Professional Documents
Culture Documents
299
2004 Kluwer Academic Publishers. Printed in the Netherlands.
Abstract
Quality of life (QOL) assessments that are easily administered and which do not impose a great burden on
the respondent are needed for use in large epidemiological surveys, clinical settings and clinical trials. Using
data from the WHOQOL-BREF eld trials, the objectives of this work are to examine the performance of
the WHOQOL-BREF as an integrated instrument, and to test its main psychometric properties. The
WHOQOL-BREF is a 26-item version of the WHOQOL-100 assessment. Its psychometric properties were
analysed using cross-sectional data obtained from a survey of adults carried out in 23 countries
(n 11,830). Sick and well respondents were sampled from the general population, as well as from hos-
pital, rehabilitation and primary care settings, serving patients with physical and mental disorders and with
respect to quotas of important socio-demographic variables. The WHOQOL-BREF self-assessment was
completed, together with socio-demographic and health status questions. Analyses of internal consistency,
itemtotal correlations, discriminant validity and construct validity through conrmatory factor analysis,
indicate that the WHOQOL-BREF has good to excellent psychometric properties of reliability and per-
forms well in preliminary tests of validity. These results indicate that overall, the WHOQOL-BREF is a
sound, cross-culturally valid assessment of QOL, as reected by its four domains: physical, psychological,
social and environment.
World Health Organisation Quality of Life as- an integrated instrument. For this reason, fresh
sessment the WHOQOL-100 is a cross-cultur- data needed to be collected. In this study we pre-
ally valid assessment of well-being. Assessment is sent new survey data that represent the rst eld
operationalized through 100 items representing 25 trial of the WHOQOL-BREF that seeks to con-
facets organised in six domains [6, 7]. The tool was rm and extend information about the properties
developed through a collaboration of 15 sites of the short form as a whole. Specically we aimed
around the world working in their own national to assess itemresponse distributions, internal
language. Centres simultaneously used common consistency reliability and itemscale correlations
protocols that were agreed through international with other aspects of construct and discriminant
consensus at each stage of development process. validity. Evidence of testretest reliability for the
The WHOQOL collaboration pooled information WHOQOL-BREF is already known [9]. It was
throughout the project and this procedure not only predicted that sick participants would report
permits a high level of semantic and conceptual poorer QOL than well participants but no pre-
equivalence to be achieved between language ver- dictions were made for other socio-demographic
sions but also creates a fast track to the rapid and centre dierences.
establishment of multi-lingual instruments [7].
This new procedure whereby centres work simul-
taneously on the same stage of instrument devel- Methods
opment, pooling their ideas and results centrally
(through WHO Geneva) and communicating with Design
each other to achieve equivalence has been de-
scribed as a spoke-wheel methodology, through Data for the WHOQOL-BREF eld trial were
analogy with the spokes and hub of a bicycle wheel collected using a cross-sectional design in 24 cen-
[8]. tres representing 23 countries. The centres were
The WHOQOL-BREF is being developed as a drawn from countries in all the WHO Regions of
short version of the WHOQOL-100 for use in situ- the world, as well as from diverse cultures and
ations where time is restricted, where respondent dierent levels of socio-economic development.
burden must be minimised and where facet-level Data were contributed from eld sites in Argen-
detail is unnecessary e.g. with large epidemiologi- tina, Australia, Brazil, Bulgaria, China, Croatia,
cal surveys and some clinical trials. Using data Germany, Greece, Hungary, Israel, Italy, India:
from 15 centres collected for the WHOQOL-100 Madras and New Delhi, Japan, Malaysia, Neth-
eld trials, items for the WHOQOL-BREF were erlands, Nigeria, Norway, Romania, Russia,
selected for their ability to explain a substantial Spain, Turkey, United Kingdom, United States.
