Assessing Occupant Reaction To Indoor Air Quality: Gary J. Raw, D.Phil

Ch53_Spengler_445549-4 11/20/00 10:41 AM Page 53.
CHAPTER 53
ASSESSING OCCUPANT
REACTION TO
INDOOR AIR QUALITY
Gary J. Raw, D.Phil.
Building Research Establishment Ltd.
Garston, Watford, United Kingdom
Occupant surveys are widely used to assess the reactions and responses of occupants to the
air quality in their indoor environments; such surveys are a powerful tool in both research
and practice in the field of indoor air quality (IAQ). This chapter describes the role of occu-
pant surveys and how to use them most effectively; the chapter is divided into the follow-
ing four main sections:
1. The role of occupant surveys. Why would someone want to carry out an occupant sur-
vey at all? How do occupant surveys fit into the wider picture of IAQ? What can they
achieve and what can they not achieve?
2. Deciding to conduct an occupant survey. Given that there is a general case for con-
ducting occupant surveys, how should someone (e.g., a building manager or a
researcher) decide specifically when, where, and for what purpose to conduct a survey?
3. Instruments for the survey. This is the first part of conducting a survey: choosing the
right instrument (usually a questionnaire) for the job.
4. Procedures for the survey. Having chosen the instrument, how should the survey be
conducted?
For each of these items, the issues are more complex than many suspect, but with a little
understanding of the principles and available methods, an approach that is both manage-
able and effective can be created.
53.1 THE ROLE OF OCCUPANT SURVEYS
Introduction
This section provides the theoretical underpinning for the three more practical sections that
follow. It addresses the basic issues of why would someone want to carry out an occupant
53.1
Ch53_Spengler_445549-4 11/20/00 10:41 AM Page 53.2
53.2 ASSESSING IAQ
survey at all and how such surveys fit into the wider picture of IAQ. Anyone carrying out or
commissioning an IAQ survey should seek to acquire an understanding of these fundamentals.
Much of the discussion in these chapters is focused on surveys of groups of 50 or more
people sharing a common environment (e.g., the occupants of an office building). Although
the general points will apply to other kinds of environment (e.g., individual homes), much
of the specific guidance would need to be adapted.
The Primacy of Occupant Reactions
The starting point for this discussion is to question, from a theoretical perspective, what we
mean by an IAQ problem and how we can know when one exists. This chapter is concerned
with the impact of indoor pollution on people, as distinct from effects on building and fur-
nishing materials, animals, and plants. In this context, there are three main ways of defin-
ing a problem: health (either identifiable illness or the occurrence of nonspecific
symptoms), comfort, and productivity. Of course, criteria for IAQ will often find expres-
sion in terms of environmental variables such as pollutant levels or ventilation rates, but the
bases for the criteria are human responses.
It follows that occupants have a key role in defining the quality of the air in their indoor
environments. Objective measurements (e.g., contaminant concentrations) have the attrac-
tion that they are generally reproducible and that it is possible to define precisely what is
being measured. However, it is not always so clear that they are directly relevant to human
responses. It is possible for an investigator (whether a research scientist or IAQ practi-
tioner) to be precisely and accurately measuring the wrong environmental parameters or
measuring the right parameters at the wrong time or place. This is not to suggest that the
investigator is incompetent: it is simply a fact that we have an inadequate understanding of
how complex mixtures of air pollutants (together with other environmental, social, and per-
sonal factors) determine occupant responses.
In summary, it is not sufficient to assume that conforming to published IAQ criteria will
always prevent complaints about IAQ. Environmental measurements are only as useful as
their capacity to predict human responses; therefore, if human responses can be recorded,
it makes sense to use them as direct indicators of IAQ. Whatever the environmental mea-
surements suggest, if the occupants are dissatisfied, there is a problem. The investigator
must determine exactly what is wrong.
Conversely, if the occupants were found to be satisfied with the indoor environment, it
would seem strange to say they ought to be dissatisfied on the evidence of environmental
measurements. There are exceptions to this because there are hazardous agents that the occu-
pants would be unable to perceive, for example radon or carbon monoxide. Hence, to obtain
a comprehensive assessment of IAQ, occupant reactions need to be used alongside environ-
mental measurements and medical diagnosis of illness. Although specific illnesses such as
lung cancer or Legionnaires disease would not be diagnosed by means of occupant surveys,
there are many circumstances in which IAQ problems are best assessed with such surveys.
The Value of Occupant Data
In carrying out occupant surveys to investigate IAQ, the investigator is not limited to count-
ing complaints. Indeed, unsolicited complaints are generally a poor guide to the nature and
magnitude of IAQ problems. Instead, the investigator can make use of the remarkable
capacity of people to act as measuring instruments and data loggers. The nose and mucous
membranes of the eyes and airways are sensitive to airborne chemicals, some at extremely
low concentrations. Consequently, people can detect, describe, and quantify a very wide
range of environmental factors over a huge dynamic range. They can do this for specific
ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.3
factors, locations, and times (e.g., the smell of tobacco smoke in the corridor on Monday
afternoon) or for combinations of factors, averaged over time and space (e.g., the general
air quality in the whole building during the past year). The information can be recorded at
the time of exposure or it can be reported retrospectively, even without prior warning; this
means that people are often the only source of data about how the air quality has been in
the past. Some kinds of information can be acquired only by occupant surveys (e.g., symp-
toms of ill health related to being in a particular building).
In using this rich source of data, there are two key problems: poor calibration and inef-
ficient downloading. Calibration refers to the need to use data from different individuals,
who respond and think differently from each other, to make valid judgments about indoor
air quality and how to improve it. Put simply, two people in the same environment will
rarely have identical reactions to that environment. This difference may reflect inherent
susceptibility, the modifying effects of various environmental factors, and differences in
reporting. There are two main ways to reduce this variability: (1) to train people to give
similar responses and (2) to average the responses from sufficient people to reduce varia-
tion to a level that is acceptable for the purpose of decision making. Of course, sometimes
individual variation is the subject of a study rather than a nuisance to be controlled, as in
the case of research into the distribution of susceptibility and its determinants.
Even if two people have identical reactions to a particular aspect of the environment,
they may not give identical responses in a questionnaire; not only this, but slightly differ-
ent questions on the same issue will elicit different answers. This is the problem of down-
loading data from people (obtaining information that is both valid and reliable). Peoples
breadth of measuring and logging capacity can thus become a limitation of occupant sur-
veys, particularly if the survey process is not carefully managed so that the investigator
knows what the occupants are reporting and according to what criteria. Fortunately, there
are procedures for addressing these problems, as discussed later in this chapter.
The Key Issues Addressed in Occupant Surveys

The three principal occupant reactions to be addressed using occupant IAQ surveys are
acute nonspecific symptoms, environmental discomfort, and the adverse effects of poor
IAQ on worker productivity. Surveys can also assess social and personal factors that can
modify response to IAQ, such as underlying medical conditions, management issues, per-
sonality variables, and sensitivity to air pollutants; however such measurements are outside
the scope of this chapter.
Acute Nonspecific Symptoms (Sick Building Syndrome). In comparison with specific

illnesses, there is less established knowledge about the causes of a range of acute nonspe-
cific symptoms, which some people report when they are in certain buildings. The major-
ity of surveys of occupant reaction to IAQ have been conducted in the context of seeking
to explain such symptoms, even when the questions themselves have been about environ-
mental parameters or modifying factors rather than the symptoms themselves. Hence these
symptoms are discussed at greater length under Sick Building Syndrome, below.
Environmental Comfort. Along with nonspecific symptoms, there are commonly com-
plaints about aspects of the indoor environment itself (e.g., the odor level might be too high
or the humidity might be too high or too low). These are complaints about the environment
rather than about the persons perceptions of his or her own health.
The importance of such perceptions is illustrated by their use in setting ventilation rates.
In contrast to the procedures for setting pollutant limits, ventilation rates in nonindustrial
workplaces have generally been set according to criteria of comfort or acceptability since
the work done by Yaglou et al. (1936).
53.4 ASSESSING IAQ
Productivity. Although productivity or staff efficiency would appear to be a potential

basis for standards, a principal barrier has been defining and measuring productivity. To
use work performance as a criterion is feasible in some settings, for example, where peo-
ple are doing repetitive routine tasks, but in other cases it is much more difficult to assess
whether performance has been improved or reduced by attaining a certain level of IAQ. For
some types of work it may be some years after a piece of research was performed before its
usefulness can be established. In practice there is little readily usable data linking IAQ with
productivity. Nevertheless, productivity is a key element in the motivation to improve IAQ
in the workplace because it is generally assumed that healthy, comfortable staff are also
productive staff.
Sick Building Syndrome
The investigation and study of sick building syndrome (SBS) requires occupant surveys to
obtain population-level data because it cannot be diagnosed through other means. SBS can
be not only a difficult health issue but also a costly problem.
What Is SBS? The concept of SBS has caused confusion since it was introduced. This
section seeks to break through the confusion by offering a usable definition of SBS and
showing how the definition is necessarily linked to the means of diagnosis. It is inherently
difficult to characterize SBS and its causes unless there is an agreed definition that can actu-
ally be used in practice. There has so far been substantial variability among the definitions
offered; in many studies, no definition at all has been given. The definition adopted in this
chapter is as follows:
Sick building syndrome is a phenomenon whereby people experience a range of symptoms

