United States
Environmental Protection
Agency
Environmental Monitoring Systems
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S4-84-014 Mar. 1984
SEPA Project Summary
A Study of Personal Exposure to
Carbon Monoxide in Denver,
Colorado
Ted Johnson
Under EPA Contract 68-02-3755,
PEDCo Environmental conducted a
study of personal exposure to carbon
monoxide (CO) in Denver, Colorado.
The target population for the study
included all noninstitutionalized, non-
smoking residents of the urbanized
portion of the metropolitan area who
were between 18 and 70 years of age at
the time of the study. A total of 454
study participants were obtained through
the use of a screening questionnaire
administered to several thousand house-
holds in the study area. Each participant
was asked to carry a personal exposure
monitor (PEM) and an activity diary for
two consecutive 24-hour sampling
periods and to provide a breath sample
at the end of each sampling period.
Each participant also completed a
detailed background questionnaire.
Analyses of approximately 900 person-
days of PEM and activity diary data
found that personal CO exposures were
higher in microenvironments associated
with motor vehicles such as parking
garages and automobiles. Mean indoor
residential exposure was increased
2.59 ppm by gas stove operation, 1.59
ppm by smokers, and 0.41 ppm by
attached garages. The weighted means
for daily maximum 1-hour and 8-hour
exposures during the study period were
10.3 ppm and 4.9 ppm, respectively.
Approximately 3 percent of the daily
maximum 1-hour exposures exceeded
35 ppm; approximately 11 percent of
the daily maximum 8-hour exposures
exceeded 9 ppm. Only one of the 15
fixed-site monitors operating during the
study reported daily maximum 1-hour
values exceeding 35 ppm. Eleven fixed-
site monitors reported daily maximum
8-hour values exceeding 9 ppm. Corre-
lations between CO values recorded
simultaneously by PEM's and by fixed-
site monitors were generally higher for
outdoor personal monitoring locations
than for indoor locations; however,
correlations were weak for most locations.
Diurnal patterns for weekdays, Saturday,
and Sundays were developed for hourly
average exposures and composite
fixed-site values.
This Project Summary was developed
by EPA's Environmental Monitoring
Systems Laboratory, Research Triangle
Park, NC, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering
information at back).
Introduction
The National Ambient Air Quality
Standard (NAAQS) for carbon monoxide
(CO) states that the daily maximum 1-
hour concentration shall not exceed 35
ppm more than onceperyearandthatthe
daily maximum 8-hour concentration
shall not exceed 9 ppm more than once
per year. Compliance with these standards
is usually determined by fixed-site
monitoring data. However, fixed-site
monitoring data may not provide an
accurate indication of personal exposure
within an urban population, which is a
function of both geographic location (i.e.,
downtown versus suburbia) and immedi-
ate physical surroundings (i.e., indoors
versus outdoors). Better estimates of
personal exposure can be developed by
equipping a large number of subjects
-------
with portable monitors and activity
diaries. If the subjects are properly
selected, their exposures can be extrapo-
lated to the larger urban population.
Such a study was conducted in Denver,
Colorado, by PEDCo Environmental, Inc.
for the Environmental Monitoring Systems
Laboratory (EMSL) of the U.S. Environ-
mental Protection Agency (EPA). Each of
454 subjects was asked to carry a PEMand
activity diary for two consecutive 24-hour
sampling periods and to provide a breath
sample at the end of each sampling
period. Each participant also completed a
detailed background questionnaire. The
questionnaire results and approximately
900 subject-days of PEM and activity
diary data were analyzed to determine if
factors such as microenvironment and
the presence of indoor CO sources
significantly affect personal CO exposure.
In addition, the exposure of the entire
Denver population was extrapolated from
exposures recorded by the study partici-
pants. PEDCo also compared CO levels
recorded by fixed-site monitors to levels
recorded simultaneously by the PEM's.
