Blackwell Science, LtdOxford, UKVHEValue in Health1098-30152004 ISPORSeptember/October 20047Supplement 1S22S26Original ArticleRasch Analysis in QoL Instrument DevelopmentTennant et al.

Volume 7 • Supplement 1 • 2004

V A L U E I N H E A L T H

Application of Rasch Analysis in the Development and

Application of Quality of Life Instruments

Alan Tennant, PhD,1 Stephen P. McKenna, PhD,2 Peter Hagell, PhD3

1
Academic Unit of Musculoskeletal & Rehabilitation Medicine, University of Leeds, Leeds, UK; 2Galen Research, Manchester, UK;
3
Department of Nursing, Lund University, Lund, Sweden

ABST R ACT

This paper discusses recent advances that have been made order to ensure that they provide unidimensional meas-
in the field of psychometrics, specifically, the application urement. By ensuring that scales are based on the same
of Rasch analysis to the instrument development process. measurement model and that they fit the Rasch model it is
It emphasizes the importance of assessing the fundamen- possible for QoL scores to be compared across diseases by
tal scaling properties of an instrument prior to consider- means of cocalibration and item banking.
ation of traditional psychometric indicators. The paper Keywords: classical test theory, differential item func-
introduces Rasch analysis and shows how it has been tioning, needs-based quality of life, Rasch analysis,
applied in the development of needs-based measures in undimensionality.

As long as primitive counts and raw scores are rou- way to ensure that earlier measures are updated to
tinely mistaken for measures by our colleagues in fit the Rasch model [4].
social, educational, and health research, there is no The quest for measurement is an important part
hope of their professional activities ever developing of advancing science, and the type of measurement,
into a reliable or useful science [1]. which allows for arithmetic operations such as
A crucial aspect of the application of the needs addition and subtraction, is known as fundamental
model is that it has been allied with the most measurement [5]. Most outcome measures used in
advanced psychometric methods. This paper argues health care are ordinal in nature, precluding such
the case for the use of Rasch analysis to ensure that arithmetic operations [6]. Many such measures
scales are unidimensional [2], a fundamental focus on attributes that are not directly measurable,
requirement of construct validity [3]. The paper such as pain, self-esteem, or quality of life. These
gives an overview of the nature of Rasch analysis measures give a “manifest score” of the construct
and shows how it aids valid across-disease compar- being measured. Consequently, most outcomes are
isons of quality of life (QoL) by means of co- expressed as ordinal manifest scores, indicating
calibration and the development of item banks. some rank on a perceived underlying latent trait.
Over the last two decades an analytical approach Although there is a substantial body of nonpara-
has been adopted that is pivotal to both judging metric statistics to analyze such information, the
the quality of existing outcome instruments and importance of the calculation of change scores in
in developing new instruments. This approach is clinical trial analysis (and attributes of measure-
called Rasch analysis, after its originator, a Danish ment such as the “effect size” [7]), which require
mathematician. He developed Poisson models for normally distributed interval-level measurement,
reading, intelligence, and achievement tests, the last gives urgency to achieving a quality of measurement
becoming known as the Rasch model [2]. Rasch that will sustain such arithmetic operations.
analysis has been employed in the development of In order to achieve such fundamental measure-
most of the needs-based QoL instruments, to ensure ment certain properties are required. These are
that the resulting scales are unidimensional. Only reviewed in detail elsewhere [8,9] but essentially
the earliest developed needs-based measures did not they are:
benefit from this approach and studies are under-
• the numerical properties of order (one mark on
the ruler represents more or less of the con-
Address correspondence to: Stephen McKenna, Galen
Research, Enterprise House, Manchester Science Park, Lloyd struct than another);
Street North, Manchester M15 6SE, UK. E-mail: • addition (points on rulers may be added
smckenna@galen-research.com together); and