when in specific buildings. The symptoms are irritation of the eyes (e.g., dry/watering eyes),
nose (e.g., runny/blocked nose), throat (e.g., dry/sore throat), and skin (e.g., dryness/redness),
together with headache, lethargy, irritability, and lack of concentration. Although present gen-
erally in the population, these symptoms are more prevalent among the occupants of some
buildings than of others and are reduced in intensity or disappear over time when the afflicted
persons leave the building concerned.
Because the symptoms are associated with particular buildings, they are often called build-
ing related. The time required for recovery can vary from hours to weeks, depending on the
type and severity of the symptom. Thus, to say that SBS is a real phenomenon is merely to
say that there is a variation in symptom prevalence among buildings, not a clear division
into sick and healthy buildings but a continuous variation.
Some studies have used more extensive lists of symptoms, including for example, air-
way infections and coughs, wheezing, nausea, and dizziness (WHO 1982); high blood pres-
sure (Whorton and Larson 1987); and miscarriages (Ferahrian 1984). However, although
these conditions are mentioned as occurring among staff in certain sick buildings, they
probably should not be included in SBS. Taste and odor anomalies are not necessarily
symptoms: they are better considered as environmental perceptions and are therefore best
excluded from the list of defining symptoms.
SBS is thus defined, as many health problems have been in the past, in terms of symp-
toms and conditions of occurrence rather than cause (except at the very general level that
buildings are somehow responsible). The reason is that there is no single proven causal
agent, and any attempt to introduce etiology into the definition is likely to be misleading at
present. This would apply equally to specific causes (e.g., tight buildings) and to com-
mon generalizations such as SBS being diagnosed only when there is no known cause or
where there are multiple causes. It is nevertheless possible to talk of preventive and reme-
dial measures, much as many diseases were to some extent prevented (e.g., by hygiene
practices) and treated by reducing specific symptoms (e.g., fever) long before the cause of
them was identified.
Although SBS can be defined, the definition of a case of SBS (a sick building or SBS-
affected person) is to some extent arbitrary. A theoretical definition of a case could follow
from the definition of SBS, but in practice the identification of specific cases would depend
on what is regarded as an acceptable level of symptoms among the occupants.
Diagnosis. The definition of SBS suggested in this chapter implies that the definition is
inseparable from the means of diagnosis. This is because the range of symptoms reported
in a given building population, and their prevalence rates, will depend on the number and
nature of the questions used to elicit the information. The choice of survey instrument is
thus critical for the diagnosis of SBS.
Because definition is tied closely with diagnosis, if the diagnostic approach is not stan-
dardized, the definition is effectively variable across investigations. For example, if two
different questionnaires are used in studying two different buildings, it may be that two dif-
ferent instruments are being used to measure the same phenomenon. In fact, because the
questionnaire determines what is measured, two nonidentical phenomena are being mea-
sured. Diagnosis can also become inconsistent if different clusters of symptoms are statis-
tically derived because these clusters will vary among buildings or groups of buildings (see
Raw et al. 1996a, 1996b).
Although this discussion may appear somewhat academic, it highlights a key issue
in the current approach to diagnosing SBS. We have imperfect instruments, but we need
to use them rather than to wait for agreement on the perfect diagnostic procedure.
Greater standardization is critical to advancing understanding of SBS. It is important to
keep in mind for this purpose that SBS is a complaint of people, not buildings, and can
be diagnosed only by assessing the building occupants, not by examining the building
itself.
An attempt to define a working criterion for SBS diagnosis (Raw et al. 1990) specifies
a level of more than two symptoms per person, recorded using the same questionnaire as in
the U.K. cross-sectional survey reported by Wilson and Hedge (1987). This was the level
at which respondents, on average, reported a negative effect of the indoor environment on
their productivity.
SBS Matters. It is generally recognized that SBS is not an isolated or occasional phenome-
non. A WHO working group (Akimenko et al. 1986) estimated that, although frequency of
occurrence varies from country to country, up to 30% of new or re-modelled buildings may
have an unusually high rate of complaints (these complaints may extend beyond SBS). This
estimate is again to some extent arbitrary and could be set considerably higher or lower by tak-
ing a different criterion for what would be considered a high rate of complaints.
Apart from effects on productivity when staff are at work, SBS has been shown to
affect absenteeism and quite obviously makes demands on the management and trade
unions that spend time trying to resolve the problem. Other likely effects are on unoffi-
cial time off, reduced overtime, and increased staff turnover. In extreme cases buildings
may be closed for a period. If building users were to associate SBS with energy-saving
measures such as controlling ventilation rates, this could inhibit moves toward greater
energy efficiency.
Although neither life threatening nor necessarily disabling, SBS is clearly perceived to
be important to those affected by it, particularly if they are affected at home (e.g., the
elderly or sick in residential care) and cannot leave the affected building. The economic sig-
nificance of IAQ problems is addressed further in Chapters 4 and 56.
53.6 ASSESSING IAQ
The Causes of SBS. The list of suggested causes for SBS is very long indeed, as
addressed in other chapters of this book. Although this chapter does not review the evi-
dence on the possible causes, it should be noted that, although many studies have focused
on IAQ and ventilation rates, there seem to be some contributions from a wide range of
other factors in the environment (particularly temperature, humidity, cleanliness of offices,
and personal control over the environment). Current evidence suggests that no single fac-
tor can account for SBS: there are probably different combinations of causes in different
buildings.
The statement is sometimes made that we do not know the cause of SBS; this is unhelp-
ful because we know many causes of SBS. The problem is one of identifying the cause or
causes in particular buildings because this entails consideration of interactions occurring at
the following four levels among etiological factors:
The building. The design, construction, and location of a building and its services and fur-
nishings may contribute to IAQ problems in a variety of ways, from the site microclimate
through shell design (i.e., depth of space) to the building services and build out.
The indoor environment. The effects of the building and site will generally be mediated
by characteristics of the indoor environment (e.g., temperature or allergen levels).
The occupants. Households or organizations that occupy and operate buildings may con-
tribute to IAQ complaints, for example, via the quality of building maintenance and
workforce management.
The individual. Reported experience of IAQ problems varies from one person to another
within buildings for a number of reasons, which would include personal control over the
environment, constitutional factors, behavior, and current mental and physical health.
In addition, causes of SBS may stem from the earliest origins of a building, from the orig-
inal concepts, specification, and design for a building through the construction, installation,
and commissioning to the maintenance and operation of the building. Hence it is too sim-
plistic to talk about the causes of SBS only at the level of IAQ parameters causing certain
symptoms. The investigator must remember that the determinants of SBS cannot be
addressed adequately using only an occupant survey.
53.2 DECIDING TO CONDUCT AN OCCUPANT

