The Impact of Vision Impairment Questionnaire: An
Assessment of Its Domain Structure Using Confirmatory
Factor Analysis and Rasch Analysis
Ecosse L. Lamoureux,1Julie F. Pallant,2Konrad Pesudovs,3Gwyn Rees,1
Jennifer B. Hassell,1and Jill E. Keeffe1,4
PURPOSE. To assess and validate the subscale structure of the
28-item Impact of Visual Impairment (IVI) Scale by using con-
firmatory factor analysis (CFA) and Rasch analysis for use as an
METHODS. Three hundred nineteen participants completed the
IVI questionnaire, and the responses then were subjected to
Rasch analysis by RUMM2020 software. With the person esti-
mates for each item, CFA was used to assess two hypothesized
structures: three-and four-factor models. The subscales of the
model with the best fit were then further validated by Rasch
RESULTS. CFA supported a three-factor model that included
items from the emotional well-being, reading and accessing
information, and mobility and independence subscales. Almost
all the selected goodness-of-fit statistics for the three-factor
model were better than the recommended values. The factor
loadings of the items on their respective domains were all
statistically significant (P ? 0.001) and ranged between 0.54
and 0.81. The three subscales individually fitted the Rasch
model according to the item–trait interaction test (mobility and
independence ?2[df] ? 45.9 , P ? 0.39; emotional well-
being ? 28.4 , P ? 0.65; and reading and accessing infor-
mation ? 43.5 , P ? 0.18). The item-fit residuals values of
the three subscales were ?2.5 and showed mean and standard
deviations approximating 0 and 1, respectively. The internal
consistency reliability of the subscales (?) was substantial,
ranging between 0.89 and 0.91.
CONCLUSIONS. An examination of the IVI dimension confirmed a
three-subscale structure that displays interval measurement
characteristics likely to provide a valid and reliable assessment
of restriction of participation. The findings provide an oppor-
tunity for a more detailed measurement of the effects of differ-
ent types of low-vision rehabilitation programs. (Invest Ophthal-
mol Vis Sci. 2007;48:1001–1006) DOI:10.1167/iovs.06-0361
termine the outcome of low-vision rehabilitation on quality of
life in people with low vision.1–5The IVI has been validated
and shows good discriminative ability and reliability, with
consistent results found between different forms of adminis-
tration.1It has a good range of items and, as opposed to most
vision-specific questionnaires which typically assess visual
functioning, has been designed to assess restriction of partici-
pation in daily living as well as to provide a tool to determine
the effectiveness of low-vision rehabilitation. The original IVI
questionnaire contained 32 items grouped under five domains
of participation: leisure and work, consumer and social inter-
action, household and personal care, mobility, and emotional
reaction to vision loss.
The grouping of items within domains is important because
they can form subscales that allow for the assessment of inter-
vention at more specific levels. This is particularly relevant to
low-vision care, as management is typically undertaken on a
task-specific basis, and so it is possible that real gains in one or
two areas may be obscured within a large scale assessing
overall performance. Subscale measurements of outcome pro-
vide a more detailed insight into the effectiveness aspects of
low-vision rehabilitation. In the initial validation of the IVI
questionnaire, principal components analysis (PCA) was used
to explore the underlying structure of the scale, but no Rasch
analysis was undertaken, making the findings essentially infer-
ential.1However, the analysis did not confirm the five domains
specified a priori by the authors. Rather, it identified a three-
factor solution that was supported by an examination of the
scree plot. The authors did not either formally reject the five
domains identified a priori or recommend the three-factor
structure identified by PCA. The equivocal nature of the factors
underlying the IVI may confuse potential users in how best to
interpret the results and indicates that further scrutiny of IVI
structure is needed.
