UNITID STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, O.C, 204SO
October 9. 198? SWID48-OQ2
The Honorable Lee «» Thomas
Acteinistrator
0. S. Environmental Protection Agency
401 M Streetr S. w*
Washington, t>, C. 20410
Pear Mr. Thoaasi
. On June 18, 1986, the Office of Radiation Programs askefl the
Science Advisory Board*® Radiation Advisory Committee to review the
design of the national radon survey, the CoRuiittee accepted the request
and fowled & Si&coflntittee, chaired by t$r* Qctfvar Hygaarf of Case Western
Reserve University, to eondyet the review,
• Th« SifccooBtitt** wet for the first time en Sefrtertber 24, l§8i» but
did nest prepare a report at that tine because the Office of Radiation
Programs planned substantial revisions to the design presented.
The revised document, Survey Design for the iSaticttal.jtedon. Survey,
which outlined the current plans for the national radon survey, was
transmitted to the Committee and Subcommittee April 30, lit?. The Siisoam-
oiittee met on June 10-11, 1§87, in Washington, D. C, to review the tip<5u«ient
artS prepare its report. The Radiation Advisory Comnittee heard and
discussed, the SiJasoaiRittee's report Oft Jizie 12, 1987.
Generally, the Subcommittee finds that the document presentee! a valid
approach to dtesiping a national radon survey* The SiixaMnittee's other
major conclusions and recoBmendations are summarized belowt
1, The primary objective of the national radon survey? that
of determining the nationwide fnqpuensy distribution of radon
ccficentrations in residential structures, is an important cane
and achievable with adequate precision within the study design.
Achievement of secondary objectives, such ts regional distribution
estimates* is also possible. However, over/sampling (disprqportion-
ate stratificationJ must not be employed to satisfy requirements /
for such regional estimates since ovewappling' will adversely
iHfsact the quality of the national figures. CSee «lis
-------
2. A pretest should be conducted to determine the relative advantages
of a telephone survey versus face-to-face interviews. Although
cost is an obvious and legitimate concern of the Agency, the
choice of the interview method for the national radon survey
should be chosen primarily on considerations of the quality of
the data collected rather than on cost.
3. Data should be also obtained to allow estimates of effective
dose equivalent occupancy figures to individuals, which will be of
signifieant value to EPA and other government agencies. The
effective dose equivalent provides a better representation of the
radon exposure than concentration since this unit of dose can be
readily translated in terms of potential health effects to the
public.
4. In addition to providing data on indoor radon concentrations,
the national radon survey could be a valuable mechanism for
investigating correlations between radon and certain variables.
The insights into factors affecting radon levels providedby
this form ofanalysis constitute a major contribution of the
national radon survey. To realize this contribution, additional
data should be collected including! bedrock geology, climate,
basement radon levels, housing characteristics and household
heating/air conditioning practices. (See discussion under Issue
14 of the report.)
5. The Subeotwittee identified areas of concern that the Agency
should address or clarify. For example, rental units should be
included in the national survey. It is noted that the Superfund
Amendments and Reauthorization Act (SARA, P. L. 99-4991) stated
that EPA was to include, "structures where people normally live
and work, including educational institutions" in the national
assessment of radon gas. However, only residential structures
were included in the objectives covered in the Office of Radiation
Programs' June 18, 1986 memorandum to the Science Advisory
Board,
Additional areas of concern are criteria for evaluating the
pretest results? identification of homogeneous strata for the
sample design; a detailed description of the approach to defining
sample size; identification of subgroup populations? minimum
sensitivity of radon concentration measurements? and the need
for defining how the survey will deal with multiple measurements
in a single dwelling unit.
6, The Subcommittee also recommendsthat the Agency consider how
the results of the national survey will be used and how the
infonnation. derived will be reported. Such planning is not
described in the current document and should be added. Explicit
-------
consideration of how the results will be used, prior to initiating
the survey will greatly enhance both data collection and analysis
and thereby strengthen the quality and defensibility of the
study.