SURVEY
Introduction
The process of deciding to carry out an occupant survey is important because it should iden-
tify the reason for the survey or hypotheses to be tested, which in turn will be a key factor
in deciding the method to be used. The decision-making process should also bring all the
interested parties together in consensus over the approach. The most common reasons for
initiating a survey would be as follows:
There is a suspected problem with the indoor environment, based on spontaneous com-
plaints, illness reports, sickness absence, or environmental monitoring.
There is a desire to be proactive in monitoring the quality of the indoor environment.
It fits the needs of a research project.
Each of these is addressed in turn.
Response to Suspected Problems
If an IAQ problem is suspected, there is rarely a good reason for not conducting an occu-
pant survey. Neither the act of conducting a survey nor the results found define liability for
the IAQ problem. Rather, the survey is a first step toward finding a solution. In many cases,
the occupants may consider that they have already made a diagnosis, that remedial action
should be taken, and that any further survey is likely to be uninformative. Although such
feelings are understandable, an unsystematic and anecdotal collection of complaints is a
poor guide to what action should be taken. On the other hand, some managers would pre-
fer to believe that there is no problem and that surveys will only cause further complaints.
Unless spontaneous complaints are investigated in a more rigorous manner, such conflicts
in views cannot be resolved. Either the occupants will continue to be affected by SBS, with
implications for their health and the success of the company occupying the building, or
effort may be wasted addressing a problem that never existed.
A good occupant survey should not just confirm (or otherwise) the level of complaints
in the building but should also provide information about where and when there are prob-
lems and what types of complaints are being made. It may also give an indication of the
cause of the problems, but this must always be backed up with further investigations
involving other measurements and/or interventions. A survey will also provide a basis for
evaluating the effectiveness of any remedial measures that might subsequently be under-
taken. This application could include piloting remedial measures in one part of a building
and assessing the outcome before extending the measures to other parts of the building.
This type of application has many of the characteristics of a research project, with all the
methodological rigour entailed.
Proactive Monitoring
It is sometimes claimed that faulty management is responsible for IAQ problems. At one
stereotyped extreme, problems in the workplace can always be attributed to bad man-
agement. At the other extreme, blame is placed purely on the environment. A proper bal-
ance between these views can be struck by establishing in specific terms what
management could have done to have avoided the problems. Broadly speaking, manage-
ment can be seen as contributing to IAQ problems if it does not act effectively to create
a good indoor environment or if it does not establish a good organizational environment
for reducing stressors that foster complaints about IAQ and for dealing with complaints
should they arise.
Carrying out an occupant survey only in response to complaints might be called bad
management, although it is more conciliatory to call it good management, albeit too late.
By analogy, consider a company that did not routinely check the safety of its vehicle fleet
but chose instead to wait until a truck reached the point of swerving off the road before
checking the whole fleet. If we can be proactive with machines, why not with people?
Occupant surveys indicate whether the final clients (i.e., the people who occupy the
building) are satisfied with the architect and building services engineers final product (i.e.,
the indoor environment and the control of that environment). The data provided by such
surveys have two essential purposes:
To tell the facilities manager and other parties whether a building is performing to an
acceptable standard
53.8 ASSESSING IAQ
To tell those who contribute to future buildings about how to create better environments
(whether their contribution is in design, building, installation, commissioning, operation,
maintenance, or management)
In other words, the purpose of carrying out surveys is to improve the product, whether
that product is the environment in a current building or in a future building.
In a busy facilities management situation, it can be tempting to believe that occupant
surveys are not necessary (because occupants will complain if there is a problem) or actu-
ally counterproductive (because they create or exaggerate problems or create an awareness
of problems). A forward-looking company will set aside these concerns in favor of the goal
of achieving greater client satisfaction.
Occupants do complain, but their complaints are often an unreliable indication of the
scale and nature of any problems because complaints are also motivated (positively or neg-
atively) by a range of factors unrelated to the subject of the complaint. Besides, what kind
of service provider can really afford to wait until the situation becomes bad enough to pro-
voke complaints? It is good practice in any industry to identify and deal with complaints
before they break out and are labeled as a problem. Thus, the second objection (i.e.,
that occupant surveys create problems) is unsound. Surveys will identify whether a prob-
lem exists and thereby offer an opportunity to solve the problem.
By using a standard questionnaire, the results of the survey can be compared with a wider
database to show how the building is performing in relation to comparable stock (see Section
53.3). Alternatively, repetition of the survey will show whether there are changes over time
in the performance of a building, which would give an even clearer indication of impending
problems. An interval of 2 years between surveys is a reasonable norm, but the interval could
be shorter or longer depending on the pace of change in the building or workforce.
Research, Including Following Up Mitigation Attempts
Survey research on SBS can be seen to have passed through three phases, representing a
transition from exploratory to confirmatory studies. The approaches represented by each of
the three phases have validity for specific purposes, but each needs to be done well if mean-
ingful results are to be achieved.
The first phase of research was concerned with the existence of the problem. It effec-
tively commenced in the late 1970s, although there were earlier warnings of the emerging
problem of SBS (Black & Milroy 1966). By the early 1980s, it had demonstrated to the sat-
isfaction of most researchers that there was a phenomenon that we now call SBS.
Second, notably in the 1980s, there were many investigations that relied on comparisons
of occupants symptom prevalences between buildings. These studies provided evidence
on what can be termed risk factors (e.g., open plan offices and low perceived control over
the indoor environment). These factors cannot necessarily be regarded as direct causes
because of the many confounding factors and confusion over causal pathways. For exam-
ple, air-conditioning was identified as a risk factor, but the causal factors could include var-
ious building characteristics that are commonly associated with air-conditioned buildings,
such as deep building plans, reliance on artificial lighting, and lack of personal control of
the indoor environment.
Now, in the third phase of SBS research, the risk factors constitute important clues as to
the causes, clues that are being followed up by making experimental changes to buildings.
The basic plan of such intervention studies is first to apply theoretical knowledge and an
examination of a building to generate hypotheses about causes of SBS in the particular
building being studied. Modifications are then made to the building and/or the indoor
environment, and measurements are made to determine whether the modifications have
been successful in reducing the symptoms experienced. This approach allows for control
of potential confounders and thus provides stronger evidence on the causes of SBS.
Identifying the type of research that is to be carried out is the key to choosing the right
research design and the right survey instruments. Research that seeks only to identify sick
buildings requires a relatively simple screening questionnaire. A search for risk factors
requires the collection of additional data on potential determinants and modifying factors.
Intervention studies require yet another layer of sophistication: Their design has recently
been discussed at length (Berglund et al. 1996) and is summarized later in this chapter.
53.3 INSTRUMENTS FOR THE SURVEY
Choosing an Approach
Occupant data can be collected by a number of means, including structured or unstructured

interviews (medical examinations generally include an interview with the patient, and this
interview may be more or less structured), discussion groups, diaries, and self-completion
questionnaires. Medical examinations are outside the scope of this chapter, but they do
have a role to play, especially where complaints have become sufficiently serious for rapid
action to be required or where the nature of the symptoms is unusual.
Although the presence of symptoms of SBS is normally assessed by self-completion
questionnaire, this is for convenience, and most of the symptoms can be assessed by other
means and shown to be correlated with questionnaire responses (see the later discussion of
reliability and validity). Demonstrations that symptoms can be reduced markedly in blind
trials of remedial measures (Raw et al. 1993) also support the validity of the questionnaires
used. The remainder of this section assumes the use of self-completion questionnaires.
Most of the information provided would apply equally to structured interviews or diaries.
Selecting or Designing a Questionnaire
Introduction. For most surveys, the use of an existing questionnaire is preferred, whether
the purpose is proactive monitoring, response to complaints, or a screening survey carried
out in a research project. Designing a new questionnaire is time consuming and difficult, if
done properly. By using a standard questionnaire, the results of the survey can be compared
with a wider database. In some cases, there will be a need to modify an existing question-
naire. In such cases, or where questions are borrowed from existing questionnaires to cre-
ate a new one, pretesting is important because the meaning of questions can be affected by
even small wording changes and by the context provided by neighboring questions in the
questionnaire (Rathouse and Raw 2000).
Various general guidelines on questionnaire design have been produced (e.g., Sudman
and Bradburn 1982, Converse and Presser 1986). The following sections can be used to
gain an understanding of questionnaires for the purpose of selecting, adapting, or design-
ing. Readers should be aware that no single SBS questionnaire has been selected as a gold
standard. For example, the phrasing of questions can be used to bring about large varia-
tions in measured symptom prevalence; such data therefore have little value unless norms
are available for a particular standard questionnaire. In every study, some key common
questions should be used that have proven validity relative to other similar surveys.
53.10 ASSESSING IAQ
Reliability and Validity. One problem in IAQ research is that questionnaires have been
accepted without sufficient evidence of validity or reliability. It is important to character-
ize these properties of any instrument, including questionnaires.
The reliability of a measurement refers to how precisely it measures. The reliability is
usually expressed as a reliability coefficient, which is the proportion of obtained variance
that is due to true variance in the variable being measured. Repeatability is one indicator of
reliability because if an instrument is imprecise, there will be a low correlation between
repeated measurements; this is how reliability was established for the U.K. Office
Environment Survey (OES).
The validity of a measurement refers to how well it measures what it intends to measure.
One type of validity is empirical validity, the degree of association between the measure-
ment and some other observable measurement. For example, the OES was validated in
comparison with medical interviews (Burge et al. 1990), whereas others have used objec-
tive measures such as tear film breakup (Franck & Skov 1991). An alternative type of valid-
ity is construct validity, which means that the measurement should correlate with all other
tests with which theory suggests it should correlate and should not correlate appreciably
with other tests with which theory suggests it should not correlate. The OES symptom
prevalences correlate with environmental discomfort and productivity but much less with
control over the job.
Symptoms are, by their nature, subjectively reported. Hence, symptom reports have an
implicit validity because what the respondent says is important in its own right. This is
helpful only up to a point because if the report has no relation to physiological states, the
investigator could be misled about the nature of the problem. The important point is that
good practice should be used in recording symptom reports.
Recall of the Past. Valid descriptive data are obtained from questionnaires that focus on
the current, the specific, and the real (Turner and Martin 1984). Questions on the past in
general appear to be more difficult than questions on the present, especially if (1) a deci-
sion was made almost without thought in the first place, (2) an event was so trivial that peo-
ple have hardly given it a second thought, (3) questions refer to events that happened long
ago, and (4) recall is required of many separate events. Even important events can fade with
time or require specific cues to bring them into focus.
The following five techniques have been recommended to improve the validity of
reporting on past events (Converse and Presser 1986):
Bounded recall addresses overreporting due to forward telescoping outside the requested
time range (it may be controlled by establishing the baseline in an initial survey).
Narrowing the reference period for survey reporting is a good corrective means.
Averaging refers to questions about typical conditions, which provide more representa-
tive data than single day focused questions.
Landmark events may be referred to instead of specific dates to anchor the timing of other
events. The question could ask about a symptom experienced since Christmas, instead
of during the last month.
Cueing means that cues are provided to help memorizing. The purpose of cues is to stim-
ulate recall by presenting a variety of associations.
Simplicity Is the Rule, Complexity the Exception. There is a need for simplicity, intelli-
gibility, and clarity. It is imperative that common language should be used, questions
should be short, and confusions should be avoided (Sheatsley 1983). If the respondents are
faced with a task they cannot manage or they believe they cannot manage, the responses
have low information value. It is generally easier to answer questions bearing on ones own
experience and behavior (facts) than questions on opinions and attitudes (evaluations). The
latter are assumed to be more open to the respondents own definition than the former.
Ranking scales have a long history in survey research. Alwin and Krosnick (1985)
showed that rankings (rank order is given) do not show the same relationship to predictor
variables as ratings (category scale value is given) even though the same factors were inves-
tigated. Magnitude estimation scales are a third, more complex, technique of responding.
Magnitude scaling of attitudes has been calibrated against numerical estimation and
physical line-length estimation of physical stimuli such as light, sound, and odors
(Berglund et al. 1975, Berglund and Lindvall 1979, Lodge 1981). Although these more
complex techniques have considerable interest and potential usefulness, they have been lit-
tle used in survey research into IAQ (Garriga-Trillo and Bluyssen 1999).
Questionnaires with closed questions are easiest to standardize. A widespread criticism
of closed questions is that they force people to choose among offered alternatives instead
of answering in their own words. Nevertheless, because closed questions give the same
response options, they are more specific than open questions and therefore more apt to
communicate the same frame of reference to all respondents. The typical survey question
incorporates assumptions not only about the nature of what is to be measured but also about
its very existence.
Questionnaire Construction. A number of basic issues have to be decided in construct-

