United States
Environmental Protection
Agency
Health Effects Research
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S1-81 -065 August 1982
Project Summary
Effects of Short-Term
Intermittent Air Pollutants on
Incidence and Severity of Acute
Respiratory Disease: Data
Collection and Quality
Assurance
R. David Flesh, Margaret L. Riha, and Michelle F. Miller
The purpose of this study was to
measure the acute response of respi-
ratory disease from peak hourly and
daily average exposures to nitrogen
dioxide alone and in combination with
other pollutants. The study population
was made up of families with children
attending public elementary schools
and living in four California South
Coast Air Basin study areas: West Los
Angeles/Santa Monica, Garden
Grove/Westminster, Glendora/Covina,
and Upland/Ontario.
The role and responsibility of the
contractor were to collect all back-
ground and health data, to process
and assure the quality of the data, and
to prepare the report. The contractor
was assisted by the Statewide Air Pol-
lution Research Center at the Univer-
sity of California, Riverside (UCR), in
the data collection tasks.
Families were recruited for partici-
pation during the spring and summer
months of 1978. Recruitment included
face-to-face interviews for the pur-
pose of collecting family background
data. Over 3,000 families were inter-
viewed. From these, 1,402 families
(about 350 per study area) were
selected.
Participation required that families
record daily the disease symptoms
that they experienced. According to
instructions mailed to them, the symp-
toms were to be recorded on a spe-
cially prepared calendar. Each week,
the symptoms recorded the previous
week were given by telephone to a
staff of trained interviewers employed
by the contractor. Symptom surveil-
lance began in September 1978 and
ended in March 1979.
The background and symptom data
were reviewed and edited in the field
and in the contractor's main offices
prior to data entry. Computer files of
the background and symptom data
were prepared in three steps. First, the
data were transferred from the back-
ground and symptom questionnaires
to magnetic tapes through the use of
an optical reader. Second, supple-
mentary information including up-
dates to the background data and doc-
tors' diagnoses of illnesses reported
during symptom surveillance were key-
taped into a separate computer file.
Third, quality assurance procedures
were used to minimize error in the
computer files. Maximum allowable
error was set at 2%. The computer
-------
files and all questionnaires were then
sent to the EPA for analysis.
This Project Summary was deve-
loped by EPA's Health Effects Re-
search Laboratory, Research Triangle
Park. NC, to announce key findings of
the research project that is fully doc-
umented in a separate report of the
same title (see Project Report order-
ing information at back).
Introduction
This report covers the health data
collection phase of the project.
Delineating the Study Areas
Four study areas had been tentatively
chosen by the EPA for this study in
1976.
In December 1977, the EPA decided
to place a second air monitoring station
in each of the study areas. The purpose
of installing a second station was to permit
better characterization of air pollution
exposure of families enrolled in the
study. An added benefit of having a
second station was to expand the portion
of each area in which families could be
interviewed.
The criterion placed on the locations
of the second stations was that they be
located no more than two miles from
the existing EPA air monitoring stations.
An additional criterion stated that the
2-mile area around the new station
must be socio-economically comparable
to the 2-mile area around the existing
station.
The final shape of each of the study
areas was a rectangle, measuring two
miles by four miles. This was done to
improve estimates of human exposure
over what would have been obtained if
only one air monitoring station had
been located in each of the areas andto
increase the number of households
from which families could be recruited.
The study areas were divided into inter-
viewing zones so that the addresses of
elementary school families residing in
the areas could be plotted on the maps
and an orderly schedule of interviewing
could be established.
Background Survey Intervie wing
Lists of student and/or family names
and addresses were obtained by the
contractor. Using EPA-supplied maps,
the contractor's clerical staff listed the
streets and street number limits which
fell into each zone, making proper zone
assignment of each address an easy
task.
