure resulting from: 1) the use ambient air-
pollution measurements at a single site as a proxy for
ambient-source personal exposure, and 2) increased
potential for indoor exposures to woodsmoke from
increased use of wood stoves during cold, stable condi-
tions that are associated with increased outdoor wood-
smoke levels. These limitations are offset by some
important additional strengths: we had a large number
of chemically resolved filter samples over the seven-year
time period, the study population lives within 8 km of
the monitor, and in Spokane vegetative burning is an
important source of PM
2.5
during the heating season.
References
1. Laden F, Neas LM, Dockery DW, Schwartz J. Association of fine
particulate matter from different sources with daily mortality in
six U.S. cities. Environ Health Perspect. 2000;108:941-7.
2. Pope CA, Thun MJ, Namboodiri MM, et al. Particulate air pol-
lution as a predictor of mortality in a prospective study of U.S.
adults. Am J Respir Crit Care Med. 1995; 151:669-74.
3. Schwartz J. Air pollution and daily mortality: a review and meta
analysis. Environ Res. 1994;64:36-52.
4. Mar TF, Norris GA, Larson TV, Wilson WE, Koenig JQ. A
reanalysis of the association between air pollution and cardio-
vascular mortality in Phoenix, 1995–1997. Environ Health Per-
spect. 2000;108:347-53.
5. Tsai FC, Apte MG, Daisey JM. An exploratory analysis of the
relationship between mortality and the chemical composition
of airborne particulate matter. Inhalation Toxicol. 2000; 12
suppl 2: 121-35.
TABLE 7 Relative Risks (RRs) and 95% CIs by Season of Hypothesized Health Outcomes Lagged One Day
Cardiac Hospital All Respiratory ED
IQR Increase RR (95% CI) RR (95% CI)
TC (vegetative burning)
Heating season
*
3.5 µg/m
3
1.024 (0.968–1.084) 1.051 (1.010–1.094)
**
Non-heating 2.2 µg/m
3
0.992 (0.966–1.018) 1.013 (0.998–1.028)
As (As-rich)
Heating 0.53 ng/m
3
0.992 (0.958–1.026) 1.013 (0.992–1.036)
Non-heating 0.30 ng/m
3
1.000 (0.981–1.021) 0.998 (0.986–1.011)
Zn (Motor vehicle)
Heating 14.9 ng/m
3
1.002 (0.969–1.037) 1.014 (0.994–1.036)
Non-heating 7.9 ng/m
3
1.004 (0.980–1.029) 1.000 (0.986–1.014)
Si (Airborne soil)
Heating 162 ng/m
3
1.002 (0.986–1.019) 1.001 (0.992–1.010)
Non-Heating 376 ng/m
3
1.024 (0.973–1.078) 1.002 (0.972–1.033)
PM
2.5
Heating 10.1 µg/m
3
1.015 (0.968–1.063) 1.018 (0.985–1.052)
Non-heating 5.5 µg/m
3
0.995 (0.969–1.021) 1.009 (0.994–1.025)
*
October–February.
†
p = 0.01.
VOL 12/NO 2, APR/JUN 2006 • www.ijoeh.com Woodsmoke and Health • 153
6. Thurston GD, Ito K, Mar T, et al. The Workshop on Source
Apportionment of PM Health Effects: inter-comparison of
results and implications. Environ Health Perspect. 2005;
113:1768-74.
7. Zeger S, Thomas D, Dominici, F, et al. Exposure measurement
error in time-series studies of air pollution: concepts and con-
sequences. Environ Health Perspect. 2000;108:419–26.
8. Lumley T, Liu H. How Can Source Apportionment and Recep-
tor Modelling Data be Used in Epidemiology? Presented at the
2003 AAAR PM Meeting, Particulate Matter: Atmospheric Sci-
ences, Exposure and the Fourth Colloquium on PM and
Human Health, April 2003, Pittsburgh, PA. <http://www.
abstracts2view.com/aaarpm/sessionindex.php)>.
9. Janssen NAH, Schwartz J, Zanobetti A, Suh HH. Air condition-
ing and source-specific particles as modifiers of the effect of
PM
10
on hospital admissions for heart and lung disease. Environ
Health Perspect. 2002;110:43-9.
10. Kim E, Larson TV, Hopke PK, Slaugher C, Sheppard L, Clai-
born C. Source identification of PM
2.5
in an arid northwest U.S.
city by positive matrix factorization. Atmospheric Res. 2003;66:
291-305.
11. McGowan JA, Hider PN, Chacko E, Town GI. Particulate air pol-
lution and hospital admissions in Christchurch, New Zealand.
Aust NZ J Public Health. 2002;26:23-9.
12. Peel JL, Tolbert PE, Klein M, et al. Ambient air pollution and
respiratory emergency department visits. Epidemiology. 2005;
16:164-74.
13. Zanobetti A, Schwartz J. Airborne particles and hospital admis-
sions for heart and lung disease. In: Revised Analyses of Time-
series Studies of Air Pollution and Health. Special Report,
Boston, MA: Health Effects Institute, 2003: 241-8. <http://www.
healtheffects.org/Pubs/TimeSeries.pdf[18>. October 2004
14. Schwartz JA, Litonjua AA, et al. Traffic related pollution and
heart rate variability in a panel of elderly subjects. Thorax. 2005;
60:455-61.
