Validation of the three-factor model of the PSQI in a large sample of chronic fatigue
syndrome (CFS) patients
An Mariman⁎, Dirk Vogelaers, Ignace Hanoulle, Liesbeth Delesie, Els Tobback, Dirk Pevernagie
Department of General Internal Medicine, Infectious Diseases and Psychosomatic Medicine, University Hospital Ghent, Belgium
a b s t r a c t a r t i c l ei n f o
Received 1 August 2011
Received in revised form 27 October 2011
Accepted 2 November 2011
Chronic fatigue syndrome
Confirmatory factor analysis
Objective: To evaluate whether a 3-factor model of the Pittsburgh Sleep Quality Index (PSQI) scale would fit
the constellation of sleep disturbances in patients with a diagnosis of chronic fatigue syndrome (CFS).
Methods: Consecutive CFS patients filled out the PSQI. Scores from this self-report questionnaire were exam-
ined with exploratory and confirmatory factor analysis (CFA).
Results: 413 CFS patients were included for analysis in this study. CFA showed that the 7 PSQI component
scores clustered into the 3 factors reported by Cole et al. (2006), i.e. Sleep Efficiency, Perceived Sleep Quality
and Daily Disturbances. In contrast with the single-factor and all 2-factor models, all factor loadings were sig-
nificant, and all goodness-of-fit values were acceptable.
Conclusion: In CFS, the PSQI operates as a 3-factor scoring model as initially seen in healthy and depressed
older adults. The separation into 3 discrete factors suggests the limited usefulness of the global PSQI as a sin-
gle factor for the assessment of subjective sleep quality, as also evidenced by a low Cronbach's alpha (0.64) in
this patient sample.
© 2011 Elsevier Inc. All rights reserved.
Chronic fatigue syndrome (CFS) is a disabling condition character-
ized by chronic fatigue of a new or definite onset that lasts for at least
6 months and that is not explained by medical or psychiatric causes
. Next to this major criterion, the 1994 case definition requires
the co-occurrence of at least four out of eight minor criteria: unusual
postexertional malaise, impaired memory or concentration, unre-
freshing sleep, headaches, muscle pain, joint pain, sore throat and
tender cervical nodes . These 1994 CDC diagnostic criteria prevail
as a standard in current clinical practice and scientific research.
Complaints of unrefreshing sleep and poor sleep quality are com-
mon in CFS patients. The Pittsburgh Sleep Quality Index (PSQI) is one
of the most used and validated questionnaires to measure sleep qual-
ity and disturbances during the past month . The self-report ques-
tions are divided into 7 clinically derived components of sleep
difficulties: subjective sleep quality, sleep latency, sleep duration, ha-
bitual sleep efficiency, sleep disturbances, use of sleeping medications
and daytime dysfunction. Individual component scores are summed
to yield one global score or a single factor, with higher scores indicat-
ing poorer sleep quality. Psychometric properties of the PSQI have
been examined and found to be appropriate in relation to internal
consistency [2,3], concurrent validity [3,4] and discriminative validity
[3,4] in a range of clinical and healthy populations.
Using a cross-validation approach in healthy and depressed elder-
ly US adults, Cole et al. found that a single summed global score did
not best capture the multidimensional nature of sleep disturbances.
An exploratory factor analysis (EFA) followed by a confirmatory fac-
tor analysis (CFA) on the 7 quality components revealed that a 3-
factor scoring model significantly better fitted than either the original
single-factor or a 2-factor model. This model documents sleep distur-
bances in the separated factors Sleep Efficiency, Perceived Sleep Qual-
ity and Daily Disturbances.
Three other studies provided evidence that a multiple factor scor-
ing method of the PSQI could be more appropriate to assess sleep
problems compared to the originally proposed single- factor method.
In a sample of Nigerian university students, a 3-factor model of the
PSQI was identified performing EFA, however, the factors differed
from Cole's findings . EFA and subsequent CFA on the PSQI results
deriving from a sample of Australian adults determined a 2- and 3-
factor scoring model with slight differences in the optimal factor
structures compared to the model of Cole et al. . Conducting CFA,
the original 3-factor model  was also found to better fit than a
single-factor model in renal transplant recipients . Although the
fit indices noticed were not as good as those found by Cole et al. ,
an additional pathway significantly improved its fit .
Differences in sample characteristics may account for the different
factor structures identified in various studies since sleep patterns,
sleep quality and perception of sleep are influenced by a range of fac-
tors related to age, health and culture [9–11]. As a consequence, there
Journal of Psychosomatic Research 72 (2012) 111–113
⁎ Corresponding author at: Department of General Internal Medicine, Infectious
Diseases and Psychosomatic Medicine, University Hospital Ghent, De Pintelaan 185,
9000 Gent, Belgium. Tel.: +32 9 3323708; fax: +32 9 3323895.
E-mail address: email@example.com (A. Mariman).