proportion of variance within their parent facet Data collection methods were similar to interna-
and domain, for their relationship with the overall tionally agreed protocols designed during the de-
WHOQOL model and for their discriminant va- velopment of the WHOQOL-100 [6, 7].
lidity [9]. Analysis of these extracted items showed Adult participants (adult was culturally dened)
that a four-factor structure best tted the data [9]. were recruited from a variety of in-patient and out-
Although this contrasted with the original concept patient health care facilities, and from the general
of a 6-domain model for the WHOQOL, it was population. Using a common and consensually
consistent with empirical results from the previous agreed protocol, quota sampling was used to
WHOQOL-100 eld trials [8]. Based on these re- structure the sample so that equal numbers of each
sults, the WHOQOL-BREF was developed in the gender and the two age groups (bisected at
context of four domains of QOL: physical, psy- 45 years) were targeted. Recruitment would span
chological, social and environment [9]. Although the continua of the adult age range, four educa-
extensive analysis had been carried out on the tional levels and types of marital status. Well
WHOQOL-100 eld trial data to reduce items and samples were targeted similarly. The sites aimed to
assess the preliminary psychometric properties of a recruit sick participants from all the main groups of
short form [9], this extracted data was insucient health care users but did not use a quota for di-
to conrm the WHOQOL-BREFs properties as agnosis or severity. Wellness or sickness was de-
301
ned by self-report, from diagnostic categories as- usually for reasons of literacy or disability. Stan-
signed by health professionals, and with reference dard instructions, socio-demographic details and
to contextual knowledge about (non) patient status an item on current health status were completed
relating to the nature of the population(s) ap- before answering the 26 items of the WHOQOL-
proached in collection sites set up by each centre. BREF.
Together the centres would obtain a richly heter- During development of the WHOQOL-100,
ogeneous sample of sick people covering 28 groups four types of 5-point Likert interval scale were
of physical or mental health problems (linked to designed and tested to reect intensity, capacity,
ICD-10 categories) and with varying levels of dis- frequency and evaluation, and one of these was
ease severity and functioning. A fully structured attached to each item [10]. These response scales
design and common protocol was not feasible for were also used in the WHOQOL-BREF. Items
these variables due to an absence of relevant na- inquire how much, how completely, how often,
tional statistics in some parts of the developing how good or how satised the respondent felt in
world, and limited resources for research. the last 2 weeks; dierent response scales are dis-
tributed across the domains [10, 11]. The transla-
tion process used by the WHOQOL Group to
Instrument develop linguistically and culturally appropriate
new versions of the measure has been revised and
The WHOQOL-BREF is an abbreviated 26-item updated from the WHO standard procedures and
version of the WHOQOL-100 containing items is reported elsewhere [6, 12, 13].
that were extracted from the WHOQOL-100 eld
trial data. The WHOQOL-BREF contains one Analysis
item from each of the 24 facets of QOL included in
the WHOQOL-100, plus two benchmark items Frequency, reliability and correlational analyses
from the general facet on overall QOL and general Frequency analyses were performed to assess re-
health (not included in the scoring) (see Table 1). sponse distributions at the item level, globally and
The facets were originally subsumed within one of by country. In line with the WHOQOL-100 pro-
six domains but factor analysis of the WHOQOL- cedure, problematic items were identied as those
100 indicated that Domain 1 could be merged with where the response distribution was skewed such
Domain 3 (physical with independence), and Do- that fewer than 10% of responses fell in any two
main 2 with Domain 6 (psychological with spiri- adjacent scale points for at least 12 of the 24
tuality, religion and personal beliefs) thereby centres. Internal consistency was assessed using
creating four domains of QOL [8]. Similar results Cronbachs a and the contribution of each item to
were found during the extraction of data for the the total a. The average inter-item correlations
WHOQOL-BREF [9] which is currently scored in for domains, and correlation of items with their
four domains: Domain 1: Physical health, Domain intended domain (using corrected itemtotal cor-
2: Psychological, Domain 3: Social relations and relations) were also calculated. Multi-trait/multi-
Domain 4: Environment, with all facet items item analyses were performed to assess internal
scored as part of their hypothesised domain. Do- consistency reliability, and to identify any items
mains are not scored where 20% of items or more that were more highly associated with another
are missing, and are unacceptable where two or domain than its intended domain, or those highly
more items are missed (or 1-item in the 3-item associated with both. (This analysis was based on
social domain). The scores are transformed on a the MAP Multi-trait/Multi-Item Analysis Pro-
scale from 0 to 100 to enable comparisons to be gram by Ware et al. [14] and has been used in
made between domains composed of unequal previous WHOQOL work [7, 8]).