ing a new questionnaire, including the following:
Type of response format (e.g., ranks, ratings, magnitude estimation scales).
Open or closed questions (closed questions are easiest to standardize but they are some-
times criticized for limiting the respondents options).
The effect of the context of other questions, especially neighboring questions, in the
questionnaire.
The overall length and difficulty of the questionnaire (consider the amount of informa-
tion collected per respondent, how useful the information is, and how many sampled peo-
ple will respond at all).
In some cases, other issues will need to be considered; for example, the questionnaire might
need to be completed by children or by adults who have restricted literacy. If the question-
naire is to be translated into other languages, it should be checked to ensure that adequate
translation is possible and that the questions are likely to be culturally acceptable wherever
they are asked.
Investigators should characterize the instruments to be used by the following:
Piloting or otherwise validating the instrument
Knowing the meaning of each measure, score or index
Assessing and correcting for any predictable source of error such as habituation, practice,
response sets, and false responses
This would require reading the literature on an instrument (including questionnaires) and
being trained in its use. Sometime researchers may rely on developing an understanding of
the instrument in the course of the study, but this carries obvious risks.
The following should be considered in the pilot: variation in responses, meaning, task
difficulty, respondent interest and attention, flow and naturalness of the sections, the order
of questions, skip patterns, timing, respondent interest and attention overall, and respon-
dent well being (Converse and Presser 1986).
53.12 ASSESSING IAQ
The Example of the ROES Questionnaire
Introduction. In the United Kingdom, an expert group set up by the Royal Society of
Health has agreed on a standard questionnaire, the Revised Office Environment Survey
questionnaire (ROES) together with instructions for its use and normative data (Raw 1995).
ROES is intended to be used for screening surveys to determine the prevalence of SBS in
a particular building. An account follows of the issues that need to be covered when design-
ing or selecting a questionnaire for indoor environment surveys, with examples taken from
the development of ROES (see Appendix A for sample of the ROES). More detail on ROES
itself can be found in Raw (1995).
Two other major questionnaires, used extensively in the United States by the
Environmental Protection Agency (EPA) and National Institute of Safety and Health
(NIOSH) are discussed at greater length in Chapter 3. Another questionnaire that has been
widely used is the Swedish MM Questionnaire (Andersson et al. 1988).
The Symptoms to Be Included. The starting point for the selection of symptoms to be
included was the list of symptoms in the largest U.K. study of SBS, the Office Environment
Survey, or OES (Burge et al. 1987, Wilson and Hedge 1987). The same list has been used
in many subsequent and previous U.K. studies. Using the same list of symptoms provided
immediate reference to an established database. These symptoms were as follows:
Dry eyes
Itching or watering eyes
Blocked or stuffy nose
Runny nose
Dry throat
Headache
Tiredness or lethargy
Flulike symptoms
Difficulty breathing
Chest tightness
The last three of these are probably not correctly regarded as typical SBS symptoms. Runny
nose is also problematic as a defining symptom because it appears to indicate primarily res-
piratory infections. Itching or watering eyes may also be nonessential because itching
partly duplicates dry eyes, and watering may reflect specific allergic reactions. An analy-
sis carried out in the course of the development of the questionnaire (Burge et al. 1993)
showed that a building symptom index (BSIthe mean number of symptoms reported per
person in a building) based on the remaining five symptoms is almost perfectly correlated
with an index based on all 10 symptoms.
The following five symptomsdry eyes, blocked/stuffy nose, dry throat, headache, and
tiredness/lethargywould therefore be enough to provide an index of SBS. Consideration
was given to including only these five symptoms on the questionnaire, but this approach
was rejected for two reasons. First, the above analysis was based on removing the symp-
toms at the analysis stage; removing them from the questionnaire might have a very differ-
ent effect; some evidence for this is provided by Raw et al. (1996a). Second, a shorter list
of specified symptoms would place a greater load on the final item concerning other
symptoms.
The first point represents a lesser problem regarding flulike symptoms, difficulty
breathing, and chest tightness because these appeared at the end of the list in previous
surveys and are unlikely to be misreported as earlier symptoms on the list. It was therefore
agreed to delete these symptoms.
A number of questionnaires, particularly in Scandinavia (Andersson et al. 1988, Skov
et al. 1989), have included skin symptoms on the list (e.g., dry skin, skin rash, redness of
the skin). For the sake of greater international compatibility it was thought advisable to
introduce a question about skin symptoms at the end of the list.
If particular investigators or researchers added symptoms to the end of the list, it could
still be possible to compare with a database unless the symptoms added were similar to
those on the main list. However, it would be advisable to test this assumption. If questions
were subtracted, or if the symptom descriptions were changed, comparison with the data-
base would be invalidated.
The layout of the symptom questions was changed from that used in the OES (Burge et
al. 1987, Wilson and Hedge 1987) because the opening question (in the past 12 months
have you had more than two episodes of any of the following symptoms) was considered
ambiguous and, with the questions now covering more than one page, likely to be forgot-
ten by the respondent. The question is therefore now asked separately for each symptom.
This creates some monotony but at least the question is clear.
Recall Period. The response of any one individual to IAQ will vary over time. This vari-
ation might occur over minutes (e.g., because of adaptation to odor or changes in the inter-
pretation of perceptions), hours (e.g., delayed reactions of sensory irritation), or years (e.g.,
as awareness and understanding of IAQ issues develops). Some psychological variables
will be continuously varying over time, whereas other variables will be present or absent
or will be discrete events. The selected time period of a study will censor the data by
design: If the study period is made longer, for example, there is a greater possibility of
symptom occurrence.
The reference period for reporting symptoms in the ROES questionnaire is 12 months,
as in most U.K. questionnaires, based on two or more occurrences over the period. Seasonal
variation of the symptoms can be addressed separately if it is suspected that this was likely
to occur in a particular building.
Recall over a 12-month period is unlikely to be reliable in absolute terms: it is likely to
represent mainly the previous few weeks, possibly moderated by recall of particularly
severe symptoms prior to this or any marked seasonal variation. It is necessary to empha-
size here that the questionnaires main function is to make comparisons among buildings
or over time for a particular building. Thus it is not appropriate to attempt to assign absolute
meanings to the questionnaire responses. The use of relative ratings largely circumvents the
problem of recall because it is the same for each building and each occasion.
The test-retest reliability of the symptom questions is good (Wilson and Hedge 1987),
but they should not be repeated within too short a period because this tends to create a
decline in the number of symptoms reported (Raw et al. 1993). The critical interval for this
is not known, but an interval of a year is probably adequate to prevent it (there would nor-
mally be no reason to conduct screening surveys at shorter intervals).
If the requirement is to assess a building in relation to the whole year, not just the time
of the survey, the only alternative to using 12-month recall would be to repeat the ques-
tionnaire during the course of a year, perhaps four times with 3-month recall on each occa-
sion. A requirement to carry out a survey more frequently would increase costs and would
discourage use of the questionnaire. Frequent repetition would also affect interpretation of
the symptom scores, as noted above. It would therefore be necessary to collect data from
repeat surveys in a sample of buildings to generate a new database for comparison.
Questionnaires are available to cover shorter recall periods, where this is required. The
Swedish MM questionnaire (Andersson et al. 1988) has been translated into English and a
number of other languages. It uses a 3-month recall period. The U.S. EPA (BASE)
53.14 ASSESSING IAQ
questionnaire uses a 4-week period, as does the related NIOSH questionnaire. The ROES
was adapted for 1-month recall and for reporting current conditions, as part of the European
IAQ audit project (Groes et al. 1995). These questionnaires would generally be suitable for
intervention studies in which changes over a period of much less than a year normally need
to be detected.
Building Relatedness. The ROES questionnaire seeks to establish whether symptoms are
related to being in the target building by asking the question Was this better on days away
from the office? From the point of view of comparison with the OES data, it is preferable
to maintain this approach. However, the question is not specific about the comparison to be
made and demands interpretation on the part of the respondent. For example, does it mean
on whole days away from the office, away from the office at home or away from the office
in other buildings or outdoors or on holiday. For building-level comparison, it should be
valid, assuming that people will, on average, adopt the same kind of interpretation. It also
appears, in fact, that the phrasing of the building-relatedness question has little effect on
symptom reports (Raw et al. 1996a).
Frequency of Occurrence. For screening purposes, it is not necessary to include ratings

