Documente Academic
Documente Profesional
Documente Cultură
67]
On: 28 October 2014, At: 03:32
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK
To cite this article: María Gómez-Gallego, Juan Gomez-García & Manuel Ato-García
(2014) Confirmatory Factor Analysis of the Quality of Life in Alzheimer’s Disease Scale
in Patients with Alzheimer’s Disease, Experimental Aging Research: An International
Journal Devoted to the Scientific Study of the Aging Process, 40:3, 266-279, DOI:
10.1080/0361073X.2014.896664
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
Downloaded by [193.147.25.67] at 03:32 28 October 2014
María Gómez-Gallego
Downloaded by [193.147.25.67] at 03:32 28 October 2014
Juan Gomez-García
Manuel Ato-García
Over the past 20 years, the use of quality of life (QoL) measures in
Downloaded by [193.147.25.67] at 03:32 28 October 2014
METHODS
Subjects
The sample was recruited from psychogeriatric clinics, day hospitals, and
day care centers. Patients had received diagnosis of probable AD according
to the criteria of the National Institute of Neurological and Communicative
Disease and Stroke and the Alzheimer’s Disease and Related Disorders
Association (McKhann et al., 1984). Dementia severity was rated with
the Global Deterioration Scale (GDS) (Reisberg, Politis, Lyketsos, &
Mavreas, 1982) and Mini-Mental State Examination (MMSE) (Folstein,
Folstein, & McHugh, 1975). Only patients with GDS score of 3–5 and
MMSE score higher than 11 were recruited because the psychometric suit-
ability of the QoL-AD Scale for more severely impaired patients has not
been established. MMSE scores of 10 or below indicate severe dementia,
whereas scores of 11–20 indicate moderate dementia. This procedure also
Factor Structure of the QoL-AD Scale in AD Patients 269
Measures
In this study, the Spanish version of the QoL-AD Scale was used
(Appendix). This scale consists of 13 items ordered like the original instru-
ment (Logsdon et al., 2002). As Edelman et al. did, Items 7 “marriage”
and 12 “money” were replaced by items “people who live with” and “abil-
ity to decide,” respectively. Contrary to what these authors did, we did
not add any other item. Like the original version, items are scored on
a 4-point multiple-choice options and range of the total score is 13–52.
The scale was translated into Spanish by two professional translators,
who finally synthesized the two versions. Then, this version was trans-
lated into English to ensure that the wording of the Spanish items reflected
the meaning of the original version items. An expert committee (neurolo-
gist, geriatrician, and psychologist) reviewed the final version to check for
content validity.
The Spanish version of the QoL-AD Scale has good internal consis-
tency (Cronbach’s alpha coefficient = .85,) and good test-retest reliability
(intraclass correlation = .87), similarly to those reported by Logsdon et al
(2002). Likewise, our version showed strong convergent validity, since it
was highly correlated to measures of QoL, depression, functional state,
and neuropsychiatric symptoms (Gomez-Gallego et al., 2012).
Statistical Analyses
five items were hypothesized to load on Factor 2 (V4, V6, V7, V8, and
V12), and five items were hypothesized to load on Factor 3 (V3, V5, V9,
V11, and V13).
Although Likert-scaled items are ordinal variables, they are currently
treated as continuous variables in many studies. This procedure is not
statistically risky with items having five or more categories (Zumbo &
Zimmerman, 1993). QoL-AD items have four levels that could not be
equidistant, since only one item can receive a below-average rating and
results could be biased in favor of a positive outcome. Thus, in this study,
QoL-AD items have been treated as ordinal variables. Statistical analysis
of ordinal categorical variables assumes that they are categorized versions
of hypothetical latent variables that are truly continuous and multivariate
Downloaded by [193.147.25.67] at 03:32 28 October 2014
The model fit was evaluated using the χ 2 statistic, the comparative fit
index (CFI), the Tucker-Lewis index (TLI), the incremental fit index (IFI),
the root mean square of approximation (RMSA), and the standardized
root mean square residual (SRMS). The χ 2 statistic evaluates “the mag-
nitude of discrepancy between the sample and fitted covariances matrices”
(Hu & Bentler, 1999). A nonsignificant χ 2 suggests an adequate fit of
the hypothesized model. Relative indices (IFI and TLI) compare a chi-
square for the model tested to one from an independence model (or null
model). Like IFI and TLI, CFI assumes that all latent variables are uncor-
related (independence model) and compares the sample covariance matrix
with this null model (Bentler, 1990). However, CFI considers sample size
and performs well even in small samples. CFI, TLI, and IFI values above
.95 are indicative of good fit (Hu & Bentler, 1999; Schumacker & Lomax,
2004). RMSEA is an index useful for correcting for parsimony, based on
Factor Structure of the QoL-AD Scale in AD Patients 271
RESULTS
Model I
Model II
Some modifications of the Model I structure were necessary to fit the data
adequately. Initially, we systematically removed nonsignificant paths and
272 M. Gómez-Gallego et al.
items. Removing the path from F1 to V11 caused only a small improve-
ment in fit indices (Y-B χ 2 (65) = 84.677, p = .029; F(65, 37) = 3.528, p =
.000; CFI = .957; TLI = .946; IFI = .958; RMSEA = .081 [.052–.107]).