© ISPOR 1098-3015/04/$15.00/S22 S22–S26 S22

Rasch Analysis in QoL Instrument Development
• specific objectivity (the calibration of the ruler
(item set or questions) is independent of the
persons used to calibrate and vice versa).
Where data fit the Rasch model these properties
are confirmed and fundamental measurement fol-
lows. On a more formal level, the theory of simul-
taneous conjoint measurement [10] provides the
mechanism for translating manifest to latent scores:
Rasch analysis delivers conjoint measurement when
data fit the model.
The Rasch model is a unidimensional model that
has two main assertions:
1. that the easier the item is, the more likely it will
be passed (affirmed); and
2. the more able the patient, the more likely they
will pass (affirm) an item (or do a task) com-
pared to a less able patient.
Unidimensionality is a prerequisite to the summa-
tion of any set of items [3,11,12]. The Rasch model
assumes that the probability of a given patient
“passing” an item or task is a logistic function of
the relative distance between the item location
parameter (the difficulty of the task) and the
respondent location parameter (the ability of the
patient), and only a function of that difference.
Expressed formally, this gives:
e (q -bi )
pi (q) =
1 + e (q -bi )
where pi(q) is the probability that patients with abil-
ity q will be able to do item (task) i, and b is the item
(task) difficulty parameter. The model can be
extended to cope with items with more than two
response categories. From this, the expected pattern
of responses to a set of items or tasks is determined
given the estimated q and b. When the observed
response pattern coincides with or does not deviate
greatly from the expected response pattern, the
items fit the measurement model and constitute a
true Rasch scale [13]. Various fit statistics determine
whether or not the data do fit the model, and these
tend to be software dependent, although all work
on the principal of looking at the deviation of the
observed data from the model expectation. Finally,
where there is local independence of items (that is,
no residual associations in the data after the Rasch
trait has been removed), this, taken together with fit
to the model, supports the contention that the scale
is unidimensional [14].
Assuming that the data fit, the Rasch model
transforms them from ordinal scores into interval
level measurement with the logit (log odds unit) as
S23
the unit of measurement. A logit is the distance
along the line of the variable that increases the odds
of observing the event by a factor of 2.718. There is
a clear relation between the ability–difficulty differ-
ence, and the probability of affirming an item or
undertaking a task. For example, where the differ-
ence between a patients’ ability and the item or task
difficulty is zero, the probability is 0.5. Where the
difference is +1 logit (that is, the patient has greater
ability—or more of the trait—than expressed by the
item) the probability is 0.73, or 0.27 if the differ-
ence is -1.0. Where the difference is ±3 logits then
the probabilities are 0.95 and 0.05, respectively.
Differential Item Functioning (DIF) can also be
examined by fitting data to the Rasch model [15].
Essentially, the scale should work in the same way,
irrespective of the group assessed. Thus, the proba-
bility of being able to do a task, or affirming an
item, for patients at the same level of ability (or, for
example, with the same QoL) should remain the
same across groups. Assessment of DIF can yield
crucial information about the measurement equiva-
lence of an instrument between various cultural
groups [16] but should also be applied across gen-
der and age groups within those cultures.
In comparison with classical test theory, the
Rasch model provides a means of assessing a range
of additional measurement properties, increasing
the information available about a scale’s perform-
ance [17–19]. The model is one of many used in this
way, which are generally subsumed under the rubric
of Item Response Theory (IRT) [20,21]. The Rasch
model is known as the one-parameter model within
this framework, but it has unique properties, which
are crucial to attaining conjoint measurement [22],
a prerequisite for the calculation of change scores
[7].
The Rasch model was readily adopted in rehabil-
itation in the late 1980s [23], as the language of
ability and difficulty easily transferred from educa-
tion. Patients undergoing rehabilitation have a
given level of ability. In order to assess this level they
can be presented with a range of tasks requiring dif-
fering degrees of ability. Since then the approach has
become used with a wide range of clinical and diag-
nostic groups [24,25]. All recent needs-based qual-
ity of life instruments are developed using this
approach [26–32].
Given that both patients and items are calibrated
on the same underlying metric trait, the potential
for innovation in measurement is considerable.
Consider for example the current debate about
disease-specific and generic QoL measures. Where
scales are based on the same theoretical unidimen-
S24
sional construct, items from different diseases can