For a given interviewing zone, the
study area with the lowest number of
multiple family units determined the
number of multiple family dwellings to
be interviewed in all four areas. For
example, if a particular zone in West Los
Angeles had the least number of multi-
ple family dwellings of the four areas,
then that number of multiple family
addresses would be randomly chosen
from the corresponding zones in each of
the other three study areas. The same
procedure was used for single family
dwellings. The zone sample size was,
therefore, the sum of the lowest number
of multiple family dwellings and the
lowest number of remaining addresses
making the zone sample size the same
for all four study areas. This technique
resulted in an initial sample size of
1,294 homes in each study area.
A training session for interviewers
was held on May 18 and 19. It began
with a presentation given by the con-
tractor on the proper techniques of
interviewing in studies of scientific
concern. Following this, the inter-
viewers were given copies of the ARD
Study Background Questionnaire and
interviewers' instructions, which con-
tained information on the background
and purpose of the study, a sheet of
questions and answers about the study
that the interviewers were to use when
they contacted the school families for
interviews, an interviewing schedules,
and maps of the study areas.
After the formal presentations were
completed the interviewers practiced
administering the questionnaire to
each other. Sample interviews were
presented and critiqued, and then
actual interviews were practiced in the
field. The survey began in West Los
Angeles on May 22,1978, the first work
day following the training session. The
interviewers met with their team lead-
ers late in the afternoon, giving them
the questionnaires they completed that
day. During this meeting the interview-
ers received a new supply of blank ques-
tionnaires and address sheets for the
study area and zone where interviews
were to be conducted the next day.
Families who could not be contacted
on a previous visit were reassigned to
the interviewers. Every day, the inter-
viewers would first attempt to make
callbacks in previous zones before mak-
ing contacts with families in newly
assigned zones.
Interviewing was rotated from study
area to study area to avoid day of the
week effects and to maximize the
chances of finding the families at home.
A "round" of interviewing was said to
have been completed after each four
days of the survey. The final round
(Round 12) of interviewing ended on
August 1. The total numbers of inter-
views completed in each of the study
areas during the 12 rounds of interview-
ing were as follows: 740 in West Los
Angeles/Santa Monica, 754 in Garden
Grove/Westminster, 824 in Glendora/-
Covina, and 842 in Upland/Ontario.
Telephone validation of 10 percent of
the interviews completed by each inter-
viewer was conducted through Round
12 by non-interviewing personnel. No
significant differences were found be-
tween the initial interviews and the val-
idation interviews.
Selecting Families for
Participation
As the face-to-face interviews of fam-
ilies progressed through the 12 rounds,
the questionnaires were edited for
completeness and inconsistencies. The
answers given were checked to deter-
mine eligibility, also. Five criteria, deve-
loped by the EPA, had to be met for
families to qualify for participation.
1. Have access to working telephone.
2. Have lived in the study area for one
or more years.
3. Do not plan to move within 12
months.
4. Agree to participate in the weekly
telephone surveillance.
5. Have one or more elementary school
children living in the home.
Initially, priority for selection was to
take the following factors into consider-
ation:
1. Presence of the pre-school children
in the family.
2. Length of residence in the study
area:
a. Three or more years
b. One to three years
3. Parents livng the home:
a. Both parents
b. Female parent
c. Male parent
4. Type of fuel used for cooking:
a. Gas
b. Electricity
-------
5. Type of air conditioning in the home:
a. Central
b. Room
c. None
6. Type of dwelling unit:
a. Single Family
b. Multi-family
7. Presence of older family members
(e.g., grandparents) living in the
home.