Both the Radiation Advisory Committee and the National Radon Survey
Design Subcommittee conclude that this study is important from a national
health point of view, and that all efforts must be made to insure that a survey
of high quality is conducted. As it develops an actual survey design
fran the plan presented to the Science Advisory Board, the Agency should
bear in mind that an inadequate national radon survey would be a disservice
because it might well preempt the execution of any future study of
significant scientific value.
The Committee appreciates the opportunity to conduct this review and
would be pleased to discuss it further with you. We also acknowledge the
cooperation of the Office of Radiation Programs. We request a fotroal
response to the conclusions and recommendations presented in the attached
report.
' Sincerely,
Norton ftelson
Chairman
Committee
Till
irraan
'on Advisory Gonniittee
CGJ S. Meyers
T. Yosie
-------
REVIIN OP THE
OFFICE OF RADIATION PROGRAMS'
NATIONAL RADON SURVEY DESIGN
by the
NATIONAL RADON SURVEY DESIGN SUBCOMMITTEE
RADIATION ADVISORY QC*KITrEE
SCIENCE ADVISORY BOARD
U. S. mWIRONMENTAL PROTECTION AGENC*
October 1987
-------
NOTICE
This report has been written as a part of the activities of the
Science Advisory Board, a public advisory group providing extramural
scientific information and advice to the Administrator and other officials
of the Environmental Protection Agency. The Board is structured to
provide a balanced expert assessment of scientific matters related to
problems facing the Agency. This report has not been reviewed for approval
by the Agency and, hence, the contents of the report do not necessarily
represent the views and policies of the Environmental Protection Agency,
nor of other agencies in the Executive Branch of the Federal government,
nor does mention of trade names or comnercial products constitute
endorsement or recommendation for use.
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISOR?, BOARD
RADIATION ADVISORY COMMITTEE
NATIONAL RADON SURVEY DESIGN SUBCOMMITTEE
ROSTER
CHAIRMAN
Dr. Qddvar Nygaard, Department of Radiology, Case Western Reserve
University, Cleveland, Ohio 44106
MEMBER/CONSULTANTS
Dr. Graham Kalton, Institute for Social Research, Department of
Biostatistics, University of Michigan, Ann Arbor, Michigan 48106-1248
Dr. Leonard LoSciuto, Institute for Survey Research, Temple University,
Philadelphia, Pennsylvania 19122
Dr. Jaqueline Michel, Research Planning Institute, Post Office Box 328,
Columbia, South Carolina 29202
Dr. Donald Sehut2, Teledyne Isotopes, 50 Van Buren Avenue
Westwood, New Jersey 07675
Dr. Richard Sextro, Building Ventilation and Indoor Air Quality Program,
Lawrence Berkeley Laboratory, Building 90, Poofn 3058, Berkeley,
California 94720
Executive Secretary
Kathleen W, Conway, Deputy Director, Science Advisory Board
U.S* Environmental Protection Agency, 401 M Street, S.W.,
AlOl-F, Washington, D.C. 20460
Staff Secretary
Dorothy M. Clark, Secretary, Science Advisory Board, U.S. Environmental
Protection Agency, 401 M Street, S.W., AlOl-F, Washington, D.C. 20460
-------
TABLE OF CONTENTS
PAGE
I* Responses to the Issues Raised by, Of*P 1
A. Issue tl-Qbjeetives 1
B. Issue *2-Stratifieation 3
C. Issue |3-Samples Size 4
D. Issue 14-Subgroup Populations . 6
II. Additional Sifccommittee Contents 7
A. Mode of Data Collection 7
B. Relationship Between Indoor Radon Levels and Soils 8
& Inclusion of Housing Construction Data in the Questionnaire 8
D. Detector Placement 8
E, Questionnaire •. 8
P. Incentives 9
G. Minimum Radon Concentration Sensitivity 9
H. Derivation of Representative Radon Concentrations 9
-------
I, RESPONSES TO THE ISSUES RAIS1D B¥ ORP
On June 18, 1986 the Office of Radiation Programs referred four
issues relating to the national radon survey to the Radiation Advisory
Committee of the EPA's Science Advisory Board. The issues as stated by
the Agency and the comments of the National Radon Survey Design Subcommittee
follow.