be calibrated on the same scale, given that some
items (that are free of DIF by diagnosis) common to
both scales are employed. This provides disease-spe-
cific and comparative QoL measures by “item bank-
ing” items or questions onto the same underlying
metric [33–35]. Currently this approach, based on
the needs-based model of QoL, is being used to
establish an item bank for disease-specific QoL
measures in the rheumatic diseases [36–38]. A sim-
ilar exercise is planned for dermatology and links
between these two disease areas could be made pos-
sible by means of the Psoriatic Arthritis Quality of
Life (PSAQoL) measure [37].
The needs-based QoL measures all have the
same theoretical basis, are unidimensional (insofar
as their items fit the Rasch model) and have good
traditional psychometric properties. Not only do
they work as effective outcome measures in clinical
trials but they also offer the potential for allowing
valid comparisons of QoL to be made across dis-
eases [39] and between healthy and diseased popu-
lations [40]. While it has been common practice to
use generic health status measures such as the SF-
36 to make such comparisons, the results have
been both misleading and invalid [39]. This is
because, although a question is expressed in the
same way for all respondents, different types of
patients who have had different experiences inter-
pret it differently. For example, a “yes” response to
a question about feeling tired can represent a very
different response for a healthy person and one
with rheumatoid arthritis. This explains why sur-
prising results are frequently obtained for cross-
disease comparisons. For example, data collected
with the SF-36 suggest both that individuals with
psoriasis have worse scores than patients with
arthritis, cancer, and myocardial infarction [41]
and that such patients have comparable or even
better scores than those experienced by an average
population [42].
Summary
Only occasionally do we see concerns raised about
inappropriate analysis of data that are erroneously
assumed to be at the interval level [6]. The extent to
which analyses of reliability, validity, and respon-
siveness are compromised by ignoring such assump-
tions is unknown. It is also unknown at present to
what extent the misuse of ordinal manifest scores
compromises the results of clinical trial analyses
when these scores are used to calculate changes
across experimental and control groups. However,
Tennant et al.
the potential implications should not be underesti-
mated [43–45].
The ability of a scale to provide fundamental
measurement should be established prior to the
more commonly reported psychometric attributes.
Rasch analysis offers a method of ensuring that key
measurement assumptions are tested and, where
data fit the model, arithmetic operations may be
undertaken. It has particular value in the develop-
ment of new measures, specifically in guiding item
reduction. Traditional methods of item reduction
that rely on item–total correlations and/or indices of
internal consistency can have unfortunate effects on
the sensitivity of measures and their ability to pro-
vide valid scores at the extremes of the construct
range. This is because items at the extreme of the
measurement range are generally discarded because
too many or too few respondents affirm them. In
reality, these “extreme” items may be the most
important in a scale—extending its range of cover-
age of the construct.
References
1 Wright BD. Common sense for measurement.
Rasch Meas Trans 1999;13:704–5.
2 Rasch G. Probabilistic Models for Some Intelli-
gence and Attainment Tests. Chicago: University of
Chicago Press, 1960 (Reprinted 1980).
3 Streiner D, Norman G. Health Measurement
Scales. Oxford: Oxford University Press, 1989.
4 McKenna SP, Whalley D, Cook S. Improving the
sensitivity of the Quality of Life in Depression
Scale (QLDS). Qual Life Res 2002;11:625.
5 Ellis B. Basic Concepts in Measurement. Cam-
bridge: Cambridge University Press, 1966.
6 Svensson E. Guidleines to statistical evaluation of
data from rating scales and questionnaires. J Reha-
bil Med 2001;33:47–8.
7 Kaziz L, Anderson JJ, Meenan RF. Effect sizes for
interpreting changes in health status. Med Care
1989;27:S178–S189.
8 Andrich D. Rasch Models for Measurement.
Series: Quantitative Applications in the Social Sci-
ences No. 68. London: Sage Publications, 1988.
9 Embretson SE, Reise SP. Item Response Theory for
Psychologists. NJ: Lawrence Erlbaum, 2000.
10 Luce RD, Tukey JW. Simultaneous conjoint meas-
urement: a new type of fundamental measurement.
J Math Psychol 1964;1:1–27.
11 Rasch G. On general laws and the meaning of
measurement in psychology. In: Neyman J, ed.,
Proceedings of the Fourth Berkeley Symposium on
Mathematical Statistics and Probability, IV. Berke-
ley CA: University of California Press, 1961.
12 Wright BD, Masters GN. Rating Scale Analysis.
Chicago: MESA Press, 1982.
Rasch Analysis in QoL Instrument Development
13 Van Alphen A, Halfens R, Hasman A, Imbos T.
Likert or Rasch? Nothing is more applicable than a
good theory. J Adv Nurs 1994;20:196–201.
14 Smith RM. Fit analysis in latent trait measurement
models. J Appl Meas 2000;2:199–218.