of the frequency of the symptoms, although such scales can be useful. A frequency scale
has been included, placed after the assessment of building relatedness to maintain compat-
ibility with the U.K. database. The scale meets the dual requirements of (1) covering the
complete scale of frequency without (2) having overlap between categories. Including such
a scale gives the potential for greater sensitivity in comparing symptom prevalence
between buildings or over time.
Layout. The layout of the questions (i.e., whether it is question by question or with the
responses given in a grid) may be significant. Where responses are given in a grid, Raw et
al. (1996a) report that respondents have a greater tendency simply to pick the symptoms
that particularly apply but not give a response to the others at all (i.e., there are more miss-
ing responses). They also found that there is less variance in response within respondents,
as though there is a tendency to stay within or close to a particular column (frequency cat-
egory) of the grid. Therefore, although a grid would require less space (probably one page
rather than two), separate questions are preferred.
Questions about the Environment. Ratings of the environment can assist in the identifi-
cation of causes but cannot be always taken at face value; their primary purpose should be to
indicate what aspects of the environment give rise to most concern and therefore which
aspects offer the best chance for improvement. A very large number of ratings could be
included in a questionnaire, depending on the level of detail with which environmental fac-
tors need to be specified. For a screening survey, only the main likely problem areas should
be evaluated, normally with separate ratings for summer and winter. The key ratings would
generally be temperature, humidity, air movement, and air quality. Lighting, noise, and vibra-
tion are probably less important in most cases of SBS, and therefore ROES has only a single
question on each of them for each season. More detailed questions, or a follow-up question-
naire, could be added in specific surveys if the investigator wished to do so. As with symp-
toms, the impact of doing this should be assessed before comparing with the database.
A rating of office cleanliness has also been added. Some studies have used a rating of
office cleaning, but it was felt that this could be ambiguous, including for example the
extent to which cleaners interrupt work or feelings about the use of chemicals that may
damage the indoor or global environment. The previous OES question on personal control
over heating, ventilation, and lighting has been rephrased to remove any ambiguity con-
cerning which part of the office the control refers to. The question about exposure to
tobacco smoke has also been made less ambiguous.
If the environment questions are not required in a particular survey, they can be omitted by
taking out the central pages of the questionnaire. This has subtle effects on symptom reporting,
depending on the gender of the respondent and the overall quality of the indoor environment,
but it does not affect the overall symptom prevalence (Rathouse and Raw 2000).
Questions on Confounding Factors. Two kinds of confounding factors can be included

in an SBS questionnaire: variables that permit adjustment of the building symptom score
and variables that may provide insight into the causes of problems in the building (and
modifying factors and confounders). The building symptom index (BSI) can be corrected
for gender, job type, and visual display unit use (the latter is a relatively minor modifying
factor). If some staff are part time or spend time in different parts of the building, this also
may need to be taken into account.
In a screening questionnaire, there is only a limited role for seeking to identify the cause of
SBS: The existence of the problem should be determined before its causes. However, a limited
number of questions were included about the office environment as discussed above, plus
questions on speed and effectiveness of the management in dealing with indoor environment
problems, privacy, office layout, and decor. Ratings of overall working conditions, productiv-
ity, personal medical history, alcohol consumption, and work breaks are not included because
they are not likely to add significantly to data at the building level. Questions about the job and
quality of management are also not included because these are not primary issues in screening
(they may be relevant in certain research studies) and may inhibit some managers from agree-
ing to the survey and some staff from returning the questionnaire.
Consideration was given to including some kind of check for honest and consistent
reporting, to improve the validity of responses. This could significantly increase the
length of the questionnaire, and the usefulness of a lie scale is likely to reduce over time
as people become aware of its existence. The required comparisons are in any case prob-
ably valid without this kind of check because the database used for comparison would
have any tendency to misreport built into it. The presence of unusual patterns of
response to the questionnaire could in principle be used as a form of lie scale, but this
idea has not been developed.
Adding Questions. It is always tempting to collect too much information, much of which
will never be subjected to any useful analysis. The ROES questionnaire is designed with
this in mind: It is a basic screening questionnaire to determine whether there are problems
with occupant health and comfort, not a method of showing what is causing the problems.
If there is any intention, in a particular survey, to add to this questionnaire, it can be useful
to ask the following questions:
Is it possible or necessary to carry out a statistical analysis of the information to be
gathered?
How much, approximately, should the study cost, how long should it take to complete,
and what uncertainty can be tolerated in the results?
What are the motives and purposes of the study and for gathering particular items of
information, and would they be credible to the respondents?
53.4 PROCEDURES FOR THE SURVEY
Introduction
A questionnaire is not in itself a method; it is an instrument, which will produce valid