numbers of items.
The WHOQOL-BREF was self-administered by Discriminant validity
respondents but exceptionally, an experienced in- In a preliminary test of discriminant validity, the
terviewer assisted administration by reading items ability of the domain scores to discriminate be-
aloud where self-completion was not possible, tween ill and well groups of respondents was tested
302
by comparing mean scores in the two groups, us- of the data. Eigenvalues, relative magnitude and
ing t-tests. This feature was also assessed by testing direction of factor loadings explaining variance
the relationship between domains and the two and communality, were examined in these analyses.
general facet items, using linear regression analy- Conrmatory factor analyses (EQS.5.7b) using
sis. All domains were expected to be strongly and structural equation modelling [15] were conducted
positively associated with the concept of overall to obtain objective measures of model t.
QOL and health. The impact of gender and age on
scores from those who were sick and well (depen-
dent variable) was assessed through a hierarchical
multiple regression where these socio-demographic Results
variables were entered together as a block, fol-
lowed by mean scores for the domains. Twenty-four centres contributed a total of 11,830
respondents to the WHOQOL-BREF data set
Data structure and model t (range 2408 (Germany) to 41 (Netherlands)). The
Exploratory factor analyses (with Varimax rota- study population consisted of adults aged 1297
tion) were conducted to explore the factor structure years, with a mean age of 45 (SD 16) (mean
303
Table 2. Internal consistency shown by Cronbachs as for domains and centres (n = 11,830)
Table 3. Discriminant validity: t-tests of domain scores for illness vs. well samples
population, only seven items had strong correla- tended domain. Three of these items on energy,
tions (>0.50) with domains other than their in- activities of daily living and work were from the
305
physical domain (Domain 1), and they correlated two items, showed a strong association with the
strongly with the psychological domain (Domain four domains indicating that each one should be
2). The self-esteem item from Domain 2 was considered when evaluating QOL. All nal equa-
strongly correlated with all of the other domains. tion b values were signicant.
The other items were positive feelings, relation- A review of all the itemtotal correlations in the
ships and safety. Specic sites showed more ex- total population showed generally good results
tensive cross-domain correlation, e.g. more than overall. Poor itemtotal correlations (<0.30) were
30 out of 96 per centre in Argentina, Madras, only found for negative feelings and in one centre
Netherlands, Nigeria, Romania, USA, with most only. In 7 out of 24 centres, items on pain and/or
of these items arising from Domains 1 and 2 and dependence on medication were generally prob-
most of these correlations occurring between items lematic in the physical domain, but no other items
in these two domains. However in centres where were consistently so by this criterion, across sites.
sample sizes were small (as with most in the list Itemdomain correlations ranged between 0.48 for
above), this scale of cross-domain correlations pain, to 0.70 for activities of daily living (Domain
would be expected, so these analyses should be 1), from 0.50 for negative feelings to 0.65 for
interpreted with this in mind. spirituality (Domain 2), from 0.45 for sex to 0.57
However, no item for the total sample corre- for personal relationships (Domain 3) and from
lated more strongly with another domain than 0.47 for leisure to 0.56 for nancial resources
with its own domain, but centre-specic analysis (Domain 4). Summary Pearson correlations (one-
identied two items that occasionally correlated tailed test) between domains for the total sample
more strongly with domains other than their in- were strong, positive and highly signicant
tended domain. In seven sites, the item on safety (p < 0.0001), ranging from 0.46 (physical vs. so-
was more strongly correlated with the psycholog- cial) to 0.67 (physical vs. psychological).