Further examination of the domain structure is also war-
ranted, as the IVI has recently undergone further validation
using Rasch analysis which is a sophisticated approach to
questionnaire development using modern psychometric meth-
ods.6Rasch analysis converts categorical data into a linear
scale, calculates item difficulty in relation to patient ability, and
provides estimates of item and person measures on an interval
scale.7–15Rather than an exploratory approach as used in the
initial study (i.e., PCA),1the IVI subscale needs to be validated
using a confirmatory process, especially if the outcome of
low-vision rehabilitation and the sensitivity to change of spe-
cific aspects of quality of life (i.e., mobility, emotion, and
leisure) are desired. The main objectives of this study were
therefore to (1) assess the dimensions of the 28-item IVI using
CFA performed on person-item measures derived from the
Rasch analysis of the instrument and (2) to use Rasch analysis
to validate the factors of the best-fitting model as viable traits
for measuring specific aspects of restriction of participation for
people with impaired vision.
he Impact of Vision Impairment (IVI) questionnaire was
designed to assess participation in daily activities and de-
From the1Centre for Eye Research Australia, The University of
Melbourne, Melbourne, Victoria, Australia; the2Swinburne University
of Technology, Hawthorn, Victoria, Australia; the3NH&MRC (National
Health and Medical Research Council) Centre for Clinical Eye Re-
search, Flinders University and Flinders Medical Centre, Adelaide,
South Australia, Australia; and the4Vision CRC (Correction Research
Center), Sydney, Australia.
ELL is a recipient of a Australian National Health and Medical
Research Council Public Health Fellowship.
Submitted for publication April 3, 2006; revised September 20,
2006; accepted January 5, 2007.
Disclosure: E.L. Lamoureux, None; J.F. Pallant, None; K. Pesu-
dovs, None; G. Rees, None; J.B. Hassell, None; J.E. Keeffe, None
The publication costs of this article were defrayed in part by page
charge payment. This article must therefore be marked “advertise-
ment” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Corresponding author: Ecosse L. Lamoureux, Centre for Eye Re-
search Australia, Department of Ophthalmology, University of Mel-
bourne, 32 Gisborne St., East Melbourne Victoria, 3002, Australia;
Investigative Ophthalmology & Visual Science, March 2007, Vol. 48, No. 3
Copyright © Association for Research in Vision and Ophthalmology
MATERIALS AND METHODS
Individuals referred to low-vision rehabilitation centers across the state
of Victoria (Australia) were recruited. An ophthalmologist’s report,
providing the cause of vision loss and visual acuities, was obtained for
each participant. The eligibility criteria for the study included present-
ing visual acuity ?6/12 (or ?6/12 with restricted fields), ?18 years of
age and the ability to converse in English. Individuals who agreed to
participate signed a consent form that allowed access to low-vision
rehabilitation records. Ethical approval was obtained from the Royal
Victorian Eye and Ear Hospital’s Human Research and Ethics Commit-
tee, and the research adhered to the tenets of the Declaration of
The IVI questionnaire, sociodemographic, and clinical data were col-
lected. Participants also completed the SF-12 from which the physical
and mental component summaries (PCS-12 and MCS-12, respectively)
were computed.16Each summary scale was scored from 0 to 100,
where a score of 100 indicates the best possible score and 0 represents
the worst possible score. The PCS-12 and MCS-12 scores were in-
cluded, to validate the IVI subset of items.
The IVI Questionnaire
The 32-item IVI instrument was either self- or interviewer-adminis-
tered, to measure vision-related restriction of participation in daily
living, as described previously.1,2Responses to the IVI items were
rated on a five-category Likert scale: not at all, 0; hardly at all, 1; a little,
2; a fair amount, 3; a lot, 4; and can’t do because of eyesight, 5; with
an additional response category, don’t do because of other reasons, for
19 items. The latter response was not included in computing the
average overall or domain score. The wording preceding these items
was, In the past month, how much has your eyesight interfered with
the following activities. For the remaining 13 items, the rating scale
used was: not at all, 0; very rarely, 1; a little of the time, 2; a fair amount
of the time, 3; a lot of the time, 4; and all the time, 5. The wording
preceding these items was, In the past month, how often has your
eyesight made you concerned or worried about the following.
