Psychometric properties of the Chalder Fatigue Scale revisited: an exploratory structural equation modeling approach
Psychometric properties of the Chalder Fatigue Scale revisited: an exploratory structural equation modeling approach
Ted C. T. Fong 0 1 2
Jessie S. M. Chan 0 1 2
Cecilia L. W. Chan 0 1 2
Rainbow T. H. Ho 0 1 2
Eric T. C. Ziea 0 1 2
Vivian C. W. Wong 0 1 2
Bacon F. L. Ng 0 1 2
S. M. Ng 0 1 2
0 E. T. C. Ziea V. C. W. Wong B. F. L. Ng Chinese Medicine Department, Hong Kong Hospital Authority , Hong Kong , China
1 J. S. M. Chan C. L. W. Chan R. T. H. Ho S. M. Ng (&) Department of Social Work and Social Administration , Jockey Club Tower, The Centennial Campus , The University of Hong Kong , Pokfulam, Hong Kong , China
2 T. C. T. Fong R. T. H. Ho Centre on Behavioral Health, The University of Hong Kong , Hong Kong , China
Objective Previous validation studies of the Chalder Fatigue Scale (CFS) suffer methodological shortcomings. The present study aimed to re-evaluate its psychometric properties using exploratory structural equation modeling (ESEM). Methods A Chinese sample of 1259 community-dwelling residents completed the 11-item Chinese CFS and a variety of health measures (anxiety, depression, exhaustion, sleep disturbance, and quality of life). In addition to traditional confirmatory factor analysis, ESEM was performed to assess the fit of two- and three-factor models using robust maximum likelihood estimation and oblique geomin rotation. Convergent validity of the CFS was examined via associations with five covariates (gender, age, exercise, perceived health, and life event) and the health measures in the ESEM model. Results The ESEM models displayed a superior fit to confirmatory factor models. The three-factor ESEM model showed a satisfactory model fit to the data but not for the two-factor model. The three factors were physical fatigue (three items, a = .800), low energy (four items, a = .821), and mental fatigue (four items, a = .861). The factors exhibited convergent validity with the model covariates and health measures. Conclusion The results demonstrate the satisfactory reliability and convergent validity for the three-factor structure of the CFS as a valid measure of fatigue symptoms in the general population. Future psychometric studies could adopt the ESEM approach as a practical alternative to traditional confirmatory factor analysis.
Chinese; Chronic fatigue; Convergent validity; Cross-loadings; Factor structure
-
Chronic fatigue is a symptom commonly reported by
patients in primary care practice and by the general
population, with prevalence of 11.3 % among British primary
care patients [1] and of 10.7 % among the general
population of Hong Kong [2]. Patients with relapsing and
unexplained fatigue that persists for at least 6 months are
said to suffer from chronic fatigue syndrome. This
debilitating syndrome is associated with significant disability
in the functioning capacity of the cognitive and
psychosocial domains [3]. The 11-item Chalder Fatigue Scale
(CFS) [4, 5] was developed as an assessment tool for
fatigue in both general and clinical populations [6, 7]. The
scale has shown adequate degrees of reliability and
convergent validity [5, 8, 9].
Regarding the factor structure of the CFS, a two-factor
structure was originally proposed [4]. Despite some
empirical support for the two-factor structure [8, 10], previous
validation studies of the CFS suffer methodological
shortcomings. First, most of these studies adopted the
outdated principal component analysis and varimax
rotation approaches. Principal component analysis does not
distinguish shared variance from unique variance [11] and
is a biased estimator in factor analysis [12]. The unrealistic
orthogonal factors resulting from varimax rotation likely
lead to distorted factor structures [12]. The Kaisers
criterion of retaining factors with eigenvalues that exceed one is
known to be unreliable and biased. The frequent use of
these outdated approaches diminishes the credibility of
these results on the factor structure of the CFS.
Second, Wong and Fielding [13] applied confirmatory
factor analysis (CFA) to evaluate the factor structure of the
CFS in a Chinese sample. They compared the fit of a
twofactor correlated model with a two-factor model with a
second-order factor (see their Fig. 1, p. 91). Although they
claimed to successfully replicate the original two-factor
structure by showing a superior model fit for the latter
model, the second-order factor model with only two
firstorder factors was actually statistically unidentified and
addition of a second-order factor should not result in a
decrease in model Chi-square. It remains open to question
whether their findings replicated the original two-factor
model or provided evidence in support of a three-factor
model.
Given the methodological limitations of the existing
validation studies, there is a clear need for systematic
psychometric analysis on this widely used scale.
Traditional CFA has been criticized for being overly restrictive
in fixing all cross-loadings to zero [14]. The
(...truncated)