results if used in accordance with the manufacturers instructions. There is very little
53.16 ASSESSING IAQ
value in using a standard questionnaire without following the data collection method rec-
ommended for that questionnaire. A questionnaire study is an important part of monitoring
the health of people in the workplace, and it is worth a little effort to organize it well. Much
effort can be wasted through simple oversights, but a few basic principles will make the
task more manageable. The following guidance is based on that given with the ROES ques-
tionnaire. If available, the guidance provided with whichever questionnaire is used should
be followed.
Planning the Survey
The survey should be carried out by an organization that can guarantee (to the satisfaction
of the staff) that the survey is confidential and that information on individuals will not reach
management or other staff in the building without the consent of the individuals concerned.
Eligible organizations could be, for example, a body that is independent of the building
management or, if one exists, the occupational health department of the organization occu-
pying the building.
It is of value to plan with all parties concerned with the study and to do the following:
Hold preliminary discussion between the organization that will carry out the survey,
management, unions, and other representatives of the building occupants, safety officer,
maintenance staff, and so forth (the survey should nonetheless be seen by staff to be inde-
pendent of management and unions).
Ensure that staff know the survey is approved by management and other parties as appro-
priate and can therefore be regarded as part of their work.
Establish agreements about confidentiality and lines of communication between all par-
ties.
Agree to inform building management immediately about any health risks that are dis-
covered.
Confidentiality is particularly important, not only from the point of view of motivating the
respondents but also for evaluation of the results.
If the survey is being conducted as part of a research project, an early stage of the work
will often be to select a building. This selection will depend on the purpose of the project
and cannot be covered in detail here. However, in the case of an intervention study, the fol-
lowing should be considered when selecting a building:
A single large building (with many rooms and people) allows better specification of
experimental/control conditions than several small buildings, unless the small buildings
are all very similar in design, operation, occupancy, and managementin either case the
objective is to reduce confounders by making the experimental and control groups as
similar as possible.
The initial level of SBS symptoms should be high in the building, to be able to demon-
strate an improvementone way of identifying such a building is by examining the level
of spontaneous complaints from building users, but the actual level of symptoms should
be confirmed by a structured survey.
There needs to be a high level of cooperation from all parties concerned with the build-
ing, especially in relation to carrying out the intervention.
The management of the study will probably be simplified if the organization that occu-
pies the building is also responsible for its maintenance.
There is of course the additional question of when the researcher should embark on a
major study: Should obvious problems with the building be put right first and should the
first questionnaire survey be carried out before any such remedial measures? The answer
will depend largely on the researchers resources and research objectives. It may be of
interest to assess whether any obvious problems in a building are actually responsible for
occupant complaints. In such cases it will be necessary to carry out a first stage of moni-
toring before carrying out any remedial measures. If, however the requirement is to iden-
tify the causes of problems once the building appears to be operating within normally
accepted conditions, the first stage of the monitoring could be delayed. If it is delayed, it
should be well after the first remedial measures have been completedat least as long as
the recall period of the questionnaire.
Survey Sampling
The survey sample design should be developed as an integral part of the overall study
design. Survey sampling is a highly specialized and developed component of the survey
process. Therefore the wisest decision for a researcher with limited sampling knowledge is
to consult an experienced survey statistician, particularly in relation to the size of the sam-
ple to be used. Some general guidance is given here.
The total number of people who could, in principle, complete the questionnaire may be
referred to as the target population. This might be, for example, all the staff in a particular
building or in certain parts of the building. The advantage of defining the target population
is that the exclusion of any subgroups is explicit and the restrictions of the survey will be
known.
When practical constraints are considered (for example, the target population may be
very large), the target population is often replaced by a survey population, or sample. About
100 workers need to be included in a sample to produce reasonably reliable results (Raw et
al. 1996b). If the target population is larger than this, a sample can be used of approximately
100 workers. The sample size should be increased if different areas of the same building
are to be compared (e.g., 100 from an area where complaints have been made and 100 from
a comparison area). If fewer than 100 workers are available, a survey can still be conducted,
but as the number of workers is reduced there is a progressive decrease in the reliability of
the results and an increase in risk of bias due to variation in individual sensitivity among
occupants. Unless the target population is very large, it is often easier (logistically and
politically) to include every person present during the survey than to go through the process
of selecting a sample and then finding the selected persons.
The most basic sampling procedure is simple random sampling, which requires that
each person or workstation has an equal probability of being included in the sample.
Strictly, this means that a list of people or a plan of workstations should be available. For
example, people might be chosen randomly from a staff list. In this way, selection biases
are avoided. A reasonable approximation to random sampling can generally be achieved by
selecting from a plan of workstations. However, if sampling is based on workstations rather
than persons, rigorous care is needed to follow a plan and not to select only the workers
who are present at the time of the first visit to the workstation. With any sampling strategy,
a complete sample will only be obtained if those who are unavailable through absence or
for other reasons are contacted at a later date.
Several common practical sampling designs are modifications of simple random sam-
pling (Kalton 1983, Lee et al. 1989), as follows:
1. With systematic sampling, each nth element is selected after a random start in a list or a
chosen route around the building.
53.18 ASSESSING IAQ
2. Stratified random sampling classifies population elements (e.g., people) into strata (e.g.,
departments, job grades), and random sampling is then carried out separately from each
stratum. This can be complex but may be useful to ensure an adequate statistical sam-
ple of any small population subgroups. Any mean values calculated for the building as
a whole would then have to be corrected to take account of the overrepresentation of
particular groups.
3. Multistage cluster sampling can be used when the population is very large, for example,
an estate of many buildings. Clusters of elements are selected randomly in one or more
stages (e.g., 5 out of 20 buildings, then half the floors on each selected building), and
then at the final stage individuals are randomly sampled.
4. Probability proportional to size sampling would select, for example, a sample of
rooms weighted by number of individuals in each room (i.e., the more people in the
room, the more likely it is to be selected).
In general, samples should be balanced for workers near windows and near the center and
on different faces and floors of the building.
The above approaches are all examples of probability sampling. Nonprobability sam-
pling covers a variety of procedures, including the use of volunteers and other bases for
choice of elements for the sample with the purpose that they are representative of the
population (Kalton 1983). Of course, the weakness of all nonprobability sampling is its
subjectivity. A sample of volunteers or a representative sample chosen by an expert can be
assessed only by subjective evaluation, not by assumption free statistical methods.
Motivating the Respondents
Questionnaire studies can be perceived by busy respondents as wasting time, and they do
not always understand the purpose of the survey. The loss of respondents from the sample
is therefore a risk if proper care has not been taken, and this can have two consequences.
First, there may be insufficient responses for satisfactory statistical analysis (particularly if
there is a small sample to start with). Second, there may be biased sampling. In particular,
people with more complaints may become overrepresented in the sample.
To achieve a reasonably representative sample, response rates of over 80 percent are needed
from either whole building populations or from occupants randomly selected from a popula-
tion. In practice it should be possible to achieve over 90 percent for a single survey (this can be
difficult to maintain if repeated surveys are conducted of the same population at short intervals).
The following paragraphs provide recommendations for recruiting and retaining a sample.
First, consider the demands to be made on the respondent, for example the length and com-
plexity of the questionnaire, and the number of occasions on which it is to be completed. If the
number of respondents is sufficiently small, a meeting could be held in advance of the survey;
otherwise hold a meeting with representatives (possibly trade unions) and/or send a letter to the
respondents in advance of the survey and/or in a letter accompanying the questionnaire when
it is distributed. This exercise should seek to do the following:
Convey the value of the study (this ought to be straightforward if the study may lead to
remedial measures to improve the indoor environment).
Explain the need to have the participation of everyone who has been selected.
Make clear who is carrying out the study (e.g., independent researcher who is neutral to
any conflicts within the building is likely to have an advantage).
Make the information collected completely confidential and inform the respondent of
this.
Give a contact point for queries.

Fix dates for feedback to respondents, especially the end of the study.
Direct social interaction can also be valuable by showing appreciation, understanding

that they are tired and busy, and showing a presence. A few minutes dealing with a simple
misunderstanding or objection can prevent nonresponse. Of course it is critical that the
investigators do not actively influence the answers that respondents give in questionnaires:
Social interaction should be kept at a moderate and professional level. This interaction
should be achieved by delivering the questionnaire personally to each selected person and
collecting the questionnaire a short time later. The questionnaire should be collected the
same day if possible in case the respondent is absent the following day. On collection, the
questionnaire should be briefly checked for any obvious errors and for completeness.
Errors can then be corrected at this time, or the respondent can be encouraged to complete
the whole questionnaire. It can be helpful to monitor nonresponses and, where possible, to
understand the reasons for them. This may make it possible to reinstate a respondent or to
avoid the nonresponse of others. Analysis of nonresponse is not necessary if the target of
80 percent is achieved.
Analysis
Definition of Outcome. Usually, ratings of environmental conditions are considered as

independent scales. Symptoms of SBS are also sometimes treated as independent but often
as all relating to a common phenomenon, which can be represented by an index or score
based on all the symptoms. An intermediate approach is to use several indexes, based on
hypothesized mechanisms or anatomic location of the symptoms. For example, Jaakkola
(1986) calculated a score consisting of six components: skin, eye, nasal, pharyngeal symp-
toms, headache, and lethargy. A presence of one or more symptoms of each component
during the past 7 days added 1 to that score (range 06).
The intensity or severity of the symptoms has rarely been considered in detail; however,
see Jaakkola et al. (1991), Reinikainen et al. (1991), Berglund et al. (1990a, 1990b), and
Lundin (1991) for studies of the frequency of symptoms over a longer time period.
Index or Score to Describe the Total Phenomenon. The principal measure to be

obtained from the ROES questionnaire will normally be the sum of the building-related
symptoms reported by each person, giving the person symptom index (PSI). The mean PSI
of a random sample of building occupants is the BSI. The BSI can be used in one of two
ways: for comparison with the OES database or for comparing over time using repeated
surveys of the same building.
Eight symptoms are listed on the questionnaire; seven are listed in Table 53.1, and a
final symptom (dry, itching, or irritated skin) has been added to the questionnaire since the
OES, and comparison data are therefore not available. As the use of the questionnaire pro-
gresses, comparison data will be published. Of the remaining symptoms, two (itchy or
watery eyes and runny nose) have been included to maintain comparability with the OES
data and to avoid too many symptoms being entered under other. They are probably less
relevant to SBS than the remaining five symptoms.
If the results from different buildings are being compared with each other or with the
OES figures, it is recommended to use a BSI based on five core symptoms: dry eyes,
blocked or stuffy nose, dry throat, headache, and lethargy or tiredness (Burge et al. 1993).
Each symptom that a respondent experienced on at least two occasions in 12 months, and
that was better on days away from the office, scores 1. Where a respondent has not marked
either yes or no to the question about whether a symptom has been experienced, but has
53.20 ASSESSING IAQ
TABLE 53.1 Percentage of Respondents