Most of the Rasch analyses were performed with RUMM202017but
Winsteps (ver. 3.61)18was used to generate transformed individual
person scores for all items, as this feature is not currently available in
RUMM. Individual person measures for all items in Winsteps were
estimated by assigning a category threshold for each item and convert-
ing raw scores to Rasch category estimates. These data were required
for CFA. The compatibility of the results from the two Rasch analysis
software programs was tested by comparing person measures. These
results were identical with one decimal place for 94% of cases, with
greater deviation only occurring for extreme responders. This deviation
arises from the different assumptions used to calculate extreme results.
Confirmatory factor analysis (performed with AMOS, ver. 6; SPSS
Science, Chicago, IL) was used to confirm the hypothesized structure
statistically. CFA allows for assessment of the overall model fit, the
statistical significance tests for theorized relations in the model, and
the estimation of latent concepts free of measurement error. CFA was
undertaken to assess two hypothesized measurement models based on
the findings of previous exploratory investigation1,2and similar
work.19The first, a three-factor model, comprised three latent traits:
mobility and independence (11 items), emotional well-being (8 items),
and reading and access to information (9 items). The second was a
four-factor model that assessed the interrelationship of four latent
traits: mobility and safety (6 items), emotional well-being (8 items),
independence (8 items), and reading and near-vision activities (6
CFA with the maximum-likelihood estimation was conducted on
the calibrated person-item scores to evaluate model fit of each pro-
posed model. A good model fit can be indicated by a nonsignificant
item–trait interaction ?2probability value. However, because the ?2
test has been criticized for its dependence on sample size, a range of
fit statistics were assessed. A relative ?2is usually used (ratio of ?2to
degrees of freedom- ?2/df) with a recommended range of 1.0 to 2.0.20
The root mean square error of approximation (RMSEA) is the one of
the most informative statistics in determining model fit, as it takes into
account the number of variables that are estimated in the model.21–23
RMSEA values are required to be ?0.05 to indicate good fit. Values
between 0.05 and 0.08 indicate reasonable fit.21–23For the incremental
fit statistics (goodness of fit index: GFI; the Tucker-Lewis index: TLI;
and the comparative fit index: CFI) ?0.90 indicates lack of fit, between
0.90 and 0.95 indicates reasonable fit, and between 0.95 and 1.00
indicates good fit.21–23
The latent variables or subscales of the model identified by CFA as
providing the best fit were then examined with Rasch analysis using
RUMM17with the purpose of assessing how well the subscales fit the
Rasch model. Fit was evaluated by using person and item fit residual
statistics, which are transformed weighted mean squares. The trans-
formed mean squares are normally distributed with an expected value
of 0 and an expected variance of 1. An item–trait interaction score (?2)
with a statistically nonsignificant probability (P ? 0.05) indicates fit to
the model. An estimate of person separation reliability which indicates
how well the items of the instrument separate or spread out the
subjects in the sample was also reported.24The unidimensionality of
the each subscale after overall fit to the Rasch model was determined
using principal components analysis of the residuals available in
RUMM. Unidimensionality is tested by allowing the pattern of factor
loadings on the first residual to determine subsets of items. If person
estimates derived from these subsets of items differ significantly from
the estimates derived from the full subscale, local independence is
considered to be compromised.25
The personal and clinical characteristics of the 319 participants
of the study appear in Table 1. The mean PCS-12 and MCS-12
components of the SF-12 were 36.7 and 47.7, respectively, and
were comparable to mean scores reported for Americans of
similar age groups (38.7 and 50.0 for PCS-12 and MCS-12,
respectively).16Compared with the initial validation study,1
the two samples were similar on personal and clinical charac-
teristics, except for distance visual acuity and main cause of
vision loss. In the present study, there was a greater percentage
of participants with distance visual acuity ?6/18 to 6/60 (47%
vs. 29%) and age-related macular degeneration (54% vs. 39%)
compared with the initial study.