Data Collection Instruments
and Procedures
The target population of the study
included all noninstitutionalized, nonsmok-
ing residents of the urbanized portion of the
Denver, Colorado, metropolitan area who
were between 18 and 70 years of age at
the time of the study. Research Triangle
Institute (RTI) and PEDCo developed a
two-phase scheme for sampling this
population, which is estimated to be
245,000. In the first phase, a two-stage
sample of housing units was selected.
Data on the individuals residing within
these housing units were collected using
a brief screening questionnaire ad-
ministered by telephone or in the field.
Individuals who exhibited rare character-
istics with respect to CO exposure were
identified and over-sampled in the
second-phase of sample selection.
Individuals entered the sample by three
paths. The majority of study participants
(402) were identified by means of a
telephone screening questionnaire ad-
ministered to members of housing units
appearing on a list prepared by Donnelley
Marketing Information Services. The
remaining 52 study participants were
identified by field screening of housing
units which 1) appeared on the Donnelley
list but for which no telephone number
was available or 2) were identified
through a special survey of housing units
which did not appear on the Donnelley
list. The original sample selection pro-
tocol was designed to yield 500 study
participants. The reduced sample size
(454) resulted from a higher than
expected refusal rate and unexpected
equipment problems early in the study.
The data collection instruments used in
the Denver CO study included three
questionnaires [screening, computer
model input questionnaire (CMIQ), and
study] providing background data on
subjects and their families, a network of
15 fixed-site monitors, the PEM's and
activity diaries carried by each subject,
and breath sample bags. The screening
questionnaire was administered on a
household basis as a means of identifying
persons eligible for study. It requested the
name of each household member, rela-
tionship to head of household, sex, age,
smoking status, occupation, and typical
commute time. The completed screening
questionnaires yielded a list of 2232
eligible individuals from which were
selected a stratified sample of 1139
potential subjects. An attempt was made
to administer the CMIQ to each potential
subject. Part A of the CMIQ requested
detailed data about the commuting habits
of the respondent's household and
determined if any member of the house-
hold was employed in one of nine
occupational categories associated with
high CO exposure. These data were
collected for use in SHAPE, a population
exposure model developed by Wayne Ott,
and NEM, a population exposure model
developed by the Strategies and Air
Standards Division of EPA. Part B of the
CMIQ verified the respondent's address
and attempted to set up an appointment
for the first visit by an interviewer. The
study questionnaire was administered to
each of 454 persons who actually
participated in the study. It included
detailed questions about the subject's
home environment, work environment,
commuting habits, occupation, leisure-
time activities, and shopping habits. The
study questionnaire also requested age,
sex and education data.
A PEM and an activity diary were pro-
vided to each subject for each of two 24-
hour periods. The PEM was a modified
General Electric (GE) Carbon Monoxide
Detector, Model 15EC53003, mated with
a modified Magus DL-1 Data Logger and
mounted in a compact, tamperproof
casing. The PEM recorded the time and a
CO concentration value every time the
"activity button" on the top of the
instrument was pushed and every hour
on the hour. In both cases, the CO value
was the integrated average CO concentra-
tion since the last recorded value. Each
PEM was capable of operating continu-
ously for 24 hours and logging up to 113
data points. Quality assurance activities
associated with the PEM's included daily
zero-span checks, frequent multipoint
calibrations, special studies evaluating
precision, and two independent audits.
The activity diary contained instructions
for completing the diary, examples of
properly completed diary pages, and 64
blank pages for recording activities. The
subject was instructed to fill out a diary
page whenever the subject changed
location or activity. Data entered on each
diary page included activity (e.g., cooking
dinner), location (e.g., indoors residence),
address, mode of transit if applicable, and
whether smokers were present. For
indoor locations, subjects indicated
whether a garage was attached to the
building and whether a gas stove was in
use.