A. ISSUE tl-^OSJECTlVES
As stated in CRP's June 18 memorandum, "The primary objective of
the national radon survey is to determine the frequency distribution of
radon concentrations in residential structures. This will provide data
on the average indoor residential radon levels to which the population
of the United States is exposed. It will also provide information on the
number of homes that exceed various radon levels. Another objective is
to investigate relationships between several parameters, including geography,
geology, and house construction with indoor radon levels.
"We would like the Committee to review these objectives and
determine if they are consistent with the needs of the scientific conmunity,
as well as providing recommendations for other objectives that may be
appropriate."
1. Primary Objectives
The Subcommittee finds that the primary objective of the national
radon survey, to describe the overall frequency distribution of radon
concentrations within residential structures in the United States, requires
some refinement. For example, whether residential structures refers only to
owner-occupied single family homes or also includes rental units, and
apartment/condominum complexes, influences the design of the survey
and the utility of the results.
The Subcommittee does not believe that the Agency should assume, a_
priori, that the distribution of radon in dwelling units is a single Tog-
normal distribution. The Agency should empirically determine the nature
of the distribution* Although the same size may not be sufficient to
evaluate whether observed radon concentrations in the tail belong to a
separate distribution than the lower concentration values (as might be
the case), the shape of the high-concentration tail should be determined
as accurately as possible.
The Subcommittee is very concerned about the bias that may arise if
only owner-occupied housing is sampled. Approximately 60% of the dwelling
units in the United States fall into the owner-occupied category. The
-------
- 2 -
survey also plans to exclude participants who neve during the year. About 20%
of the U. S. population move annually. Because of the logistical difficulties
of getting owner and occupant permission to sample rental units, the higher
moving frequencies in such units, and the subsequent decrease in response
rates, the present study plan ignores about half of the housing stock.
The restriction of the survey to residences already contrasts with SARA
language concerning the workplace and schools.
It is possible that radon levels in rental units may be, because of
the hi^ier frequency of multiple-story apartment complexes, less energy-
efficient rental units, and other factors associated with multiple-family
dwellings, if this is the case, the exclusion of rental units would
introduce an unknown bias. By sampling only owner-occupied units, the
survey could over-estimate indoor radon levels.
If it is not possible, for legal or other reasons, to collect indoor
radon concentration measurements from rental units and units from which a
family has moved, as much data on these units {e.g., questionnaires) should
be collected to permit comparison with the dwelling units for whicfr EPA has
measurements.
2. Secondary objectives
Two secondary objectives for the study were subsequently reiterated
by QRP in the June 10, 1987 briefing package, National Radon Survey. They
were: (1) to "investigate whether building characteristics affect the
distribution of radon levels," and (2) to "investigate whether, on a broad
regional basis, the distribution is affected by geological and soil structure
features." ORP noted that the secondary objectives would be "pursued to
the extent that available resources make the overall sample size large
enough to support separate estimates for these reporting groups."
Although both secondary objectives are valid, the second does not
appear to be feasible within the limits of the present study design. If
the country were divided into sufficiently homogenous regions, there would
be so many and some would be so small that, for the majority of regions,
it would be inpossible to obtain regional estimates of adequate precision
with the kind of sample size that is envisioned.
3, A__Ngy__Seoendary Objective-^Inforrotion gn^Occupancy Patterns
The survey should attempt to address exposure to individuals.
Although the study is not optimally designed for exposure assessment,
the opportunities to learn a great deal from very little extra work at
low marginal cost are great. For example, information could be
collected via the questionnaire on the number and ages of persons,
-------
- 3 -
and the approximate occupancy patterns of the various rooms in the
house for each person. There may be important differences in room
and floor use among regions and housing types. These occupancy data
can be used, along with the radon data, to provide more realistic
exposure estimates.
B. ISSOE I2-STRATIPICAT10N
AS stated in CRF's June IB, 1986 memorandum, "Our present plan calls
for the use of stratification based upon geologic characteristics, because
we believe this is the parameter most direct ly relevant to indoor radon
concent rat ions. Ten geologic regions have been defined for stratification
of the sample used for the national radon survey.
"We request that the Committee: (1) evaluate the present plans
for stratification by geologic regions and make recommendations for
stratification based on other parameters, if appropriate, and
(2) recommend whether minimum precision requirements should be
established for each stratum.