0022-3999/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
Contents lists available at SciVerse ScienceDirect
Journal of Psychosomatic Research
is a need for further studies examining the factor structure of the
The aim of this study was to evaluate whether the 3-factor model
of the PSQI reported by Cole et al.  would fit the constellation of
sleep disturbances in a large sample of patients with CFS.
Consecutive patients with a final diagnosis of CFS according to the
Fukuda criteria in a multidisciplinary tertiary care referral center
were included in this study . The sample was approved by the
Ethical Review Board of the Ghent University Hospital.
All patients filled out the PSQI and scores were calculated accord-
ing to the scoring guidelines provided by Buysse et al. .
To investigate the validity of the 3-factor model of the PSQI pro-
posed by Cole et al. , CFA was performed using SPSS (PASW 17.0)
and the AMOS module (5.0). An EFA was performed to investigate
the validity of the single-factor and all 2-factor models. The fit of
the models was estimated with the Maximum Likelihood Algorithm.
In line with published recommendations, several indices were
used to assess the model fit [13,14]. These include χ² and its related
degrees of freedom (d.f.) and probability (p), goodness-of-fit index
(GFI), adjusted goodness-of-fit index (AGFI), comparative fit index
(CFI), root mean square error of approximation (RMSEA) and the
consistent Akaike information criterion (CAIC). Chi-square assesses
whether a significant amount of observed covariance between items
remains unexplained by the model. A significant χ² (pb0.05) indi-
cates a bad model fit. The RMSEA is a fit measure based on population
error of approximation . It is unreasonable to assume that the
model will hold exactly in the population. Therefore the RMSEA
takes into account the error of approximation in the population. A
RMSEA value b0.05 indicates a close fit and values up to 0.08 repre-
sent reasonable errors of approximation in the population. The GFI
and the AGFI assess the extent to which the model provides a better
fit compared to no model at all . These indices have a range be-
tween 0 and 1, with higher values indicating a better fit. A GFI
>0.90 and an AGFI >0.85 indicate a good fit of the model. The CFI is
an incremental fit index . It represents the proportionate im-
provement in model fit by comparing the target model with a base-
line model (usually a null model in which all the observed variables
are uncorrelated). The CFI ranges between 0 and 1, with values
>0.90 indicating an adequate fit. The CAIC is a goodness-of-fit mea-
sure which adjusts the model's chi-square to penalize for model com-
plexity and sample size . This measure can be used to compare
non-hierarchical as well as hierarchical (nested) models. Lower
values on the CAIC measure indicate better fit .
The study sample included 415 CFS patients (mean age 40.53 years, SD 7.91; 86% fe-
male)fromwhich413 completely filledoutall PSQIitems,allowing analysiswith the
Table 1 provides the descriptive statistics for the global PSQI, the 7 PSQI compo-
nents and the Spearman's intercorrelations. Generally, high PSQI scores were found
with a mean global score of 10.17 (SD 4.02, Cronbach's alpha 0.64). Poor sleep quality
was observed in 86% of the patients using the recommended cut-off point of 5 .
Several inter-component correlations were not significant; the highest correlation
was found between ‘sleep duration’ and ‘habitual sleep efficiency’ (r=0.71).
Fig. 1 shows the results of the CFA performed on the 3-factor model proposed by
Cole et al. . All factor loadings were significant and all goodness-of-fit values were
acceptable (χ2=14.70, d.f.=11, p=0.20; GFI=0.99; AGFI=0.97; CFI=0. 99;
RMSEA=0.03; CAIC=134.10). In contrast, the single-factor model proposed by
Buysse et al.  indicated a poor fit with the data (χ2=109.90, d.f.=14, pb0.001;
GFI=0.92; AGFI=0.85; CFI=0.84; RMSEA=0.13; CAIC=208.23) as was also the
case for all 2-factor models (significant χ2values; results not shown).
This is the first time that the PSQI factor structure was examined
in a large sample of CFS patients. CFA demonstrated that the PSQI
Pittsburgh Sleep Quality Index (PSQI) component correlations and descriptive statistics.
1. Subjective sleep quality
2. Sleep latency
3. Sleep duration
4. Habitual sleep efficiency
5. Sleep disturbances
6. Use of sleep medication
7. Daytime dysfunction
8. Global PSQI
⁎ Correlation is significant at the 0.05 level (2-tailed).
⁎⁎ Correlation is significant at the 0.01 level (2-tailed).
Fig. 1. Confirmatory factor analysis (CFA) of the 3-factor model for the PSQI in CFS.
Standardized β-coefficients (factor loadings, all significant) and R² values are
shown.χ2=14.70, degrees of freedom (d.f.)=11, probability (p)=0.20; goodness-
of-fit index (GFI)=0.99; adjusted goodness-of-fit index (AGFI)=0.97; comparative
fit index (CFI)=0. 99; root mean square error of approximation (RMSEA)=0.03; con-
sistent Akaike information criterion (CAIC)=134.10.