ical domain than with its intended domain, envi-
ronment, and in three sites, the energy item Factor analysis
correlated more strongly with the psychological
than the physical domain. As mentioned earlier, the WHOQOL-100 was
Because QOL is a complex construct that can- based on six theoretical domains that were subse-
not be directly measured, to establish its construct quently reorganised into four domains during the
validity, WHOQOL-BREF domain scores can be development of the WHOQOL-BREF. Empirical
compared to general single-item QOL measures evidence showed that facets from the independence
with evident face validity. It was predicted that all and spirituality domains were associated with the
four domains would show a strong and signicant physical and psychological domains respectively,
association with overall QOL and health, and so and that a 4-factor solution ts the data better in
construct validity was partly assessed by correlat- both ill and well populations. Exploratory factor
ing the domain scores with each general item analyses (Varimax rotation) provided no evidence
(Table 4). The overall assessment of QOL was of a better model. Analysis of the total population
most strongly associated with the psychological data showed four factors (eigenvalues >1.0) that
and environment domains, and the overall as- explained 53% of the variance in the data. Centre-
sessment of health with the physical domain, as specic analyses showed that most sites had four to
predicted. A combined variable representing six eigenvalues greater than 1.0 (mode 5; range
overall QOL and health through the sum of these 37) and these explained 5081% of the variance.
Table 4. Validity: association of domains with general facet items (standardized bs) (n = 11,830)
Conrmatory factor analyses [14] were run to Preliminary data from the WHOOL-BREF
re-evaluate the t values found for the original 4-
domain model [9]. These were rst conducted Mean domain scores for the total sample and for
separately on two random, split-half samples of each centre were calculated and found to be rela-
the data (n 5133 and n 5872). The results were tively similar, with means ranging from 13.5 to
acceptable, showing almost identical t indices for 16.2 (SD: 2.63.2). Because earlier analyses
each half and indicating a robust solution, (Ta- showed some dierences between centres in the age
ble 5 and Figure 1) and supporting similar nd- and sex of respondents, mean domain scores ad-
ings from previous studies [9]. Separate analyses justed for these factors are presented in Table 6.
conducted on sick (n 3313) and well sub-sam- Using 12.0 as the scale midpoint where QOL is
ples (n 3862) also demonstrated an acceptable t judged to be neither good nor poor, inspection of
for this model. the means shows that on average, QOL is accept-
For each of the above analyses, the model for able to very good physically, psychologically and
the 6-domain solution was also calculated and socially in all centres but is poorest where envi-
Table 5 shows that the t indices are marginally ronmental QOL is considered.
poorer in each case. However this decrement is not Table 7 provides preliminary comparisons be-
so large as to discount the possibility that a 6- tween groups, dened by age and sex. The psy-
domain model might be used in appropriate con- chological and social domains showed signicant
texts where it is justied. Some caution is urged in dierences indicating that women have better so-
the interpretation of results in view of the sub- cial QOL but poorer psychological QOL than
stantial data contributions from ve centres. men. This table also shows that mean domain
scores decrease with age, and the greatest changes
Table 5. Structural equation modelling t indices of WHO- are to be found in physical health.