Confirmatory Factor Analysis
Before undertaking CFA, the IVI questionnaire was re-exam-
ined by using the Rasch analysis. This resulted in a 28-item
questionnaire and a four-category response scale for 26 items
and a three-category response scale for two items. The revised
questionnaire had an excellent internal consistency, as illus-
trated by the following statistics: Item–trait interaction ?2?
118, P ? 0.32; mean ? SD person fit residual values ? 0.068 ?
0.85; mean ? SD item fit residual values ? ?0.203 ? 1.45 and
person separation reliability ? 0.95. Rasch calibrated person
measures across all items were then generated and CFA was
used to assess the hypothesized models. The various goodness-
of-fit statistics for the two hypothesized models are shown in
Table 2. The indices showed a good fit between the IVI data
and proposed measurement models. The fit indices for these
two models were almost identical (Table 2). The ratio of ?2to
degrees-of-freedom value (?2/df) is 1.41 in both models and
1002 Lamoureux et al.
IOVS, March 2007, Vol. 48, No. 3
falls well within the recommended range of 1.0 to 2.0.20The
RMSEA, CFI, and TLI in both models averaged 0.5, 0.94, and
0.94, respectively, and were better than the recommended
values. For both models, the goodness-of-fit index (GFI) of 0.85
was slightly less than the benchmark of 0.9.
Inspection of the correlations between the four-factor solu-
tion, however, indicated there were high values between the
independence and reading subscales (0.91) as well as between
the independence and mobility subscales (0.92). Given the
evidence of strong overlap between these latent traits and in
the interest of model parsimony, the best-fitting model for
these data was a three-correlated-factor model (Table 3) repre-
senting latent trails of mobility and independence (11 items),
reading and accessing information (9 items), and emotional
well-being (8 items). The ? coefficients of the 28 items were all
statistically significant (P ? 0.001) and ranged between 0.62
and 0.77, 0.65 and 0.81, and 0.54 and 0.78 for the mobility and
independence, emotional well-being and reading and accessing
information subscales, respectively. The interfactor correla-
tions were 0.85 (reading and accessing information and mobil-
ity and independence), 0.61 (reading and accessing informa-
tion and emotional well being), and 0.80 (mobility and
independence and emotional well being). The items—Finding
what you want during shopping? (? ? 0.80); Interfered with
your life in general? (? ? 0.81), and Getting about outdoors?
(? ? 0.77)—showed the strongest loading for the mobility and
independence, emotional well-being and reading and accessing
information subscales, respectively.
Fit of the Three Subscales to the Rasch Model
Subsequent to CFA supporting the three-factor model, its three
domains were assessed separately by Rasch analysis. They all
showed nonsignificant item–trait interactions, suggesting fit
between data and model (mobility and independence ?2[df] ?
45.9, P ? 0.39; emotional well-being ? 28.4 , P ?
0.65; and reading and accessing information ? 43.5 , P ?
0.18). The fit residuals of all the items recorded values ?2.5
and the three subscales showed mean and SD values close to 0
and 1, respectively, suggesting no misfit to the model by items
and respondents. The person separation reliability scores
ranged between 0.89 and 0.91, indicating a substantial ability
to distinguish four strata of person ability (Table 4). The most
difficult and easiest items (in logits) for the mobility and inde-
pendence subscale were: Stopped you doing the things you
want to do? (0.75) and Your general safety at home? (?1.42);
for emotional well-being: Felt frustrated or annoyed? (1.1) and
Have you felt lonely or isolated? (?1.42); and for the reading
and accessing information: Reading ordinary-sized print?
(1.92), and Generally looking after your appearance? (?1.13).
Test of Local Independence Assumption of the
No significant differences were found between the person
estimates of the three subscales and their respective positive
and negative subsets (t-test; P ? 0.45–0.69). This finding sug-
gests no breach of the assumption of local independence
supporting the unidimensionality of the subscales.