A. Mariman et al. / Journal of Psychosomatic Research 72 (2012) 111–113
operated as a multiple factor scoring model in CFS, which is consis- Download full-text
tent with previous findings in different subject groups [5–8]. More-
over, the 3-factor model proposed by Cole et al.  showed good fit
criteria and the 7 PSQI component scores clustered into the factors
Sleep Efficiency, Sleep Quality and Daily Disturbances. Therefore, the
3-factor model could improve the sensitivity of the PSQI in assessing
sleep problems in CFS compared to the global PSQI as a single factor.
This limitation of the single-factor global PSQI is further evidenced by
the low Cronbach's alpha in our sample (0.64) as an indicator of inter-
nal consistency, in contrast with the Cronbach's alpha (0.83) of the
original PSQI description .
In conclusion, CFA confirmed that the PSQI operates as a 3-factor
scoring model in CFS. The separation of the PSQI into 3 discrete fac-
tors suggests the limited usefulness of the global PSQI as a single fac-
tor for the assessment of subjective sleep quality in CFS patients.
Conflict of interest
The authors state no conflict of interest and have received no pay-
ment in preparation of this manuscript.
We wish to acknowledge the role of Walter Michielsen, MD PhD,
who has pioneered the rehabilitation program for CFS patients at the
chosomatic Medicine of the University Hospital Ghent, Belgium.
 Fukuda K, Straus SE, Hickie I, et al. The chronic fatigue syndrome — a comprehen-
sive approach to its definition and study. Ann Intern Med 1994;121:953–9.
 Buysse DJ, Reynolds CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep
Quality Index — a new instrument for psychiatric practice and research. Psychia-
try Res 1989;28:193–213.
 Carpenter JS, Andrykowski MA. Psychometric evaluation of the Pittsburgh Sleep
Quality Index. J Psychosom Res 1998;45:5–13.
 Backhaus J, Junghanns K, Broocks A, Riemann D, Hohagen F. Test–retest reliability
and validity of the Pittsburgh Sleep Quality Index in primary insomnia. J Psycho-
som Res 2002;53:737–40.
 Cole JC, Motivala SJ, Buysse DJ, Oxman MN, Levin MJ, Irwin MR. Validation of a 3-
factor scoring model for the Pittsburgh Sleep Quality Index in older adults. Sleep
 Aloba OO, Adewuya AO, Ola BA, Mapayi BM. Validity of the Pittsburgh Sleep Qual-
ity Index (PSQI) among Nigerian university students. Sleep Med 2007;8:266–70.
 Magee CA, Caputi P, Iverson DC, Huang XF. An investigation of the dimensionality
of the Pittsburgh Sleep Quality Index in Australian adults. Sleep Biol Rhythms
 Burkhalter H, Sereika SM, Engberg S, Wirz-Justice A, Steiger J, De Geest S. Struc-
ture validity of the Pittsburgh Sleep Quality Index in renal transplant recipients:
a confirmatory factor analysis. Sleep Biol Rhythms 2010;8:274–81.
 Rao U, Poland RE, Lutchmansingh P, Ott GE, McCracken JT, Lim KM. Relationship
between ethnicity and sleep patterns in normal controls: implications for psycho-
pathology and treatment. J Psychiatry Res 1999;33:419–26.
 Foley D, Ancoli-Israel S, Britz P, Walsh J. Sleep disturbances and chronic disease in
older adults — results of the 2003 National Sleep Foundation Sleep in America
Survey. J Psychosom Res 2004;56:497–502.
 Neu D, Mairesse O, Hoffmann G, et al. Sleep quality perception in the chronic fa-
tigue syndrome: correlations with sleep efficiency, affective symptoms and inten-
sity of fatigue. Neuropsychobiology 2007;56:40–6.
 Mariman A, Vogelaers D, Hanoulle I, Delesie L, Pevernagie D. Subjective sleep
quality and daytime sleepiness in a large sample of patients with chronic fatigue
syndrome (CFS). Acta Clin Belg in press.
 Bollen KA, Long JS. Testing structural equation models. Newbury Park, CA: Sage;
 Byrne BM. Structural equation modeling with Amos: basic concepts, applications
and programming. Mahwah, NJ: Lawrence Erlbaum Associates; 2001.
 Browne MW, Cudeck R. Alternative ways of assessing model fit. In: Bollen KA,
Long JS, editors. Testing structural equation models. Newbury Park, CA: Sage;
1993. p. 136–62.
 Jöreskog KG, Sörbom D, Lisrel IV. Scientific software. Mooresville: user's guide;
 Bentler PM. Comparative fit indices in structural models. Psychol Bull 1990;107:
 Burnham KP, Anderson DR. Model selection and inference: a practical
information-theoretic approach. New York: Springer-Verlag; 1998.
A. Mariman et al. / Journal of Psychosomatic Research 72 (2012) 111–113