QOL-BREF models
4-domain 6-domain
model model Discussion
Random split half sample A*
As with its more comprehensive counterpart the
(n = 5133)
v2 6830.8 7624.4 WHOQOL-100 the results from this eld trial of
df = 249 df = 248 the WHOQOL-BREF are noteworthy because
CFI 0.863 0.847 they provide supportive evidence for the cross-
RCFI 0.865 0.849 cultural validity of this QOL measure. Although
RMSEA 0.07 0.08
not designed to assess each of the 24 specic QOL
Random split half sample B facets in detail, with only 26 items the WHOQOL-
(n = 5118) BREF is short enough to be used where time is at a
v2 6791.0 7132.3
CFI 0.864 0.857
premium, where respondent burden is high or
RCFI 0.866 0.859 where facet detail is unnecessary. It has wide
RMSEA 0.07 0.07 ranging uses in clinical settings and clinical trials.
Sick sample (n = 3313) Although longer than some other short-forms, the
v2 3736.9 4418.6 WHOQOL-BREF covers a very broad range of
CFI 0.876 0.851 facets that were agreed by international consensus.
RCFI 0.878 0.854 A noteworthy feature is the inclusion of social and
RMSEA 0.07 0.07
environment domains for assessment.
Well sample (n = 3862)
v2 4991.3 4995.6 The purpose of this paper was to examine the
CFI 0.868 0.868 psychometric properties of the WHOQOL-BREF
RCFI 0.872 0.871 in terms of itemresponse distributions, internal
RMSEA 0.07 0.07 consistency reliability, discriminant validity and
CFI Comparative Fit Index; RCFI Robust Comparative Fit construct validity. The results showed that the
Index; RMSEA Root mean square error approximation. instrument performs well, although some areas
* Shown in Figure 1. deserve further attention. In particular, centre-
307
specic analyses identied some items that did not conceptualization of these issues appears to depart
discriminate well between domains and two that from the theoretical concept. Because no evidence
had stronger correlations with domains other than of these associations was found during the devel-
their intended domain. While the identication of opment of the WHOQOL-100 or in the item-se-
a particular item with its intended domain can be lection process, it is possible that the WHOQOL-
improved by changing the wording and semantics BREF format may have led to a change in the
of the translation to reinforce the intended con- context and thus the conceptualisation of the item
cept, the identication of such items also provides by respondents. Further development of the
information relevant to construct validity. In some translated instrument (including cognitive de-
centres, items on safety and energy were more brieng) should focus on this issue, particularly for
strongly associated with the psychological domain the safety item.
than their intended domains environment and Previous development work based on the ex-
physical respectively so in these centres the traction of item data from the WHOQOL-100 eld
308
Table 6. WHOQOL-BREF domain scores (range 420) adjusted for age and sex, by center (n = 11,830)
trials indicated that while the WHOQOL-100 was have arisen because centres did not contribute
based on a theoretical model with six domains, exactly the same prole of diagnostic groups and
empirical evidence supported a 4-domain model [8] well people. Although these analyses provide good
and similar results were found in the development support for using the WHOQOL-BREF in QOL
of the WHOQOL-BREF [9], so this was not un- assessment, work needs to be carried out to further
expected given the conceptual similarity of the examine the testretest reliability of the instrument
domains. Analysis of the present WHOQOL- (but see Ref. [9]), its concurrent validity in com-
BREF eld trial data conrmed that overall, the 4- parison to relevant other measures, and within a
domain model ts the data well, and also for sick longitudinal design to examine sensitivity to
and well respondents separately. But further changes in health states over time.