It was planned that the study areas
would be matched in terms of these fac-
tors as well as in terms of the socioeco-
nomic characteristics. However, as the
responses to the Background Question-
naires were being coded, it quickly
became clear that the study area could
not be completely matched and that all
of the factors could not be accommo-
dated if the selection process was to
yield the quota of at least 350families in
each study area. Therefore, a decision
was made by the EPA to use 80% of the
families who cook with gas and 20% of
the families who cook with electricity in
each study area where this combination
was possible (in West Los Angeles,
Garden Grove, and Glendora) and a
50%: 50% ratio of gas to electricity where
the 80%:20% ratio was not possible (in
Upland). The EPA also decided that 2-
parent families with no pre-schoolers,
living in single family dwellings, should
be given lowest selection priority and
should be randomly sampled to complete
the quota of 350 panel families in each
study area. Other factors, including the
type of air conditioning and the pres-
ence of older family members living in
the home, had to be dropped completely
as selection criteria, since too few homes
in the West Los Angeles study area had
air conditioning and too few school fam-
ilies contained older people.
A total of 1,400 families (4 study
areas x 350 families per study area)
were selected and randomly distributed
into 14 groups in preparation for weekly
telephone surveillance. Each group was
assigned to a different telephone
interviewer.
Weekly Telephone Interviewing
In August 1978, the contractor and
subcontractor each recruited seven
telephone interviewers and two alter-
nates in preparation for weekly symptom
surveillance. All interviewers had pre-
viously been members of the background
survey teams. A training session was
held on September 6 and 7 with the
interviewers, alternates, and supervi-
sors in attendance. An EPA instruction
manual on how to complete the ARD
Study Symptom Questionnaire was dis-
tributed to each of the interviewers
prior to the briefing. A review of the
instructions was followed by a demon-
stration on conducting the weekly symp-
tom surveillance calls. After a period of
questions and answers, the interview-
ers practiced conducting weekly inter-
views.
Pre-surveillance telephone calls were
conducted during the week of Septem-
ber 11. During the presurveillance calls,
an interviewer ascertained, without
directly asking, whether or not the re-
pondent still wished to participate in the
panel. Respondents were asked if they
had received their calendars and if they
had any questions relating to the use of
the calendar. Next, a day and time for
the weekly telephone calls was estab-
lished with each respondent. Finally, all
information on the background of the
respondent's family was updated.
Families which stated, during the pre-
surveillance call, that they were no
longer interested in participating in the
study, were replaced by eligible families
which were selected as alternates. A
withdrawn family that cooked with gas
was replaced with another family from
the same study area that also cooked
with gas. This was done to maintain the
existing balance of selection factors and
the quota of 350 panel families in each
study area.
The families began to keep records of
symptoms on their calendars on Sep-
tember 10. The first week of symptom
surveillance telephoning took place dur-
ing the week of September 18. Each
interviewer was provided with 100 pre-
printed ARD Study Symptom Question-
naires, one for each of the families to
which the interviewer was assigned.
All weekly calls to families were made
on Mondays and Tuesdays whenever
possible. The records of contacts with
all families were reviewed at regular
intervals by the field supervisors to
ascertain how many of the families'
surveillance calls had to be made on
days other than Monday or Tuesday,
how many weeks information was taken
from an alternate respondent, and how
many weeks the family was unable to
be contacted. If a family reported on an
alternate day, used an alternate
respondent, and/or failed to give a report
entirely, for more than eight weeks, the
family was withdrawn from the study.
Onlyduring the Week 1 were families
which withdrew from the study replaced
by alternates. After Week 1, no further
replacements were made.
Any changes in background informa-
tion previously supplied by families
were recorded as they occurred during
the 12 of surveillance. A special form
was used for this purpose.
After every six weeks of symptom
surveillance, i.e., after Weeks 6,12, and
18, the 14 groups of families were ran-
domly reassigned to the interviewers. In
accordance with the instructions of the
EPA Project Officer, no interviewer was
assigned the same group of families she
had interviewed previously.
During Weeks 12 and 24 random
samples of 10% of the completed inter-
views from each interviewer's group
of families was selected for validation.