Thirteen interviewers were employed
during the course of this study to deliver
PEM's activity diaries, and study question-
naires to the subjects according to
prescheduled appointments. Because
different PEM's and activity diaries were
used for the two sampling periods, an
interviewer visited each subject on
three consecutive days. In most cases,
the first PEM and activity diary were
delivered between 7 p.m. and 9 p.m. on
.Day A and picked up 24 hours later on
Day B. During pickup, problems encoun-
tered during the first sampling period
were addressed and a second PEM and a
second activity diary were delivered.
These were subsequently picked up 24
hours later on Day C. Breath samples
were taken during pickups on Days B and
C. A study questionnaire was delivered
on Day A and picked up on Day C.
A field data sheet was used to record
the PEM values and corresponding coded
activity diary data for each subject-day.
These sheets were validated using a
special computer program which checked
for 83 different types of data anomalies,
including missing entries, illegal entries,
and logical inconsistencies.
Breath samples were taken by having
each subject blow through a disposable
mouth piece into a 600 ml plastic
carboxyhemoglobin bag. To measure the
CO concentration of the breath sample, a
prefilter containing potassium perman-
ganate and activated carbon was inserted
between the mouthpiece and a General
Electric CO-3 portable CO monitor.
Fifteen fixed-site monitors operated in
Denver during the period of the study.
Nine of these monitors were temporary
and were discontinued at the conclusion
of the study. All of the monitors reported
-------
hourly-average CO data and operated
continuously.
Study Results
A total of 1094 subject-days of participa-
tion were scheduled. The 454 individuals
who actually participated in the study
yielded 900 subject-days; 446 subjects
participated in two sampling periods,
while 8 subjects participated in only one
sampling period. Of the remaining 194
subject-days scheduled, 120 were lost
because subjects requested rescheduling,
33 were lost because of last-minute
refusals -to participate, and 41 were lost
for other reasons (e.g., subject missed
appointment, interviewer experienced
car problems).
Of 899 data sets obtained from the
participants, 808 (90%) were coded as
acceptable for statistical analysis of PEM
values. Of the remaining 91 data sets, 50
were coded as unacceptable because the
difference between pre and post zero-
span values was judged excessive. Other
frequently occurring instrument problems
included clogged pumps, low battery
voltage, instances when the PEM logic
system switched out of the data recording
mode, and fragile parts.
Multipoint calibrations performed early
in the study revealed a potential nonlin-
earity problem in the low concentration
portion of the PEM's operating range. The
adverse affects of this nonlinearity on the
overall data quality were minimized by
insuring that the PEM GE sensor outputs
were properly balanced to the output of
the Magus data subsystem outputs.
The accuracy of PEM measurements
was determined daily based on a pre- and
post-sampling check of zero and span.
Using the change in slope as a measure of
accuracy, 93 percent of the measurements
were estimated to be within ±5 percent
of the-true concentration value. PEM's
operated in pairs showed a mean percent
difference in paired values of 5.0 percent
with a standard deviation of 14.2 percent.
PEM's attached to manifolds supplying
sample ambient air to fixed-site monitors
yielded paired values with a mean
difference of 8.3 percent (fixed-site being
higher) and a standard deviation of 22
percent.
A total of 859 data sets (96%) were
coded as acceptable for statistical analy-
sis of diary entries. In addition, 778 data
sets (87%) were coded as acceptable for
statistical analyses involving both PEM
and diary data.
A total of 859 breath samples were
obtained and successfully analyzed for
CO content. Thirty samples were lost
because of leaks in the sample bag. One
subject refused to provide a breath
sample, and another was unable to
provide a sample because of illness. Nine
samples were not obtained for other rea-
sons (e.g., subject could not fill breath bag).
The highest 1-hour CO concentration
reported by any of the 15 fixed-site
monitors during the study period was
44.1 ppm. Only one fixed-site monitor
(060580002F01) reported any daily
maximum 1-hour values exceeding 35
ppm, the current 1-hour NAAQS. The
highest 8-hour CO concentration reported
by any of the 15 fixed-site monitors was
20.7 ppm. Eleven of the 15 fixed-site
monitors reported daily maximum 8-hour
values exceeding 9 ppm, the current 8-
hour NAAQS. Five fixed-site monitors
reported daily maximum 8-hour values
exceeding 15 ppm.