analysis showed that the 6-domain model was also
a good t results which are generally consistent
with the structure of the WHOQOL-100. Al- Conclusion
though the more parsimonious 4-domain model is
applied, these results provide an empirical basis for The WHOQOL-BREF arises from 10 years of
extracting more information from the items/facets development research on QOL and health care. It
and scoring of six domains in situations where this is a person-centred, multilingual instrument for
is appropriate e.g. where spirituality needs assess- subjective assessment and is designed for generic
ment in palliative care. use as a multi-dimensional prole, so enabling a
Some caution is urged in the interpretation of wide range of diseases and conditions to be com-
results, given the substantial data contributions pared. A new methodology has been created
from ve centres. Other biases from sampling may within this project whereby international centres
309
Table 7. Comparison of WHOQOL-BREF mean domain scores by gender and age group (n = 11,830)
Gender
Men
Mean 14.3 14.2 14.1 13.8
SD 2.9 2.8 3.2 2.7
Women
Mean 14.2 14.0 14.4 13.9
SD 3.1 2.8 3.1 2.6
F 2.3 17.5 27.1 2.0
p 0.13 0.001 0.001 0.16
Age
1220
Mean 15.6 14.8 14.9 14.4
SD 2.9 2.8 3.1 2.4
2130
Mean 15.0 14.3 14.5 13.7
SD 2.9 2.8 3.4 2.6
3140
Mean 14.0 13.9 14.0 13.6
SD 3.0 2.8 3.2 2.7
4150
Mean 13.9 14.0 14.1 13.9
SD 2.9 2.7 3.1 2.6
5160
Mean 13.3 13.8 14.1 14.0
SD 2.9 2.8 2.9 2.6
61+
Mean 14.2 14.1 14.2 13.8
SD 3.0 2.8 3.2 2.6
F 109.5 18.0 10.6 11.2
p 0.01 0.01 0.01 0.01
worked simultaneously from a common protocol plications for its use in research involving a variety
at each stage of the development. International of interventions, as well for applications in many
consensus was also obtained at each stage to guide service settings. More work on the remaining
the direction of the research. This replaces the properties of validity (e.g. concurrent), sensitivity,
standard serial translation method that is more and feasibility are required. Future research could
commonly used in cross-cultural work and this obtain more comprehensive global survey data
new procedure has reduced some of the problems (e.g. including Arabia), of more consistent quality,
of obtaining semantic and conceptual equivalence and with structured diagnostic samples of patients.
between language versions of the instrument. However its conceptual and methodological
The WHOQOL-BREF has several strengths. It strengths, combined with the good psychometric
is based on a cross-culturally sensitive concept and properties described in this paper, suggest that
is available in most of the worlds major lan- WHOQOL-BREF may have a place among the
guages; hence it is appropriate for use in multi- leading generic QOL instruments.
national collaborative research. It consists of QOL
items that are concerned with the meaning of dif-
ferent aspects of life to the respondents, and how Acknowledgements
satisfactory or problematic is their experience of
them. In addition, the WHOQOL-BREF can The WHOQOL Group acknowledges the assis-
generate a prole of four domain scores within a tance of Dr C. Nelson in the early preparation of
relatively small item set of 26 items. This has im- the manuscript.
310
Appendix 1 3. Muldoon MF, Barger SD, Flory JD, Manuck SB. What are
quality of life measurements measuring? Br Med J 1998;
The WHOQOL group. 316: 542545.
The WHOQOL Group comprises a co-ordinating group of 4. Guyatt GH, Naylor CD, Juniper E, et al. Users guide to
collaborating investigators in each of the eld sites and a panel the medical literature XII: How to use articles about
of consultants. Dr D. Rex Billington directed this project that healthrelated quality of life. J Am Med Assoc 1997; 277:
was initiated by Dr John Orley and Dr Norman Sartorius. The 12321237.
work reported here was carried out in 24 of the eld sites: 5. WHOQOL Group. Development of the WHOQOL: Ra-
Professor H. Herrman, St. Vincents Hospital, Australia; Dr S. tionale and current status. Int J Mental Health 1994; 23:
Bonicatto, FUNDONAR, Argentina; Dr M. Fleck, University 2456.
of the State of Rio Grande do Sul, Brazil; Dr V. Petkov, Na- 6. WHOQOL Group. The World Health Organization Qual-
tional Centre for Interdisciplinary Human Studies, Bulgaria; ity of Life assessment (WHOQOL): Position paper from the
Professor S. Szabo, Prof Z. Metelko and Mrs M. Pibernik- World Health Organization. Soc Sci Med 1995; 41: 1403
Okanovic, University of Zagreb, Croatia; Professor M.C. An- 1409.