After Rasch analysis, the person measures of each subscale
were used to assess the criterion validity of the subscales of the
IVI. The reading and accessing information subscale recorded
its strongest correlations with visual acuity, as these activities
are critically dependent on near and distance vision (Table 5).
TABLE 1. The Characteristics of the 319 Study Participants
Mean ? SD
Presenting visual acuity
? 6/12 with field loss
?6/12 to 6/18
?6/18 to 6/60
N8 or better
Main cause of vision loss
Age-related macular degeneration
Duration of vision impairment (y)
Median (min, max)
Comorbidity affects daily living?
Not at all
A great deal
Mean ? SD
Mean ? SD
78.4 ? 12.9
36.7 ? 11.6
47.7 ? 11.5
The summary scale ranges between 0 and 100 where 0 and 100
indicate the best and worst possible scores, respectively. PCS-12,
physical component summary of the SF-12; MCS-12, mental compo-
nent summary of the SF-12.
TABLE 2. Comparative Goodness-of-Fit Statistics for Two Hypothesized Models
Root mean square error of approximation (RMSEA)
Goodness of fit index (GFI)
Comparative fit index (CFI)
Tucker-Lewis Index (TLI)
IOVS, March 2007, Vol. 48, No. 3
Domain Structure Assessment of the IVI Questionnaire1003
The emotional well-being factor recorded its strongest associ-
ation with the mental component of the SF-12 (?0.56) which
includes items pertinent to emotional and mental health.
Equally, the mobility and independence domain recorded its
strongest association (?0.43) with the physical component of
the SF-12 health (PCS-12) which includes items associated with
mobility. These correlations overall tend to support the defini-
tion of the new IVI subscale structure.
Internal Consistency Reliability
The internal consistency reliability of the three-factor model
was estimated by Cronbach’s ? using the person measures
across items. The mobility and independence, emotional well-
being, and reading and accessing information subscales re-
corded ? values ranging between 0.89 and 0.91. These values
are above the suggested moderate level of internal consistency
among the instrument items26and indicate that the items
under each subscale consistently measure the same construct.
Scoring of the IVI Subscales
Other investigators wishing to use the IVI subscales can use
these validation data to convert raw scores into Rasch person
measures without having to perform Rasch analysis. This con-
version mainly holds for patients with complete data. Raw
scores are calculated by first reversing the scores (0, 1, 2, 3, 4,
5) (5, 4, 3, 2, 1, 0) to give better IVI scores to those experi-
encing less restriction of participation. The categories are then
collapsed to 4(3, 2, 2, 1, 1, 0) or 3(2, 1, 1, 1, 1, 0), as described
previously.6Then, for each subscale the average of the items
gives the IVI raw score. This score is related to the IVI Rasch
person measure, as illustrated in Figure 1. The relationship is
double-asymptotic because the average raw rating has a floor
and a ceiling (at 0 and 3). The relationship can be described as
double-asymptotic nonlinear regression.27The equations listed
in Table 6 can be used to convert raw scores to Rasch person
measures for each subscale.
In the present study, CFA was used to assess the dimension of
the IVI and provide evidence of the factorial structure and
reliability of the recently Rasch-scaled IVI instrument.6Two
models were proposed a priori and a resultant three-factor
model provided the best-fit statistics and most parsimonious
TABLE 3. The Item Labels and Factor Loadings of the Three-Factor Model of the IVI Questionnaire