"Two main purposes of stratification have been identified for
the national radon survey;
(1) To increase the level of precision of national estimates.
(2) To permit estimates of adequate precision to be obtained
for specific strata.*
the first purpose, and assuming no cost difference in data collection
for different strata, the optimum allocation of the sample across the strata
is achieved by setting the sampling fraction in a stratum proportional to
the standard deviation of the variable of interest in that stratum. Often
the standard deviation is approximately constant across the strata, in
which case the optimum allocation is obtained by use of the same sampling
fraction in each stratum. This proportional allocation produces a sample
with the same distribution across the strata as the population.
For the second purpose, the sample fractions in small strata of
analytic interest might be increased to make the sample sizes in those
strata large enough to produce strata estimates of adequate precision,
However, such a disproportionate allocation means that, for the same
overall sample size, the national estimates will be less precise than
would be the case with a proportionate allocation. The loss of
precision can become substantial when a sizeable variation in sampling
fractions occurs.
-------
- 4 -
The Subcommittee endorses the plan to include geogr^jhical and other
variables (such as soil and climate, but not housing structure) in the
stratification for the national radon survey. However, it does not
consider that separate estimates for individual regions are of sufficient
importance to warrant the use of a disproportionate allocation to insure
adequate sample sizes in the smaller regions. Rather, it recommends the
use of a proportionate, or at least ajn approximately proportionate*
allocation. Although a proportional allocation will not provide estimates
of adequate precision for the smaller strata, it will enable separate
estimates to be made for larger strata and for groupings of strata.
">.,.,,
The gains in precision of national estimates from the use of
proportionate stratification depend on the way in which the strata
are formed* The greater the homogeneity of the survey variables
within strata, or equivalently the greater the heterogeneity of the
variables among strata, the greater the gains in precision. Thus,
the aim is to choose stratification factors that divide the population
into strata, each of which is homogeneous with respect to the survey
variables,
While the eight regions delineated on the basis of generalized
physiographic, hydrologic, and geologic features as presented in
the proposed national radon survey design have some validity with
regard to grouping homogeneous features, they actually encompass many
diverse geologic environments. These diverse environments could be
expected to produce radon in both high and low concentrations in
soil gas. For the purpose of sample design, the Subcommittee
recommends that these regions should be further subdivided to create
smaller, more homogeneous strata* This would improve the precision
of the national survey with little additional cost. This siiadivision
could involve creating sarolleir strata that are more homogeneous
with respect to geology, soils and climate. Further stratification
in the North-South and Bast^West directions would provide control
with regard to cold-hot, wet-dry conditions.
C. ISSUE 13-SAMPLE SIZE
As stated in QRP's June 18, 1986 memorandum, "The sample size
for the national distribution is directly related to two questions.
First, the degree of accuracy with which we want to define the national
distribution, and second, whether there are subgroups for which
separate estimation capabilities are needed. The sample size will
depend primarily on the national distribution, on the number and
characteristics of subgroup populations for which frequency distributions
are desired, and on the level of accuracy needed for the distributions.
-------
- 5 -
"We request that the Committee address the issue of sample
size and make recommendatleans on the sample size appropriate for
the survey."
The sample size for the national radon survey must be large
enough to provide overall national estimates of adequate precision.
The critical estimates are the percentages of dwelling units in the
upper tail of the radon level distribution, for instance with radon
levels above 4 or above 10 pCi/1* The question of what is an adequate
level of precision for such estimates is debatable. If, for example,
the sample estimate of the percentage of dwelling units with radon levels
above 4 pCi/1 is 7%, a 95% confidence interval for the population
percentage of 6% to 8% would be acceptable, i.e., a standard error of
plus or minus 0.5%, The sample size required to give a standard error
of this magnitude depends on the sample design employed.