The daily maximum 1 -hour and 8-hour
exposures calculated for the study
sample were extrapolated to the Denver
target population using weighting factors
which accounted for the probability of
selecting a particular subject into the
sample and for nonresponse caused by
refusals, instrument problems, and
unacceptable activity diary data. The
weighted means for daily maximum 1-
hour and 8-hour exposures during the
study period were 10.3 ppm and 4.9 ppm,
respectively. Approximately 3 percent of
the daily maximum 1-hour exposures
exceeded 35 ppm; approximately 11
percent of the daily maximum 8-hour
exposures exceeded 9 ppm.
Weighted linear regression analyses of
the daily maximum 1-hour and 8-hour
exposures predict that a member of the
Denver target population who receives a
daily maximum 8-hour exposure of 9 ppm
would receive a daily maximum 1-hour
exposure of 16.3 ppm. Similarly, a person
receiving a daily maximum 1-hour
exposure of 35 ppm would receive a daily
maximum 8-hour exposure of 16.1 ppm.
Using valid individual PEM values with
durations of 60 minutes or less, the
weighted means and standard deviations
of PEM values grouped by microenviron-
ment code were calculated. Listing the
microenvironments in descending order
by mean CO concentration suggests that
microenvironments associated with
motor vehicles had the highest CO levels
in Denver during the study period.
Occupancy period was defined as the
time a subject spends in a microenviron-
ment during a single visit. Mean occupancy
periods for in-transit microenvironments
associated with motor vehicles and high
CO levels were 30.8 minutes for trucks,
28.0 minutes for buses, 25.9 minutes for
cars, and 23.0 minutes for motorcycles.
An analysis was conducted of residen-
tial indoor exposures to determine the
contribution of three potential CO sources.
Mean exposure was increased 2.59 ppm
by gas stove operation, 1.59 ppm by
smokers, and 0.41 ppm by attached
garages.
Some models used for estimating
population exposure assume that a
strong, linear relationship exists between
CO levels in certain microenvironments
and CO levels measured simultaneously
at fixed-site monitors. This assumption
was investigated by performing linear
regression analyses that used PEM
values grouped by microenvironment as
the dependent variable and fixed-site
values as the independent variable. For
in-transit microenvironments, the inde-
pendent variable was the mean of the
simultaneously-recorded values at all 15
sites. For nontransit microenvironments,
the independent variable was the simul-
taneously-recorded value at the nearest
fixed-site monitor. Coefficients of deter-
mination (R2) ranged from 0 to 0.58. Most
were less than 0.50. Microenvironments
with R2 values exceeding 0.30 included
parks and golf courses, motorcycles, and
buses. The residential garage microenvi-
ronment yielded an R2 value of zero.
Diurnal patterns for weekdays, Saturdays,
and Sundays were developed for hourly
average exposures and composite fixed-
site values. In general, diurnal patterns
for exposure were similar in shape to
those for fixed-site data, although the
exposure patterns contained midday
peaks missing from the fixed-site patterns.
In general, this study suggests that 1)
the methodology proposed by EPA for
using personal monitors to estimate
population exposure to CO in urban
populations is sound, 2) CO exposures in
microenvironments associated with
motor vehicles are higher than exposures
in microenvironments not associated
with motor vehicles, and 3) CO exposures
in the microenvironments defined for
this study are not strongly correlated with
CO concentrations simultaneously recorded
at fixed-site monitors.
-------
Ted Johnson is with PEDCo Environmental, Inc., Durham, NC 27701.
G. G. Akland is the EPA Project Officer (see below).
The complete report, entitled "A Study of Personal Exposure to Carbon Monoxide
in Denver, Colorado," (Order No. PB 84-146 125; Cost: $23.50, subject to
change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use S300
i, U.S. GOVERNMENT PRINTING OFFICE: 1984—759-015/7617
------- |