1. Taking part in sporting activities?
2. Visiting friends or family?
3. Getting about outdoors?
4. Made you go carefully to avoid falling or tripping?
5. Interfered with travelling or using transport?
6. Going down steps, stairs, or curbs?
7. Your general safety at home?
8. Spilling or breaking things?
9. Your general safety when out of your home?
10. Stopped you doing the things you want to do?
11. Needed help from other people?
12. Felt embarrassed?
13. Felt frustrated or annoyed?
14. Have you felt lonely or isolated?
15. Have you felt sad or low?
16. Worried about your eyesight getting worse?
17. Concerned or worried about coping with
18. Interfered with your life in general?
19. Felt like a nuisance or a burden?
20. Ability to see and enjoy television?
21. Finding things during shopping?
22. Reading ordinary-sized print?
23. Recognizing or meeting people?
24. Getting information that you need?
25. Generally looking after your appearance?
26. Opening packaging?
27. Reading labels or instructions on medicines?
28. Operating household appliances and the
TABLE 4. The Mean ? SD Person and Item Fit Residual Values, Person Separation Reliability Scores, and
Range of Item Fit Residual Values of the Three Domains of the IVI
Mean ? SD
Mean ? SD
Mobility and independence
Reading and accessing information
?0.31 ? 1.23
?0.28 ? 1.10
?0.41 ? 0.91
0.08 ? 0.99
0.06 ? 1.31
?0.42 ? 0.71
1004 Lamoureux et al.
IOVS, March 2007, Vol. 48, No. 3
measure of restriction of participation in quality of life in
patients with low vision. The confirmed dimensions of the IVI
represent items from the emotional well-being, reading and
accessing information, and mobility and independence, respec-
tively. Our finding supports the recent work of Massof et al.,19
who also used person measures estimated from subsets of
functionally grouped items to assess the dimensionality of a
vision disability questionnaire. They confirmed a two-factor
structure: reading and mobility. Massof et al. did not find an
emotional well-being subscale, as their questionnaire, the Ac-
tivity Breakdown Structure (ABS), did not have such content,
because it assesses the difficulty undertaking daily tasks. On
the other hand, a critical component of the IVI is the assess-
ment of the emotional impact of visual impairment in people
with low vision. With the confirmation of an emotional sub-
scale, it is now possible to provide a more specific assessment
of the emotional consequences associated with vision loss as
well as to evaluate the effect of rehabilitation programs on
Only one factor identified a priori in the initial validation
study1(i.e., emotional well-being) was confirmed in the study.
Discrepancy between the initial and current studies could be
linked to different factor analysis techniques. The initial study
used exploratory methods, whereas the current one used a
confirmatory approach. Also, in the present study, Rasch-cali-
brated person measures were used compared with raw scores
used in the initial study. Factor analysis of raw scores can lead
to item grouping based on item difficulty. The difference in
sample size could also explain the dissimilarity between the
two studies. Only 86 participants were included initially, com-
pared with 319 in the present study. It is likely that a bigger
sample size provided a better capacity to undertake factor
analysis, as it has been suggested that small data sets tend not
to generalize as well as those derived from large samples.
Tabachnick and Fidell28have suggested that a ratio of five cases
to one item is adequate. In the present study, our ratio was
10.3 compared with 3.1 in the initial study.
In this study, the use of Rasch analysis has enabled a de-
tailed examination of the operation of the subscales. Our find-
ings show that the three domains possess viable measurement
characteristics to assess specific aspects of restriction of par-
ticipation in daily activities for individuals with impaired vi-
sion. The domains also possess demonstrated reliability and
validity and show no evidence of multidimensionality. Because
it had been shown that the Rasch scoring method had greater
precision compared with standard Likert scoring and plays an
important role in improving sensitivity to change,14,29,30our
findings suggest that the greater accuracy of the Rasch-ana-
lyzed subscales could result in improved measurement of the
specific outcomes of low-vision rehabilitation trials.30,31Future
studies, however, are needed to substantiate this claim.
One important finding of the new IVI domain structure is
the identification of a reading and accessing information do-
main. Activities related to near and distance vision have con-
sistently been associated with increased difficulty for people
with low vision8,32,33and with this revised IVI domain struc-
ture, the impact of these critical activities of daily living could
be assessed individually and collectively. Of importance, a
significant component of most low-vision rehabilitation pro-
grams includes the prescription of low-vision devices as well as
strategies intended to improve visual functioning, and the
Rasch-assessed domain structure of the IVI can now potentially
assess the outcome of low-vision rehabilitation specific to read-
ing and ability to undertake vision-dependent activities.