Reporting the Occurrence of Each Symptom
Symptom % Reporting
Lethargy 57
Blocked or stuffy nose 47
Dry throat 46
Headache 43
Itchy or watery eyes 28
Dry eyes 27
Runny nose 23
indicated that the symptom is better when away from the office, this can be counted as a
building-related symptom and scored 1. All other responses score 0. The PSI5 and BSI5 will
therefore range from a minimum of 0 to a maximum of 5.
Of the 46 buildings in the OES, the best 8 (all naturally ventilated) had a BSI5 of less
than 1.5 symptoms. This level can be regarded as indicative of minimal problems with SBS.
The worst 13 buildings had a BSI5 of over 2.5 (maximum 3.4), and of these, 11 were air-
conditioned. This can be regarded as an action level, above which steps should be taken to
reduce the BSI. Between 1.5 and 2.5 there is a case for taking action, but the levels are more
open to interpretation, depending on the frequency of the symptoms, other health and safety
problems in the workplace, and the degree of commitment to health in the workplace. The
percentage of respondents reporting each symptom is shown in Table 53.1.
A case can be made for correcting the BSI5 for gender and job category; this produces
a basis for comparing buildings while reducing any bias that might be due to the particular
people who happen to be occupying the building at the time of the survey. This is most eas-
ily done by applying weightings to the individual scores. Dividing a PSI5 by the appropri-
ate weighting will standardize the score to that which would be expected of a male
manager. Care should be taken in interpreting corrected scores because part of the variance
attributed to gender and job type may in fact be a result of nonrandom allocation of staff to
working locations. For example, people in lower-paid, more routine jobs might have lower-
quality accommodation and less power to get conditions changed. Uncorrected scores rep-
resent the building as it is, with its current occupants. Its meaning is therefore transparent,
and it will normally be sufficient for most purposes. However, the same building could give
a different score if occupied by a different population.
The BSI can be based on all eight symptoms on the questionnaire if comparisons within
a particular building are being made. This may arise for example if the questionnaire is
repeated at intervals as a monitoring procedure to determine whether good environmental
conditions are being maintained. A different sample of respondents can, in such cases, be
used on different occasions so long as the sampling procedure is the same. In such cases
there is little advantage to be gained from adjusting scores for gender and job type. The
results should be stable over time unless the environment has changed, given an interval of
a year or more. Short intervals between surveys will reduce the BSI if the same respondents
are used in each survey.
Interpretation of Environmental Ratings. There is no absolute interpretation of the rat-

ings of environmental comfort. Each individual rating should be taken as what it is
claimed to be: a subjective rating. This means that there are three major limitations on
the interpretation of the ratings. A poor rating means that something is wrong with the
environment, but
The obvious interpretation is not necessarily the correct one (e.g., ratings of dry air can
mean that the air is dusty or polluted with organic vapors, ratings of stuffiness can mean
that it is too warm, and reports of offices that are too warm can be due to low air move-
ment rather than air temperatures in excess of recommended levels).
The suggested failing in the environment may well be present, but it is not necessarily
related to SBS in the building.
Symptoms could cause adverse perceptions of the indoor environment, rather than vice
versa.
As a guide, some figures from the OES are given in Table 53.2. The mean is not necessar-
ily the optimum, but it does give an indication of what can reasonably be expected.
Most of the scales are unipolar: One extreme is good and the other bad. In these
cases, any score higher than the mean should be investigated further, and any figure more
than one standard deviation above the mean should be a cause for concern. Three scales
(temperature, air movement, humidity) are bipolar: Neither end of the scale is ideal, and a
deviation above or below the mean of more than one standard deviation represents a cause
for concern. In all cases, interpretation of the environmental ratings should be comple-
mented by local knowledge of the conditions in the building and/or by objective monitor-
ing of the indoor environment.
Ethical Considerations
Environmental change intended as treatment of subjects who have SBS may cause ethical
problems for the researcher. In the laboratory experiment, the researcher may avoid the
TABLE 53.2 Means and Ranges of Environmental

Ratings from the OES
Rating Mean S.D.
Winter
Comfort 3.43 1.79

Temperature 3.65 1.58
Ventilation 3.24 1.73
Air quality 3.10 1.44
Humidity 2.76 1.36
Satisfaction 3.97 1.89
Summer
Comfort 3.28 2.08

Air quality 2.72 1.63
Humidity 2.62 1.66
Satisfaction 3.60 2.16
Control of

Lighting 3.31 2.39
53.22 ASSESSING IAQ
ethical problems by exposing only voluntary subjects to known concentrations of specific

pollutants for controlled periods. In field settings, the building occupants have to be fully
informed about their participation in an experiment and about the possible consequences of
the environmental change. Because the occupants best interests have to be met by the sci-
entific manipulation, what the researcher may accomplish will be restricted in field
research. For ethical reasons the researcher should be able to reasonably well assure that
the occupants are provided the best treatment by the planned environmental change.
Rothman (1986) lists a number of constraints that have to be considered for ethical reasons.
One obvious constraint is that exposures assigned to occupants should be limited to potential
preventives of disease or disease consequences, thus including SBS. Another constraint is that
the exposure alternatives should be equally acceptable under present knowledge. A third con-
straint is that by being admitted to the study, occupants should not be deprived of some prefer-
able form of treatment or preventive measure that is not included in the study. For example, it
is unethical to include a placebo therapy measure (e.g., an unconnected ventilation inlet) in cir-
cumstances for which there is an accepted remedy or preventive measure.
Intervention Studies
The power of intervention studies has recently become more widely realized, but many
attempts at this kind of research have been subject to methodological problems. The design
of intervention studies has been discussed at length in a recent report (Berglund et al. 1996),
which makes recommendations on minimum requirements for the study design, measurement
procedures, assessment of outcomes and determinants, and data analysis.
53.5 CONCLUSION
Making sense of IAQ problems depends not on any single research finding but on putting
together the right conceptual framework and using it in research that has been well designed
and implemented. The following are necessary interrelationships among three important
issues regarding SBS:
The definition of SBS
Diagnosis of SBS in specific buildings
Establishing and comparing the prevalence of SBS in different buildings and contexts
This chapter has set out a definition that makes diagnosis possible and a diagnostic
method that produces consistent and useful results. The method comprises both a ques-
tionnaire and a procedure for using the questionnaire; both are essential. The method is not
unique, and an indeterminate number of other approaches might be taken. It is better to
choose a single approach even if it is for no other reasons than that this particular approach
has been tried and has produced a database of comparison figures. Against this conceptual
framework, the benefits and methods of intervention studies have been described.
Future research and problem solving will need to be directed in an integrated and mul-
tidisciplinary manner to all stages in the life of the building and will need to cover the build-
ing itself (and its location), the indoor environment, the organizations that occupy
buildings, and the needs of individual workers. There are many possible causes of com-
plaints about IAQ, and they are interrelated and interactive, creating multifactorial
problems that demand a multidisciplinary approach: a comprehensive view and systematic
checking of possible problems, not a standard solution applied to all buildings.
APPENDIX: REVISED OFFICE ENVIRONMENT

SURVEY 1
1
This questionnaire remains the copyright of Building Research Establishment Ltd of Garston, Watford,
Hertfordshire, WD2 7JR, United Kingdom, and is reproduced by permission. The right to use the questionnaire can
be acquired by purchasing the questionnaire and guidance for its use, as provided in the following publication: Raw
GJ ed. 1995. A questionnaire for studies of sick building syndrome. BRE Report Construction Research
Communications, London. Data collected using this questionnaire can be compared with the benchmark values pro-
vided in the publication only if the prescribed method of use is followed.
53.24 ASSESSING IAQ