germeyer and Dr R. Kilian, Department of Psychiatry, Uni- 7. WHOQOL Group. The World Health Organization Qual-
versity of Leipzig, Leipzig, Germany; Dr L. Kullmann, ity of Life assessment (WHOQOL): Development and
National Institute for Medical Rehabilitation, Hungary; Dr S. general psychometric properties. Soc Sci Med 1998a; 46:
Kumar, Chennai Medical College, Madras, India; Dr S. Saxena 15691585.
and Dr Rachna Bhargava, All India Institute of Medical Sci- 8. Skevington SM. Measuring quality of life in Britain: In-
ences, New Delhi, India; Dr M. Amir, Ben-Gurion University troducing the WHOQOL-100. J Psychosom Res 1999; 47:
of the Negev, Beer-Sheva, Israel; Dr G. de Girolamo, Instituto 54495459.
Superiore di Sanita, Rome, Italy; Dr M. Tazaki, Science Uni- 9. WHOQOL Group. Development of the World Health Or-
versity of Tokyo, Japan; Dr H. Che Ismail, University of Sains ganization WHOQOL-BREF quality of life assessment.
Malaysia, Kelantan, Malaysia; Dr M. Kalfoss, Oslo College, Psychol Med 1998b; 28: 551558.
Norway; Dr A. Lomachenkov, V.M. Bekhterev Psychoneuro- 10. Szabo S, Orley J, Saxena S on behalf of the WHOQOL
logical Research Institute, St Petersburg, Russian Federation; Group. An approach to response scale development for
Dr R. Lucas Carrasco, Barcelona, Spain; Dr C. Fidaner, Izmir cross-cultural questionnaires. Eur Psychol 1997; 2: 3270
Cancer Registry Center, Turkey; Professor S. Skevington, Ms 3276.
K. OConnell and Ms A. Wright, University of Bath, Bath, 11. Sartorius N, Kuyken W. Translation of health status
United Kingdom; Professor D. Patrick, Ms M. Martin and Dr measures. In: Orley J and Kuyken W (eds), Quality of
D. Buesching, University of Washington, Seattle, United States. Life Assessment: International Perspectives. Heidelberg:
Some data was collected as part of the Longitudinal Investi- Springer Verlag, 1994.
gation of Depression Outcomes (LIDO) study directed by 12. WHO The World Health Organization. WHOQOL User
Health Research Associates Inc., and in several eld centres of Manual. Geneva, World Health Organization, 1998.
the WHO Disability Assessment Schedule (WHODAS) vali- 13. Skevington SM. Advancing cross-cultural research on
dation study (ACE Group, WHO, Geneva). Further informa- quality of life: Observations drawn from the WHOQOL
tion on the WHOQOL can be obtained at http:// development. Qual Life Res 2002; 11: 135144.
www.who.ch.\msa\mnh\mhp\q1.htm 14. Ware JE, Harris WJ, Gandek B, Rogers BW, Reese PR.
MAP-R for Windows: Multi-trait/Multi-Item Analysis
Program-Revised. Users Guide. Boston, MA: Health As-
sessment Lab, 1997.
References 15. Bentler, PM. EQS Structural Equation Program Manual.
Encino, C.A: Multivariate Software, Inc, 1995.
1. Wood-Dauphine S. Assessing quality of life in clinical re-
search: From where have we come and where are we going? Address for correspondence: Suzanne Skevington, WHO Centre
J Clin Epidemiol 1999; 52: 355363. for the Study of Quality of Life, Department of Psychology,
2. McHorney CA. Health status assessment methods for University of Bath, Bath BA2 7AY, UK
adults: Past accomplishments and future challenges. Ann Phone: 44-1225-826830; Fax: 44-1225-826752
Rev Public Health 1999; 20: 309335. E-mail: s.m.skevington@bath.ac.uk