Finally, moderate correlations were found between the
mental component of the SF-12 and the emotional well-being
subscale; the physical component of the SF-12 and mobility
and independence subscale; and visual acuity and the reading
and access to information subscale. This finding provides fur-
ther support of the new factorial model of the IVI, as the
MCS-12 and PCS-12 contain several items specific to emotional
well-being and mobility, respectively, and visual acuity func-
tions are critical to distance reading and near vision perfor-
In conclusion, through a confirmatory factor analysis and
Rasch analysis, our examination of the dimensionality of the IVI
questionnaire supported a three-subscale structure with inter-
val level measurement characteristics likely to provide a reli-
able assessment of specific aspects of restriction of participa-
tion in daily living and effectiveness of rehabilitation in people
with low vision. This new structure of the IVI opens the door
to the exploration of three components of restriction of par-
ticipation in daily living and a better understanding of the
score). The fit lines are generated with double asymptotic nonlinear regression. (A) Mobility and independence, (B) Emotional well-being, and (C)
Reading and accessing information.
Scatterplots of the person measure estimated from Rasch analysis versus the average rating for each person across items (raw subscale
TABLE 5. Correlation between the Participants’ Characteristics and
the Person Measures of the New IVI Subscales
* P ? 0.05.
† P ? 0.001.
IOVS, March 2007, Vol. 48, No. 3
Domain Structure Assessment of the IVI Questionnaire1005
effects of different types of intervention. Future work should Download full-text
evaluate the sensitivity of the IVI subscales to measure out-
comes of low-vision rehabilitation.
1. Weih LM, Hassell JB, Keeffe JE. Assessment of the impact of vision
impairment. Invest Ophthalmol Vis Sci. 2002;43:927–935.
2. Lamoureux EL, Hassell JB, Keeffe JE. The determinants of partici-
pation in activities of daily living in people with impaired vision.
Am J Ophthalmol. 2004;137:265–270.
3. Lamoureux EL, Hassell JB, Keeffe JE. The impact of diabetic reti-
nopathy on participation in daily living. Arch Ophthalmol. 2004;
4. Hassell JB, Lamoureux EL, Keeffe JE. Impact of age-related macular
degeneration on quality of life. Br J Ophthalmol. 2006;90:593–
5. Noe ¨ G, Ferraro J, Lamoureux E, Rait J, Keeffe JE. Associations
between glaucomatous visual field loss and participation in activ-
ities of daily living. Clin Exp Ophthalmol. 2003;31:482–486.
6. Lamoureux E, Pallant J, Pesudovs K, Hassell J, Keeffe J. The Impact
of Vision Impairment Questionnaire: an evaluation of its measure-
ment properties using Rasch analysis. Invest Ophthalmol Vis Sci.
7. Norquist JM, Fitzpatrick R, Dawson J, Jenkinson C. Comparing
alternative Rasch-based methods vs raw scores in measuring
change in health. Med Care. 2004;42:125–136.
8. Pesudovs K, Garamendi E, Keeves JP, Elliott DB. The Activities of
Daily Vision Scale for cataract surgery outcomes: re-evaluating
validity with Rasch analysis. Invest Ophthalmol Vis Sci. 2003;44:
9. Fisher WP Jr, Eubanks R, Marier RL. Equating the MOS SF36 and
the LSU HSI Physical Functioning Scales. J Outcome Meas. 1997;
10. Fisher WPJ. The Rasch debate: validity and revolution in educa-
tional measurement. In: Wilson M, ed. Objective Measurement:
Theory into Practice. Norwood, NJ: Ablex; 1994:36–72.
11. Massof RW. The measurement of vision disability. Optom Vis Sci.
12. Wright BD, Linacre JM. Observations are always ordinal, measure-
ments must be interval. Arch Phys Med Rehabil. 1989;70:857–
13. White LJ, Velozo CA. The use of Rasch measurement to improve
the Oswestry classification scheme. Arch Phys Med Rehabil. 2002;
14. Raczek AE, Ware JE, Bjorner JB, et al. Comparison of Rasch and
summated rating scales constructed from SF-36 physical function-
ing items in seven countries: results from the IQOLA Project.
