A Rasch analysis of a self-perceived change in quality of life scale in patients
with mild stroke
Jau-Hong Lin1,2, Wen-Chung Wang3, Ching-Fan Sheu4, Sing Kai Lo5, I-Ping Hsueh6
& Ching-Lin Hsieh6,7
1Faculty of Physical Therapy, College of Health Science, Kaohsiung Medical University, Taiwan;
2Department of Rehabilitation Medicine, Kaohsiung Medical University Hospital, Taiwan;3Department of
Psychology, National Chung Cheng University, Taiwan;4Department of Psychology, DePaul University,
Occupational Therapy, College of Medicine, National Taiwan University, Taiwan;7Department of Physical
Medicine and Rehabilitation, National Taiwan University Hospital, Taiwan (E-mail: firstname.lastname@example.org)
5The Faculty of Health and Behavioural Sciences, University of Deakin, Australia;
Accepted in revised form 27 May 2005
A Rasch analysis was used to assess the unidimensionality and appropriateness of the scoring level of a 13-
item self-perceived change in quality of life scale (CQOL) for stroke patients. A total of 158 patients with
mild stroke completed the CQOL themselves at home. The results showed that a unidimensional CQOL
can be created by deleting the three items related to speaking, vision, and thinking. The 4 scoring categories
of the shortened scale were deemed appropriate from the analysis. These results provide preliminary
evidence of the 10-item CQOL in assessing self-perceived change in quality of life in stroke patients. Further
studies are needed to examine the test-retest reliability, criterion validity, and responsiveness of the 10-item
CQOL in stroke patients.
Key words: Quality of life, Stroke, Unidimensionality
Stroke patients can be impaired in the physical,
social, emotional, and other health-related quality
of life (HRQOL) domains . Even among pa-
tients with good functional recovery, many still
decline substantially after discharge from the
hospital and have very poor HRQOL for several
years [2, 3]. Longitudinal research into the
HRQOL of stroke survivors is therefore war-
Although self-perceived change in HRQOL is
critical in the patient-centered approach to out-
come measurement [4, 5], few HRQOL indices
directly assess patient’s own perceptions of clini-
cally important changes. Williams and coworkers
 developed the Stroke-Specific Quality of Life
Scale (SS-QOL) and found that there were 12
domains that were reliable, valid, and responsive
in stroke patients: energy, family roles, language,
mobility, mood, personality, self-care, social roles,
thinking, vision, upper extremity function, and
work/productivity. These authors constructed a
13-item scale (i.e., the 12 items based on the
aforementioned 12 domains and an item relating
to the overall QOL), called the change of QOL
(CQOL). The patients were asked to rate the items,
compared with their QOL prior to stroke, as ‘a lot
worse = 1’, ‘somewhat
worse = 3’, or ‘the same = 4’. The CQOL cur-
rently is the only instrument measuring stroke
patients’ self-perceived change in HRQOL. How-
ever, its psychometric properties in stroke patients
were largely unknown.
Rasch analysis is useful in testing whether
items from a test measure a unidimensional
worse = 2’,‘alittle
Quality of Life Research (2005) 14: 2259–2263
? Springer 2005
construct [7, 8], which is required to justify the
summation of scores. It is also useful in deter-
mining appropriateness of the scoring levels of a
scale, which refers to whether or not participants
can be differentiated by their responses as clearly
as the levels allow . In addition, Rasch analysis
provides the person separation statistic and tar-
geting information [10, 11]: The former can be
used to evaluate the extent to which a question-
naire can distinguish those with different levels of
HRQOL, and the latter, to examine the extent to
which the items of a questionnaire are of appro-
priate difficulty for the participants studied. This
study used a Rasch analysis to assess the unidi-
mensionality, appropriateness of the scoring level,
person separation, and appropriateness for the
sample (targeting) of the CQOL for stroke pa-
Participants were recruited from the registry of the
Quality of Life after Stroke Study in Taiwan
between December 1, 1999 and December 31,
2001. Details of the selection procedures have been
reported elsewhere .
The Chinese version of the CQOL was used in this
study. The CQOL was translated from English
into Mandarin Chinese based on general proce-
dure . The Canadian Neurological Scale (CNS)
was used to assess severity after stroke . The
score ranges from 0 to 11.5. The Barthel Index (BI)
is a measure of the severity of basic activities of
daily living (ADL) function . The total score is
a simple additive summary score across 10 items.
Its score ranges from 0 to 20.
The patients completed the CQOL questionnaire
themselves at home. Initial stroke severity was
ascertained from applying the CNS retrospectively
to medical records. An occupational therapist
administered the BI to the patients and their
caregivers in a face-to-face interview.
The unidimensionality of the CQOL was examined
using the Rasch rating scale model  with the
WINSTEPS computer program . To examine
the unidimensionality of the CQOL, the infit and
outfit statistics were used to examine whether the
data fit the model’s expectation. Items with infit or
outfit mean square error (MNSQ) greater than 1.3
indicated potential misfits [11, 17]. Furthermore,
when items fit the model’s expectation, the resid-
uals (observed scores minus expected scores)
should be randomly distributed. A principal
component analysis was conducted to verify whe-
ther any dominant component existed among the
residuals. If dominant components were found, the
unidimensionality assumption was violated.