15. Gold DR, Litonjua AA, et al. Air pollution and ST-segment
depression in elderly subjects. Environ Health Perspect. 2005;
113:883-7
16. Peters A, von Klot S, Heier M, et al. Exposure to traffic and the
onset of myocardial infarction N Engl J Med. 2004;351:1721-30.
17. Neuberger M, Schimek MG, Horak F, et al. Acute effects of par-
ticulate matter on respiratory diseases, symptoms and functions:
epidemiological results of the Austrian Project on Health
Effects of Particulate Matter (AUPHEP). Atmospheric Environ-
ment. 2004;38:3971-81.
18. Norris G, Larson TV, Koenig JQ, Claiborn C, Sheppard L, Finn
D. Asthma aggravation, combustion, and stagnant air. Thorax.
2000;55:466-70.
19. Schwartz, J, Norris G, Larson T, Sheppard L, Claiborn C, Koenig
J. Episodes of high coarse particle concentrations are not asso-
ciated with increased mortality. Environ Health Perspect. 1999;
107:339-42.
20. Slaughter JC, Kim E, Sheppard L, Sullivan JH, Larson TV, Clai-
born C. Association between particulate matter and emergency
room visits, hospital admissions and mortality in Spokane, Wash-
ington. J Exposure Analysis Environ Epidemiol. 2005;15: 153-9.
21. Brunekreef B, Forsberg B Epidemiological evidence of effects
of coarse airborne particles on health. Eur Respir J. 2005;26:
309-18.
22. Claiborn CS, Larson TV, Sheppard L. Testing the metals
hypothesis in Spokane, Washington. Environ Health Perspect.
2002;110 suppl 4: 547-52.
23. Claiborn CS, et al. Final Investigators’ Report: Testing the
Metals Hypothesis in Spokane: Daily Associations Between Res-
piratory and Cardiovascular Outcomes and Airborne Fine Par-
ticulate Metal Species. Prepared for The Mickey Leland
National Urban Air Toxics Research Center, March 2005.
24. Chow JC, Watson JG Pritchett LC, Pierson WR, Frazier CA, Pur-
cell RG. Measurement methods to determine compliance with
ambient air quality standard for suspended particles. J Air Waste
Management Assoc. 1995;45:320-82.
25. Henry RC, Park ES, Spiegelman CH. Comparing a new algo-
rithm with the classic methods for estimating the number of fac-
tors. Chemometrics and Intelligent Laboratory Systems. 1999;
48:91-7.
26. Henry RC. Multivariate receptor modeling by n-dimensional
edge detection. Chemometrics and Intelligent Laboratory Sys-
tems. 2003;65:179-89.
27. Paatero P. Least square formulation of robust non-negative
factor analysis. Chemometrics and Intelligent Laboratory Sys-
tems. 1997;37:23-35.
28. Paatero P. The multilinear engine—a table-driven least squares
program for solving multilinear problems, including the n-way
parallel factor analysis model. J Computational and Graphical
Statistics. 1999;8:1-35.
29. Polissar AV, Hopke PK. Atmospheric aerosol over Alaska 2. Ele-
mental composition and sources. J Geophysical Res. August 20,
1998; Vol. 103 No. D15: 19045-57.
30. Hastie TJ, Tisshirani RJ. Generalized Additive Models. New
York: Chapman and Hall, 1990.
31. Dominici F, McDermott A, Hastie T. Improved semiparametric
time series models of air pollution and mortality. J Am Statisti-
cal Assoc. 2004;99:938-48.
32. Maykut NN, Lewtas J, Kim E, Larson TV. Source apportionment
of PM
2.5
at an urban IMPROVE site in Seattle, WA. Environ Sci
Technol. 2003;37:5135-42.
33. Mosey NJ, Muser MH, Woo TK. Molecular mechanisms for the
functionality of lubricant additives. Science. 2005;307:1612-5.
34. Jansen K, Larson TV, Koenig JQ, et al. Associations between
health effects and PM and black carbon in subjects with respi-
ratory disease. Environ Health Perspect. 2005,113:1741-6.
35. Koenig,JQ, Larson TV, Hanley QS, et al. Pulmonary function
changes in children associated with fine particulate matter. Env-
iron Res. 1993;63:26-38.
36. Larson T, Koenig J. Wood smoke: emissions and non-cancer res-
piratory effects. Annu Rev Public Health. 1994;15:133-56
37. Lipsett M, Hurley S, Ostro B. Air pollution and emergency
room visits for asthma in Santa Clara County, California. Envi-
ron Health Perspect. 1997;105:216-22.
38. Larson T, Gould T, Simpson C, et al. Source apportionment of
indoor, outdoor and personal PM
2.5
in Seattle, WA, using posi-
tive matrix factorization. J Air and Waste Management Assoc.
2004;54:1175-87.
39. Slaughter JC, Kim E, Sheppard L, Sullivan JH, Larson TV, Clai-
born C. Association between particulate matter and emergency
room visits, hospital admissions and mortality in Spokane,
Washington. J Exposure Analysis Environ Epidemiol. 2005;15:
153-9.