International Quality of Life Assessment. J Clin Epidemiol. 1998;
15. Pesudovs K, Garamendi E, Elliott DB. The Quality of Life Impact of
Refractive Correction (QIRC) Questionnaire: development and val-
idation. Optom Vis Sci. 2004;81:769–777.
16. Ware J, Kosinski M, Keller S. SF-12: How to Score the SF-12
Physical and Mental Health Summary Scales. Lincoln, RI: Quality
17. Andrich D, Lyne A, Sheridan B, Luo G. RUMM 2020. Perth,
Australia: RUMM Laboratory; 2003.
18. Linacre J. A User’s Guide to Winsteps: Rasch-Model Computer
Program. Chicago: Mesa Press; 2002.
19. Massof RW, Hsu CT, Baker FH, et al. Visual disability variables. II:
The difficulty of tasks for a sample of low-vision patients. Arch
Phys Med Rehabil. 2005;86:954–967.
20. Hair JF, Black WC, Babin BJ, Anderson RE, Tatham RL. Multivariate
Data Analysis. 6th ed. Englewood Cliffs, NJ: Prentice Hall; 2005.
21. Byrne B. Structural Equation Modeling with AMOS: Basic Con-
cepts, Applications, and Programming. Mahwah, NJ: Lawrence
Erlbaum Associates; 2001.
22. Bentler PM. Comparative fit indexes in structural models. Psychol
23. Kline R. Principles and Practice of Structural Equation Modeling.
2nd ed. New York: Guilford Press; 2005.
24. Mallinson T, Stelmack J, Velozo C. A comparison of the separation
ratio and coefficient alpha in the creation of minimum item sets.
Med Care. 2004;42:117–124.
25. Smith EV. Detecting and evaluation the impact of multidimension-
ality using item fit statistics and principal component analysis of
residuals. J Appl Meas. 2002;3:205–231.
26. Nunnally J. Psychometric Theory. New York: McGraw-Hill; 1978.
27. Massof RW. Application of stochastic measurement models to
visual function rating scale questionnaires. Ophthalmic Epide-
28. Tabachnick B, Fidell L. Using Multivariate Statistics. 4th ed. New
York: HarperCollins; 2001.
29. McHorney CA, Haley SM, Ware JE, Jr. Evaluation of the MOS SF-36
Physical Functioning Scale (PF-10): II. Comparison of relative pre-
cision using Likert and Rasch scoring methods. J Clin Epidemiol.
30. Fitzpatrick R, Norquist JM, Jenkinson C, et al. A comparison of
Rasch with Likert scoring to discriminate between patients’ eval-
uations of total hip replacement surgery. Qual Life Res. 2004;13:
31. Hays RD, Morales LS, Reise SP. Item response theory and health
outcomes measurement in the 21st century. Med Care. 2000;38:
32. Hart P, Stevenson M, Montgomery A, Muldrew K, Chakravarthy U.
Further validation of the Daily Living Tasks Dependent on Vision:
identification of domains. Br J Ophthalmol. 2005:1127–1130.
33. Stelmack J, Szlyk J, Stelmack T, et al. Use of Rasch person-item map
in exploratory data analysis: a clinical perspective. J Rehabil Res
TABLE 6. Equations Converting Raw Score to the Rasch Person Measure
Mobility and independence
Reading and accessing information
IVIperson measure? 33.79log(IVIraw score/3-IVIraw score) ? 53.39
IVIperson measure? 33.90log(IVIraw score/3-IVIraw score) ? 50.61
IVIperson measure? 30.50log(IVIraw score/2.8-IVIraw score) ? 52.63
1006 Lamoureux et al.
IOVS, March 2007, Vol. 48, No. 3