Althou^i the Subcommittee believes it would be inappropriate
for it to calculate the requisite sample size for the Agency, and in any
case lacks the information with which to do so, the following comparison
illustrates the impact design may have on sample size. For illustrative
purposes, suppose that for an unclustered telephone sample design, a
sample size of about 2500 might suffice. A clustered face-to-face
interview survey design would require a larger sample size (perhaps 4000
dwelling units). The actual difference in sample size depends on how much
cluster there is. Clustering of face-to-face interviews results from
the practical and economic need to make good use of the interviewer's
time by selecting dwelling units in groups ("clusters") that can be
interviewed efficiently.
Another issue in the determination of sample size is the need to
provide subgroup estimates of reasonable precision. This issue is often
the driving force in determining sample size. It would be desirable to
expand the sample size above the minimum for national estimates in order
to enable subgroup estimates of adequate precision to be produced, and
to enable useful comparisons of subgroup estimates to be made*
Under issue 14 below, the Office of Radiation Programs states its
intention to develop separate estimates for single versus multiple family
houses. The Subcommittee, as noted above, believes that it is also
necessary to look at rental units. The need to address more than owner
occupied single family houses in a national survey is important and
should be taiken into account in the sample size determination. Whether
the sample size will be adequate for this task will o*epend on the
proportion of such houses in the population. In determining final
sanple size, consideration should be given to whether the number of
multiple family houses will be adequate to provide separate estimates
of useful precision.
-------
- 6 -
Provided that the sample size is adequate, the Subcommittee
believes that the survey will be a valuable source for investigating
the correlates of radon. Thus, for instance* the survey data dan
be analyzed to examine the relationship between radon level and
soil types, housing characteristics, housing insulation, and household
practices in heating, and air conditioning. The insights into factors
affecting radon levels provided by this form of analysis constitute
a major potential contribution of the survey.
The Subcommittee agrees that an achieved sample of 2500 dwelling
units is the minimum needed to accomplish the primary objective of
the study. This size does not include any safety factors to allow
for the uncertainties in design factors and sources of error. It
will allow some of the subgroup analyses that wight be of interest. For
further subgroup analysis, which is an impjcnrtant part of the study, the
sample size should be increased.
The Subcommittee recommends that the assumptions and calculations
regarding determination of sample size be presented in a more straight-
forward and logical manner. If possible, this should be summarized
in a table or figure.
D. ISSUE 14-SJBGRQUP POPULATIONS
As stated in ORP's June 18, 1986 memorandum, "The subgroup
populations for which separate frequency distribution estimates are
desired influence the design in one of two ways. If a specified
precision is needed for the subgroup's estimate, then the overall
sample size for the survey may be increased. If a subgroup is of
secondary interest, pertinent questions can be included in the
questionnaire but the sanple size will not be changed to guarantee
precision for that subgroup estimate. We currently plan to establish
specific precision requirements only for the subgroups of the geologic
regions and the single versus multiple family houses.
"We request recommendations as to the subgroups that the Committee
believes to be of interest. In addition, we would like the Committee
to distinguish those subgroups that are of sufficient importance that
separate estimation capability be insured (via specified minimum
precision requirements.)"
The Subcommittee strongly believes that when the study is complete,
the Agency should be able to analyze the data for two factors at once.
For example, comparisons of radon levels in dwelling units with basements
and selected soil types by geographic regions would be of great interest.
-------
- 7 -
The factors, or subgroups or greatest interest are:
1. soil-type characteristics
2. outdoor temperature
3, rainfall
4. bedrock geology
5, substructure type
6. heating and cooling systems (e.g., forced air vs. other types)
lf constnjction techniques/materials (e.g., poured vs. block walls)
II. ADDITIONAL SUBCOMMITTEE OOMHfNTS
A. MODE Of1 DRTft
One of the most critical aspects of the study is the mode of data
collection. The CRP document extensively discusses the relative merits
of face-to- face and telephone data collection, and proposes a pretest to
determine which approach would be the most cost-effective way of collect-
ing data of the highest quality, but no criteria were presented for the
evaluation of the pretest and for making a final selection.
The Subcommittee is concerned that this decision not be dictated
by cost, because the major issue here is data quality, especially with
regard to the accuracy of the information collected on the questionnaire,
proper placement of detectors in the dwelling units, recovery rate of
exposed detectors, and overall survey response rate.