The selected families who had been
interviewed on Monday or Tuesday of
Weeks 12 and 24 were reinterviewed on
Wednesday or Thursday of the same week
by the field supervisors. The integrity of
the validation procedure was enhanced
by the exchange of assignments
between the field supervisors, i.e., the
prime contractor's field supervisor vali-
dated the interviews conducted by the
subcontractor's interviewers and vice
versa. Once again, no significant differ-
ences were found between the initial
interviews and the validation interviews.
The final session (Week 26) of symp-
tom surveillance telephone calls was
conducted on March 12 and 13. The
numbers of families who participated
through the 26 weeks of the study were
as follows: 316 (90.3%) in West Los
Angeles, 290 (82.9%) in Garden Grove/
Westminister, 300(85.7%) in Glendora/
Covina, and 310 (88.6%) in Upland/
Ontario.
Assuring the Accuracy of the
Data Files
As a preliminary step in testing the
accuracy of the data files, computer
range checks were performed for all
questions in order to verify that all
answers were within a valid range.
Invalid responses were checked,
receded, and corrected on the data tapes.
The quality assurance testing proce-
dure employed sampling inspection
tabjes based on a Poisson distribution,
which sets the probability at 90% that
less than 2.0% of the forms have one or
I US.GOVERNMENTPRINTINa OFFICE. IMi-559-017/0781
-------
more errors. For a quantity of between
3,001 and 4,000 Background Ques-
tionnaires, 510 questionnaires had to
be drawn at random, and checked against
the data from those questionnaires
that exist inthe computer files. According
to this sampling plan, the maximum
number of acceptable errors was six.
For the purpose of counting errors, an
error was considered to exist if one or
more responses in the computer file of a
particular questionnaire differed from
the corresponding responses on the
actual questionnaire.
A 100% check on the Background
Questionnaires was conducted as a
penalty for failing the sample test. Cor-
rections to the background data files
were made and a new random sample
of 510 questionnaires was selected and
compared to the computer files twice
before the computer file passed the test.
The background data files, after passing
the quality assurance test, were sub-
mitted to EPA on two magnetic tapes.
For a quantity of between 1,001 and
2,000 Symptom Questionnaires a ran-
dom sample of 380 family records had
to be selected. This was done for each of
the 26 weeks of data. The computer
files corresponding to the random sam-
ple of 380 questionnaires were allowed
to have a maximum of four errors in
order to be within the acceptable levels
of the sampling plan. An error existed
when a family record in the computer
files did not exactly correspond to the
actual questionnaire for that family.
The first three weeks sampled failed
with a great many errors and the deci-
sion was made to perform a 100% check
of all 26 weeks of symptom data before
doing any further random sampling. All
26 weeks were 100% checked by com-
paring the computer files to the original
written forms.
Near the end of the contractor's au-
thorized level of effort, random samples
were drawn from each week of surveil-
lance. Weeks 1, 4, 6, 13, 20, 21, 22,
23, 25, and 26 passed the test, since
they had four or less errors per sample.
The remaining weeks were again sub-
jected to 100% error checks and correc-
tions were again made to the computer
files to eliminate the errors encoun-
tered. The symptom data files were
submitted to the EPA on 26 magnetic
tapes, one for each week of symptom
ij surveillance. EPA continued processing ;i|:
|.: the weekly surveillance forms that hadg
:jfailed the quality assurance test until*:
jthey also passed. 8
This contract was for collection of
health data only. HERL, RTP, USEPA is
in the process of analyses and interpre-
tation of this data set.
R. David Flesh, Margaret L. Riha, and Michelle F. Miller are will Del Green
Associates, Inc., San Diego, CA 92131.
Wilson B. Riggan is the EPA Project Officer (see below}.
The complete report, entitled "Effects of Short- Term Intermittent A ir Pollutants
on Incidence and Severity of Acute Respiratory Disease: Data Collection and
Quality Assurance," (Order No. PB 82-129 479; Cost: $21.00, subject to
change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Health Effects Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Postage and
Fees Paid
Environmental
Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300
------- |