We examined the Rasch estimates of the inter-
section parameters of each item of the CQOL to
see if the scaling levels were adequate. Whether the
items of the CQOL cover the full range (extent) of
self-perceived change in HRQOL was verified by
examining the gaps between the item difficulties
along the item hierarchy. In addition, the person
separation statistic refers to the ratio of spread in
item difficulties to the error in estimating them. A
value P2.0 is acceptable . Furthermore, the
average person measure was used to determine the
extent to which the set of items is of appropriate
difficulty for the patients. An absolute average
person measure P0.5 indicates slight mistargeting
A total of 158 patients who had a stroke within the
previous 6–12 months and who were living in the
community participated in this study. The patients
had mild to moderate severity at admission as
shown by their CNS scores. The BI scores indi-
cated that patients were slightly limited in basic
ADL. Detailed characteristics of these patients are
shown in Table 1.
Three items (i.e., speaking, vision and thinking)
were found to be poor-fitting (both infit and outfit
MNSQs larger than 1.3) and were deleted from
further analysis. Table 2 shows the estimates of
overall difficulties, infit MNSQ and outfit MNSQ
for the 10 remaining items. A principal component
analysis on the standardized residuals of Rasch-
transformed scores revealed no dominant principal
component (the first and the second components
accounted for only 25 and 16% of the residual
variances, respectively). These results suggested
that the shortened 10-item CQOL constitutes a
unidimensional construct measuring self-perceived
change in HRQOL.
The Rasch estimates of the three intersection
parameters were )1.16, )0.54 and 1.70, respec-
tively, all with a standard error (SE) of 0.09. As the
range of the intersection parameters (2.86) was
rather wide compared with the SD for person
measures (1.99), the 4-level scaling of the short-
ened 10-item CQOL was considered appropriate.
No floor or ceiling effects were found since only
16 (10%) of the 158 patients had extreme scores (7
receiving the highest possible score and 9 receiving
the lowest possible score). The person separation
index (3.12) was satisfactory, indicating that the
scores on the shortened 10-item CQOL could dis-
tinguish four distinct levels (strata) of change
in HRQOL. Furthermore, the average patient
measure was close to zero ()0.03), indicating that
the items were well targeted to the sample .
Figure 1 shows that the estimated perceived chan-
ges of the patients were well scattered throughout
the possible score range of the shortened 10-item
CQOL and that there were no substantial gaps
between the item difficulties (the extent of change in
HRQOL) along the item hierarchy.
Our study found that a unidimensional CQOL can
be created by deleting the three items related to
speaking, vision, and thinking. The total score of
Table 2. Difficulty (extent of self-perceived change in HRQOL), standard error (SE), and infit and outfit statistics of the modified
Item* Difficulty logitSE Infit MNSQOutfit MNSQ
Upper extremity function
Overall quality of life
* The items are arranged in descending order of self-perceived change in HRQOL.
Table 1. Characteristics of the patients with stroke (N=158)
Mean years (SD)
Median (lower-upper quartile)
Side of hemiplegia
CNS score at admission
Days after stroke at interview
Assessment results in the community
10-item CQOL raw score
Median (lower-upper quartile)
Median (lower-upper quartile)
the shortened CQOL scale can be used to represent,
on a single dimension, a patient’s perception of
change in HRQOL. A higher score indicates a
smaller degree of self-perceived change in HRQOL
before and after stroke. From the substantive point
of view, it often does not make sense to combine
items from separate (or distinct) domains to gauge
performance on a single scale. However, the
shorten scale for the overall CQOL for stroke pa-
tients proposed in this study was supported by the
Rasch analysis and it can be expected to yield more
precise estimates of changes in QOL for stroke
patients at a given level of limitation, and translate
into clinically sound decision regarding treatment.
The energy item was located in the highest part
of item difficulty (with highest difficulty logit),
indicating the greatest self-perceived change in
HRQOL as compared with the rest of the items,
while personality had the lowest difficulty logit.
The ordering of the items of shortened CQOL
generally follows clinical expectations (e.g., the
patients perceived more change in ‘‘mobility’’ and
‘‘work/productivity’’ than in ‘‘self-care’’). These
results are useful in a patient-centered approach.
For example, the ordering of the items can help
determine treatment priority in clinics. The clini-
cians may establish their treatment plans based
on the extent and hierarchy of change in HRQOL
perceived by their patients.
The validity of assessing change in HRQOL over
a response shift in perceptions and internal stan-
dards [18, 19]. They may judge their subjective
objective measures of health, or alternatively,
judgments of health may change in a situation
new information. In addition, the implicit theories
of change in subjective judgments of change from a
previous state may have an influence on assessing
change in HRQOL over time . Accordingly,
further studies are needed to examine how response
shift and the implicit theory of change affect the
measurement of change in HRQOL.