Nor should the survey design be overly driven by the higher
geographical spread obtained by the use of telephones. The increased
overall precision of the results dye to the better geographical spread
may be offset by the increased quality in data obtained with the question-
naire by the face-to-face method.
The Subcommittee agrees with ORP that a pretest should be conducted
to select the preferable data collection method, and to refine the survey
instrument. In addition, the Subcommittee recommends that ORP review the
survey conducted by the State of New Jersey to help evaluate the face-to-
face approach. The recently completed New Jersey survey consisted of
a sample of dwelling units. The Office of Radiation Programs might find
it useful to discuss the New Jersey survey with the investigators,
especially the faee-to-faee interviews, to detetraine what lessons were
learned that could be applied to the conduct of the national radon survey,
-------
- 8 -
B. RELATIONSHIP ,HE1WEEN INDOOR RADON LEVELS AND SOILS .-••?
• • Based on preliminary results of research presented at the
Subcommittee's September 24, 1986 meeting by the U.S. Geological
Survey it appears that, in considering factors which influence
indoor radon concentration levels, soil characteristics are as
important as housing construction. Information on soil type should
be included in the survey data collection. During the pretest, the
feasibility of collecting soil type data should be evaluated. Soil
information is mapped on detailed aerial photogaphs published for
each county by the Soil Conservation Service. Soil type is probably
more important than geological siistrata as a factor controlling
indoor radon levels.
C. INCmSlON OP HOUSING OONSTRUCTI(H_ DMA IN THE QUimONNAIgE
The Subconmittee .believes that the housing construction items in
the proposed questionnaire are very important and should definitely be
included. One of the most important goals of the EPA radon strategy is
to collect information on the factors that contribute to elevated indoor
radon levels. Housing'Construction is considered as important as soil
characteristics. Statistical analysis of the survey data can be used to
attempt to assess the relative importance of the various factors on a
regional basis.
D* DETECTOR PLACEMENT . • '
The current ORP plan is to place the detector for a single-level
dwelling unit in the roaster bedroom. The Subcommittee recommends
that the detector be -placed in the common living space rather than the
master bedroom to conform with current EPA guidelines. This placement
will also provide a comnon basis for comparison among the various dwelling
unit types. In addition, in all housing units with basements, a detector
should also be placed in the basement (regardless of whether it is used
as a living space) so that results of this study may be more easily
compared with other studies of indoor radon concentration. Also, such
measurements will permit correlation of radon concentration in basements
with concentration on upper floors,
E. QUESTIONNAIRE
CRP has revised the questionnaire since its initial submission to
the Sybconmitteei and its plans additional revision before it is administered.
The Subcoomittee hopes that it will be comprehensible to the respondents
and that all the proper topics (such as household composition) will be
included in the final questionnaire.
-------
— Q —
P. INCTHTIVE5
A good response rate contributes to a sound survey, and the Agency
is considering whether incentives might enhance that response rate.
Some Subcommittee members questioned whether the proposed S5 incentive
would be effective. The experience of New Jersey, which used a $5
incentive, and of the Bureau of the Census, which gives credit card
calculators as an incentive, might be informative. It is possible
that people may participate simply to obtain the survey results— the
radon measurements of their homes {for which they might otherwise
pay $50).
Q. MINIMUM RADON CSNCSm&TION SENSITIVITY
In order to determine the shape of the distribution accurately, the
minimum sensitivity will have to be better than 0,5 pCi/1 and, if possible,
closer to 0.1 pci/l. The documentation provided the Subcommittee does
not address this point.
H. DERIVATION OF REPRESENTATIVE BADON
In the national radon survey, as currently designed, some dwelling
units may be characterized by a single radon measurement. In other
dwelling units, more than one measurement may be made. Such multiple
measurements are a bonus rather than a problem because they five the
Agency flexibility in its analysis. While the data can, and should be,
analysed in a variety of ways, a common basis for summarizing data from
each dwelling unit needs to be specified so that comparisons can be made
within and among the dwelling units. The living space concentration,
measured in each dwelling unit, represents the common data point in all
sampled dwelling units and might be used for such comparison. However, for
dwelling units with multiple measures, a scheme for weighting individual
data might be developed to yield a representative radon concentration to be
used in a national radon distribution.
------- |