15 Holland PW, Wainer H, eds. Differential Item
Functioning. Mahwah, NJ: Lawrence Erlbaum
Associates, 1993.
16 Smith RM. Applications of Rasch Measurement.
Sacramento: JAM Press, 1992.
17 Cella DF, Lloyd SR, Wright BD. Cross-cultural
instrument equating: current research and future
directions. In: Spilker B, ed., Quality of Life and
Pharmacoeconomics in Clinical Trials 2nd ed.
Philadelphia: Lippincott-Raven Publishers, 1996.
18 Bond TG, Fox CM. Applying the Rasch Model:
Fundamental Measurement for the Human Sci-
ences. Mahwah, NJ: Lawrence Erlbaum Associ-
ates, Inc., 2001.
19 Smith EV Jr. Evidence for the reliability of meas-
ures and validity of measure interpretation: a
Rasch measurement perspective. J Appl Meas
2001;2:281–311.
20 Birnbaum A. Some latent trait models and their use
in inferring an examinee’s ability. In: Lord FM,
Novick MR, eds., Statistical Theories of Mental
Test Scores. Reading, MA: Addison-Wesley, 1968.
21 Van der Linden WJ, Hambleton RK, eds. Hand-
book of Modern Item Response Theory. New
York: Springer, 1997.
22 Perline R, Wright BD, Wainer H. The Rasch model
as additive conjoint measurement. Appl Psychol
Meas 1979;3:237–56.
23 Silverstein B, Kilore KM, Fisher WP, et al. Apply-
ing psychometric criteria to functional assessment
in medical rehabilitation: I. Exploring unidimen-
sionality. Arch Phys Med Rehabil 1991;72:631–7.
24 Haley SM, McHorney CA, Ware JE Jr. Evaluation
of the MOS SF-36 Physical Functioning Scale (PF-
10): I. Unidimensionality and reproducibility of the
Rasch item scale. J Clin Epidemiol 1994;47:671–
84.
25 Shulman JA, Wolfe EW. Development of a nutri-
tion self-efficacy scale for prospective physicians. J
Appl Meas 2000;1:107–30.
26 Doward LC, McKenna SP, Kohlmann T, et al. The
international development of the RGHQoL: a
quality of life measure for recurrent genital herpes.
Qual Life Res 1998;7:143–53.
27 McKenna SP, Whalley D, Renck-Hooper U, et al.
The development of a quality of life instrument for
use with post-menopausal women with urogenital
atrophy in the UK and Sweden. Qual Life Res
1999;8:393–8.
28 McKenna SP, Doward LC, Alonso J, et al. The
QoL-AGHDA: an instrument for the assessment of
quality of life in adults with growth hormone defi-
ciency. Qual Life Res 1999;8:373–83.
S25
29 Whalley D, McKenna SP, Dewar AL, et al. A new
instrument for assessing quality of life in atopic
dermatitis: International Development of the Qual-
ity of Life Index for Atopic Dermatitis (QoLIAD).
Br J Dermatol 2004;150:274–83.
30 Whalley D, McKenna SP, Dewar AL, et al. Quality
of life in adults with atopic dermatitis—the inter-
national development of the QoLIAD. Qual Life
Res 2000;9:322.
31 McKenna SP, Cook SA, Whalley D, et al. Devel-
opment of the PSORIQoL, a psoriasis-specific
measure of quality of life designed for use in clin-
ical practice and trials. Br J Dermatol 2003;149:
323–31.
32 Whalley D, McKenna SP, Dewar AL, et al. Inter-
national development of a measure to assess qual-
ity of life in childhood atopic dermatitis—the
PIQOL-AD. Qual Life Res 2000;9:302.
33 Dobby J, Duckworth D. Objective Assessment by
Means of Item Banking. Schools Council Exami-
nation Bulletin 40. London: Evans/Methuen Edu-
cational, 1979.
34 Revicki DA, Cella DF. Health status assessment for
the twenty-first century: item response theory, item
banking and computer adaptive testing. Qual Life
Res 1997;6:595–600.
35 Wainer H. Computerized Adaptive Testing (2nd
ed.). Mahwah, NJ: Lawrence Erlbaum Associates,
2000.
36 McKenna SP, Doward LC, Whalley D, et al. The
development of the PsAQoL: a quality of life
instrument specific to Psoriatic Arthritis. Ann
Rheum Dis 2004;63:162–9.
37 Doward LC, Whalley D, Dewar AL, et al. The
development of the SLE-QoL: a quality of life
instrument specific to Systemic Lupus Erythemato-
sus. Qual Life Res 1999;8:609.
38 Doward LC, Spoorenberg A, Cook SA, et al. The
development of the ASQoL: a quality of life instru-
ment specific to Ankylosing Spondylitis. Ann
Rheum Dis 2003;62:20–6.
39 Cook SA, Whalley D. Looking for common
ground: a first step towards comparing quality of
life across diseases. Proc Br Psychol Soc 2001;9:64.
40 Wirén L, Whalley D, McKenna SP, Wilhelmsen
L. Application of a disease-specific, quality-of-
life measure (QoL-AGHDA) in growth hor-
mone-deficient adults and a random population
sample in Sweden: validation of the measure by
Rasch analysis. Clin Endocrinol (Oxf)
2000;52:143–5.
41 Rapp SR, Feldman SR, Exum ML, et al. Psoriasis
causes as much disability as other major medical
diseases. J Am Acad Dermatol 1999;41:401–7.
42 Nichol MB, Margolis JE, Lippa E, et al. The appli-
cation of multiple quality of life instruments in
individuals with mild-to-moderate psoriasis. Phar-
macoeconomics 1996;10:644–53.
S26
43 Merbitz C, Morris J, Grip JC. Ordinal scales and
foundations of misinference. Arch Phys Med Reha-
bil 1989;70:308–12.
44 Wright BJ, Linacre JM. Observations are always
ordinal; measurements, however, must be interval.
Arch Phys Med Rehabil 1989;70: 857–60.
Tennant et al.
45 Streiner DL, Norman GR. Health Measurement
Scales: a Practical Guide to Their Development and
Use 2nd ed. Oxford: Oxford University Press,
1995.