Moreover, 25% of residual values were higher than .10. The deletion of
the path from F3 to V11 led to worse fit to the data than the former.
Then, we used the modification indices to improve the model fit. Firstly,
we introduced the path from F3 to V12 and deleted the path from F2 to
V12 because it was not found significant. Fit indices indicated that the
Model II (Table 2) provided a better fit to the data, with nonsignificant
robust measures (Y-B χ 2 (62) = 68.01, p = .280; Y-B F(62, 40) = 1.36,
p = .152) and all criteria and indices inside the limits. However, more than
20% of correlation residuals were higher than .10. The correlation between
Downloaded by [193.147.25.67] at 03:32 28 October 2014
Model III
Model IV
This model reached also an acceptable fit, with fit indices similar to those
of Model III (Table 2). The mean score was 2.222 (SD = 0.867) for F1 and
3.029 (SD = 0.744) for F2. Neither floor nor ceiling effect was found for
F1 (2.1% and 1.4%, respectively) and for F2 (1.4% and 4.3%, respec-
tively). Cronbach’s α coefficient was .856 for F1 and .518 for F2. The
model values of AIC and CAIC were −36.845 and −268.843, respec-
tively. The comparison between Models III and IV showed that there were
no significant differences between them (χ 2 (2) = 3.28, p = .194).
Downloaded by [193.147.25.67] at 03:32 28 October 2014
Table 1. Confirmatory factor analysis solution for all models of the QoL-AD Scale
Model I Model II Model III Model IV Model V
Items F1 F2 F3 F1 F2 F3 F1 F2 F3 F1 F2 F1
Note. Factor loadings are standardized regression coefficients. All loadings except marked with ∗ were significant at p < .01. F1 =
Factor 1; F2 = Factor 2; F3 = Factor 3.
273
274 M. Gómez-Gallego et al.
p= p = .110 (.030–.092)
.252
V 75.88 1.77 (65, 37) .926 .911 .927 .081 .106
p= p = .031 (.079–.129)
.168
Model V
Finally, a unifactorial solution (Model V, Table 1) was also tested. All the
fit indices indicated a poor model fit (Table 2).
DISCUSSION
to mood, as it was saturated by ability-related items (9, 10, and 11). Item
5 (memory), as a previous study found, had low standardized loadings on
Factor 3 (Revell et al., 2009). This could be because memory problems are
related to disability in AD, but they are not a major source of it, opposed
to executive dysfunction (Cahn-Weiner et al., 2007).
Some changes in the goodness of fit were expected to occur for mod-
els with lower number of factors. In fact, all the fit indices plus AIC and
CAIC were better for Model III than for Model IV. However, chi-square
statistics showed that there were no significant differences between mod-
els. Model IV shows good fit and is more parsimonious than Model III.
Indeed, Factor 1 combines all items of the factors “Perceived functional
ability” and “Perceived health.” This factor called “Perceived behavioral
Downloaded by [193.147.25.67] at 03:32 28 October 2014
REFERENCES
Brod, M., Steward, A. L., Sands, L., & Walton, P. (1999). Conceptualization
and measurement of quality of life in dementia: The Dementia Quality of Life
instrument (DQoL). The Gerontologist, 39, 25–35.
Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York:
The Guilford Press.
Buckley, T., Fauth E. B., Morrison, A., Tschanz, J., Rabins P. V., Piercy, K. W.,
. . . Lyketsos, C. G. (2012). Predictors of quality of life ratings for persons with
dementia simultaneously reported by patients and their caregivers: The Cache
County (Utah) Study. International Psychogeriatrics, 24, 1094–1102.
Byrne, B. M. (2006). Structural equation modeling with EQS. Mahwah, NJ:
Lawrence Erlbaum.
Cahn-Weiner, D. A., Farias, S. T., Julian, L., Harvey, D. J., Kramer, J. H., Reed,
B. R., . . . Chui, H. (2007). Cognitive and neuroimaging predictors of instru-
Downloaded by [193.147.25.67] at 03:32 28 October 2014
APPENDIX