53.26 ASSESSING IAQ


53.28 ASSESSING IAQ

REFERENCES
Akimenko, V. V., I. Anderson, M. D. Lebowitz, and T. Lindvall. 1986. The sick building syndrome.
In B. Berglund, U. Berglund, T. Lindvall, and J. Sundell (eds.), Indoor Air 6:8797. Stockholm:
Swedish Council for Building Research.
Alwin, D. F., amd J. Krosnick. 1985. The measurement of value in surveys: A comparison of ratings
and rankings. Public Opinions Quarterly 49:535552.
Andersson, K., I. Fagerlund, L. Bodin, and B. Ydreborg. 1988. Questionnaire as an instrument when
evaluating indoor climate. In Berglund, B., and T. Lindvall, (eds.), Healthy Buildings 88 vol. 3:
Systems, Materials and Policies for Healthier Indoor Air. D21:139145. Stockholm: Swedish
Council for Building Research.
Berglund, B., U. Berglund, and T. Lindvall. 1975. Scaling of annoyance in epidemiological studies. In
Proceedings from the CEC-WHO-EPA International Symposium on Recent Advances in the
Assessment of the Health Effect of Environment Pollution. Luxembourg: CEC, vol. I: 119137.
Berglund, B., J. K. Jaakkola, G. J. Raw, and O. Valbjrn, 1996. Sick Building Syndrome: The Design
of Intervention Studies. Conseil International du Btiment Report Publication 199. Rotterdam: CIB.
Berglund, B., I. Johansson, T. Lindvall, and L. Lundin. 1990a. A longitudinal study of perceived air
quality and comfort in a sick library building. In Proceedings from Indoor Air 90. Ottawa
International Conference on Indoor Air Quality and Climate, vol. 1, pp. 489494.
Berglund, B., I. Johansson, T. Lindvall, and L. Lundin, 1990b. A longitudinal study of airborne chem-
ical compounds in a sick library building. In Proceedings from Indoor Air 90. Ottawa International
Conference on Indoor Air Quality and Climate, vol. 2, pp. 677682.
Berglund, B., and T. Lindvall, 1979. Olfactory evaluation of indoor air quality. In Fanger, P. O., and
Valbjrn, O. (eds.), Indoor Climate Effects on Human Comfort, Performance, and Health in
Residential, Commercial, and Light-Industry Buildings. Copenhagen, Danish Building Research
Institute, pp. 141157.
Black, F. W., and E. A. Milroy. 1966. Experience of air-conditioning in offices. J. Institute of Heating
and Ventilating Engineers 34:188196.
Burge, P. S., A. Hedge, S. Wilson, J. H. Bass, and A. Robertson. 1987. Sick building syndrome: A
study of 4373 office workers. Ann. Occup. Hyg. 31:493504.
Burge, P. S., A. S. Robertson, and A. Hedge. 1990. Validation of self-administered questionnaire in
the diagnosis of sick building syndrome. In Proceedings of Indoor Air 90, vol. 1, pp. 575580.
Burge, P. S., A. S. Robertson, and A. Hedge. 1993. The development of a questionnaire suitable for
the surveillance of office buildings to assess the building symptom index a measure of the sick build-
ing syndrome. In Proceedings of Indoor Air 93, vol. 1, pp. 731737. Helsinki: Indoor Air 93.
Converse, J. M., and S. Presser. 1986. Survey Questions. Handcrafting the Standardized
Questionnaire. Beverly Hills, Calif.: Sage.
Ferahrian, R. H. 1984. Indoor air pollutionsome Canadian experiences. In Proceedings of Indoor Air
93, Berglund, B., Lindvall, T., and Sundell, J. (eds.), vol. 1, pp. 207212. Stockholm: Swedish
Council for Building Research.
Franck, C., and P. Skov. 1991. Evaluation of two different questionnaires used for diagnosing ocular
manifestations in the sick building syndrome on the basis of an objective index. Indoor Air 1:511.
Garriga-Trillo, A., and P. Bluyssen. 1999. Predicting acceptability and panel selection with magnitude
estimates. In Proceedings of Indoor Air 99, Raw, Aizlewood & Warren (eds.), vol. 2, pp. 5735.
London: CRC.
Groes, I., G. J. Raw, and P. Bluyssen. 1995. Symptoms and environmental perceptions for occupants
in European office buildings. In Proceedings of Healthy Buildings 95, Maroni, M. (ed.), vol. 3, pp.
12931298. Milan: Healthy Buildings 95.
Jaakkola, J. J. K. 1986. Indoor air in office building and human health. Experimental and epidemiologic
study of the effects of mechanical ventilation. (In Finnish with an English summary). Doctoral Thesis.
University of Helsinki, Health Services Research by the National Board of Health in Finland Nr 41.
Jaakkola, J. J. K., O. P. Heinonen, and O. Seppnen. 1991. Mechanical ventilation in office buildings
and the sick building syndrome. An experimental and epidemiologic study. Indoor Air 91 1:111121.
53.30 ASSESSING IAQ
Kalton, G. 1983. Introduction to Survey Sampling. London: Sage.

Last, J. M. (ed.). 1983. A Dictionary of Epidemiology. Oxford: Oxford University Press.
Lee, E., R. N. Forthofer, and R. J. Lorimor. 1989. Analysing Complex Survey Data. London: Sage.
Lodge, M. 1981. Magnitude Scaling: Quantitative Measurement of Opinions. Beverly Hills, Calif.:
Sage.
Lundin, L. 1991. HMSO. 1992. On Building-Related Causes to the Sick Building Syndrome. Acta
Universitas Stockholmiensis, Stockholm Studies in Psychology. Stockholm: Almqvist & Wiksell.
Rathouse, K., and G. J. Raw. 2000. Context effects in a sick building syndrome questionnaire. In
Proceedings of Healthy Buildings 2000, (Helsinki, August 2000), vol. 1, pp. 615620.
Raw, G. J., (ed.) 1995. A questionnaire for studies of sick building syndrome. BRE Report.
Construction Research Communications, London.
Raw, G. J., C. E. Aizlewood, P. S. Burge, C. Whitehead, A. Robertson, C. Kelly, and P. Leinster.
1996b. A questionnaire for studies of sick building syndrome. CIBSE/ASHRAE Joint National
Conference 1996 Part Two, vol. 2, pp. 139144.
Raw, G. J., M. S. Roys, and A. Leaman, A. 1990. Further findings from the office environment sur-
vey: productivity. In Proceedings of the 5th International Conference on Indoor Air Quality and
Climate, (Toronto, Canada), vol. 1, pp. 231236.
Raw, G. J., M. S. Roys, and C. Whitehead. 1993. Sick building syndrome: cleanliness is next to health-
iness. Indoor Air 3:237245.
Raw, G. J., M. S. Roys, C. Whitehead, and D. Tong. 1996a. Questionnaire design for sick building syn-
drome: an empirical comparison of options. Environment International 22:6172.
Reinikainen, L. M., J. J. K. Jaakkola, and O. P. Heinonen, 1991. The effect of air humidification on
different symptoms in office workersan epidemiologic study. Environment International
17:243250.
Rothman, K. J. 1986. Modern Epidemiology. Boston: Little Brown.
Sheatsley, P. B. 1983. Questionnaire construction and item writing. In Handbook of Survey Research,
Rossi, P. H., Wright, J. D., & Anderson, A. B. (eds.). New York: Academic Press.
Skov, P., O. Valbjrn, B. V. Pedersen, and the Danish Indoor Climate Study Group. 1989. Influence
of personal characteristics, job-related factors and psychosocial factors on the sick building syn-
drome. Scandinavian Journal of Work and Environmental Health 15:286292.
Sudman, S., and N. M. Bradburn. 1982. Asking Questions: A Practical Guide to Questionnaire Design.
San Francisco: Jossey-Bass.
Turner, C. F., and E. Martin. (eds.). 1984. Surveying Subjective Phenomena (2 volumes). New York:
Sage.
WHO (World Health Organization). 1982. Indoor Air Pollutants: Exposure and Health Effects: report
on a WHO meeting in Norlingen. Copenhagen: WHO Regional Office for Europe.
Whorton, M. D., and S. R. Larson. 1987. Investigation and work up of tight building syndrome. J.
Occupational Medicine 29:142147.
Wilson, S., and A. Hedge. 1987. The Office Environment Survey. London: Building Use Studies.
Yaglou, C. P., E. C. Riley, and D. I. Coggins. 1936. Ventilation requirements. ASHVE Transactions
42:133162.

Assessing Occupant Reaction To Indoor Air Quality: Gary J. Raw, D.Phil

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Assessing Occupant Reaction To Indoor Air Quality: Gary J. Raw, D.Phil

Uploaded by

Copyright:

Available Formats

Ch53_Spengler_445549-4 11/20/00 10:41 AM Page 53.

53.1 THE ROLE OF OCCUPANT SURVEYS

53.2 ASSESSING IAQ

The Primacy of Occupant Reactions

The Value of Occupant Data

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.3

The Key Issues Addressed in Occupant Surveys

Acute Nonspecific Symptoms (Sick Building Syndrome). In comparison with specific

53.4 ASSESSING IAQ

Productivity. Although productivity or staff efficiency would appear to be a potential

Sick Building Syndrome

Sick building syndrome is a phenomenon whereby people experience a range of symptoms

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.5

53.6 ASSESSING IAQ

53.2 DECIDING TO CONDUCT AN OCCUPANT

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.7

Each of these is addressed in turn.

Response to Suspected Problems

53.8 ASSESSING IAQ

Research, Including Following Up Mitigation Attempts

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.9

53.3 INSTRUMENTS FOR THE SURVEY

Occupant data can be collected by a number of means, including structured or unstructured

Selecting or Designing a Questionnaire

53.10 ASSESSING IAQ

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.11

Questionnaire Construction. A number of basic issues have to be decided in construct-

53.12 ASSESSING IAQ

The Example of the ROES Questionnaire

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.13

53.14 ASSESSING IAQ

Frequency of Occurrence. For screening purposes, it is not necessary to include ratings

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.15

Questions on Confounding Factors. Two kinds of confounding factors can be included

53.4 PROCEDURES FOR THE SURVEY

A questionnaire is not in itself a method; it is an instrument, which will produce valid

53.16 ASSESSING IAQ

Planning the Survey

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.17

53.18 ASSESSING IAQ

Motivating the Respondents

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.19

Give a contact point for queries.

Direct social interaction can also be valuable by showing appreciation, understanding

Definition of Outcome. Usually, ratings of environmental conditions are considered as

Index or Score to Describe the Total Phenomenon. The principal measure to be

53.20 ASSESSING IAQ

TABLE 53.1 Percentage of Respondents

Interpretation of Environmental Ratings. There is no absolute interpretation of the rat-

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.21

TABLE 53.2 Means and Ranges of Environmental

Rating Mean S.D.

Comfort 3.43 1.79

Comfort 3.28 2.08

Temperature 2.05 1.77

53.22 ASSESSING IAQ

ethical problems by exposing only voluntary subjects to known concentrations of specific

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.23

APPENDIX: REVISED OFFICE ENVIRONMENT

53.24 ASSESSING IAQ

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.25

53.26 ASSESSING IAQ

ASSESSING OCCUPANT REACTION TO INDOOR AIR QUALITY 53.27