Our results provide good evidence of unidi-
mensionality and appropriateness of scaling of the
shortened 10-item CQOL in mild stroke patients.
In addition, the 10 items were well targeted to the
sample and covered a comprehensive range of self-
perceived change in HRQOL in our patients. To
comprehensively validate the shortened 10-item
CQOL, further studies are needed to validate the
scale in stroke patients with characteristics differ-
ent from those recruited for this study (e.g., pa-
tients having had a stroke within 6 months and
patients with moderate or severe disability) and to
examine its test-retest reliability, criterion validity
and responsiveness. Furthermore, because pa-
tients’ cognitive function plays an important role
in assessing theirself-perceived
HRQOL, cognitive function, especially memory,
Figure 1. Observed average logits (the extent of self-perceived change in HRQOL) by scaling level of each item of modified 10-item
CQOL and patient distribution. The items are arranged in descending order of self-perceived change in HRQOL. The scaling levels
(1, 2, 3, and 4) indicate ‘a lot worse’, ‘somewhat worse’, ‘a little worse’, and ‘the same’, respectively. The axis at the bottom of the
figure (as well as the one on top) indicates the location of persons on the logit scale. The two rows of digits beneath indicate the
number of persons at that location, with the top row standing for units of 10 and the bottom one, units of one.
of the subjects should be assessed in future vali- Download full-text
This study was supported by research grants from
the National Science Council (NSC 93-2314-B-
002-033 and NSC 92-2314-B-037-066) and the
National Health Research Institute (NHRI-EX93-
1. Kim P, Warren S, Madill H, Hadley M. Quality of life of
stroke survivors. Qual Life Res 1999; 8: 293–301.
2. Sturm JW, Donnan GA, Dewey HM, Macdonell RA,
Gilligan AK, Srikanth V, Thrift AG. Quality of life after
stroke: the North East Melbourne Stroke Incidence Study
(NEMESIS). Stroke 2004; 35: 2340–2345.
3. Carod-Artal J, Egido JA, Gonzalez JE, de Siejas V.
Quality of life among stroke survivors evaluated 1 year
after stroke Experience of a stroke unit. Stroke 2000; 31:
4. Buck D, Jacoby A, Massey A, Ford G. Evaluation of
measures used to assess quality of life after stroke. Stroke
2000; 31: 2004–2010.
5. Doyle PJ. Measuring health outcomes in stroke survivors.
Arch Phys Med Rehabil 2002; 83(Suppl 2): S39–43.
6. Williams LS, Weinberger M, Harris LE, Clark DO, Biller J.
Development of a stroke-specific quality of life scale.
Stroke 1999; 30: 1362–1369.
7. Rasch G. Probabilistic models for some intelligence and
attainment tests. Chicago: MESA Press, 1980.
8. Wright BD, Masters GN. Rating scale analysis Rasch
measurement. Chicago: MESA Press, 1982.
9. Wright BD, Mok M. Rasch models overview. J Appl Meas
2000; 1: 83–106.
10. Duncan PW, Lai SM, Bode RK, Perera S, DeRosa J.
Stroke Impact Scale-16: A brief assessment of physical
function. Neurology 2003; 60: 291–296.
11. Prieto L, Alonso J, Lamarca R. Classical test theory versus
Rasch analysis for quality of life questionnaire reduction.
Health Qual Life Outcomes 2003; 1: 27.
12. Mao HF, Hsueh IP, Tang PF, Sheu CF, Hsieh CL.
Analysis and comparison of the psychometric properties of
three balance measures for stroke patients. Stroke 2002; 33:
13. Aaronson N, Alonso J, Burnam A, Lohr KN, Patrick DL,
Perrin E, Stein REK. Assessing health status and quality of
life instruments: attributes and review criteria. Qual Life
Res 2002; 11: 193–205.
14. Goldstein LB, Chilukuri V. Retrospective assessment of
initial stroke severity with the Canadian Neurological
Scale. Stroke 1997; 28: 1181–1184.
15. Mahoney FI, Barthel DW. Functional evaluation: the
Barthel Index. Md State Med J 1965; 14: 61–65.
16. Linacre JM. WINSTEPS Rasch Measurement Computer
Program. Chicago: IL, 2003.
17. Wright BD. Reasonable mean-square fit values. In: Wright
BD, Linacre JM (eds), Rasch Measurement Transactions.
Part 2. Chicago: MESA, 1994.
18. Norman G. Hi! How are you? Response shift, implicit
theories and differing epistemologies. Qual Life Res 2003;
19. Ahmed S, Mayo NE, Wood-Dauphinee S, Hanley JA,
Cohen SR. Response shift influenced estimates of change in
health-related quality of life poststroke. J Clin Epidemiol
2004; 57: 561–570.
Address for correspondence: Dr Ching-Lin Hsieh, School of
Occupational Therapy, College of Medicine, National Taiwan
University, 7 Chung-Shan S. Rd., Taipei 100, Taiwan