Validation of a 3-Factor scoring model for the Pittsburgh Sleep Quality Index in Older Adults

University of Pittsburgh, Pittsburgh, Pennsylvania, United States
Sleep (Impact Factor: 4.59). 02/2006; 29(1):112-6.
Source: PubMed
The Pittsburgh Sleep Quality Index (PSQI) is widely used to assess subjective sleep disturbances in psychiatric, medical, and healthy adult and older adult populations. Yet, validation of the PSQI single-factor scoring has not been carried out.
The PSQI was administered as a self-report questionnaire. Using a cross-validation approach, scores from the PSQI were analyzed with exploratory and confirmatory factor analyses.
San Diego, Denver, and Los Angeles community-based clinics.
Community-dwelling depressed and nondepressed adults older that 60 years of age (N = 417)
Results yielded a 3-factor scoring model that obtained a measure of perfect fit and was significantly better fitted than either the original single-factor model or a 2-factor model. Components of the 3 factors were characterized by the descriptors sleep efficiency, perceived sleep quality, and daily disturbances.
These findings validate the factor structure of the PSQI and demonstrate that a 3-factor score should be used to assess disturbances in three separate factors of subjective sleep reports.


Available from: Michael R Irwin, Mar 16, 2014
SLEEP, Vol. 29, No. 1, 2006
with sleep performance, ranging from long latency periods be-
fore falling asleep and frequent awakenings at night to difficulties
returning to sleep upon awakening.
Such disturbances of sleep
impact daytime functioning, reduce quality of life, and are report-
ed to lead to declines in heath status
and increases in all-cause
The Pittsburgh Sleep Quality Index (PSQI) is a widely used
19-item self-report questionnaire that measures sleep disturbanc-
Seven clinically derived domains of sleep difficulties, in-
cluding sleep quality, sleep latency, sleep duration, habitual sleep
efficiency, sleep disturbances, use of sleeping medications, and
daytime dysfunction are assessed by the PSQI. Together, these
sleep domains are scored as a single factor or PSQI Sleep Qual-
ity. Whereas many psychometric aspects of the PSQI have been
examined and found to be appropriate, including internal consis-
concurrent validity,
and discriminative validity,
scoring validity of the PSQI has not been statistically examined.
Given that efficacy of a scoring system is an essential aspect of
it is important to know whether a single summed total
score, as is presently used in single-factor scoring of the PSQI,
best captures the multidimensional nature of sleep disturbance as
indexed by the PSQI.
In this study, we examined the factor structure of the PSQI
score using a cross-validation approach. An exploratory factor
analysis (EFA) was followed by a confirmatory factor analysis
(CFA) to ascertain the replicability of the factor structure in a
second independent sample. Furthermore, the CFA compared
the structure obtained through EFA with other logical structures
for the PSQI, namely, the single-factor model. PSQI scores were
measured in a sample of community-dwelling depressed and non-
depressed older adults. This population provides an ideal sample
in which to conduct initial factorial examination because of the
full range of PSQI scores found in older adults, as well as the
high prevalence of sleep disturbances in nondepressed as well as
depressed elderly persons.
The Depression Substudy of the Veterans Affairs Cooperatives
Trial #403, Shingles Prevention Study, provided the data pre-
sented in this article. The Shingles Prevention Study is a double-
blind, placebo-controlled, multicenter, efficacy trial to determine
whether vaccination with live-attenuated Oka/Merck varicella
vaccine decreases the incidence and/or severity of herpes zoster
and its complications in adults 60 years of age and older over
the course of 2-year longitudinal follow-up.
ing older adults (veterans and nonveterans) were recruited using
general media publicity, letters of invitation, advertising, and in-
teractions with local referral groups. The Depression Substudy
identified subjects from 3 sites: University of Colorado; Univer-
sity of California, San Diego (UCSD) and San Diego Veterans
Affairs Healthcare Center; and University of California, Los An-
Validation of a 3-Factor Scoring Model for the Pittsburgh Sleep Quality Index in
Older Adults
Jason C. Cole, PhD
; Sarosh J. Motivala, PhD
; Daniel J. Buysse, MD
; Michael N. Oxman, MD
; Myron J. Levin, MD
; Michael R. Irwin, MD
University of California, Cousins Center for Psychoneuroimmunology, Los Angeles, CA;
University of Pittsburgh, School of Medicine, Pittsburgh, PA;
University of California, San Diego Veterans Affairs Healthcare System, San Diego, CA;
University of Colorado, Health Sciences Center, Denver, CO
Validation of Pittsburgh Sleep Quality Index
—Cole et al
Disclosure Statement
This was not an industry supported study. Dr. Cole has worked as a senior
consultant for QualityMetric and is the president of Consulting Measurement
Group. Dr. Buysse is a consultant for Actelion, Cephalon, Eli Lilly, Merck,
Neurocrine, Pfizer, Respironics, Sanofi-Synthelabo, Servier, Sepracor, and
Takeda. Dr. Levin has received research support from GlaxoSmithKline; and
shares authorship of a pediatric text for McGraw Hill. Drs. Irwin, Motivala,
and Oxman have indicated no financial conflicts of interest.
Submitted for publication June 2005
Accepted for publication August 2005
Address correspondence to: Michael R. Irwin, MD, Cousins Center for Psy
choneuroimmunology, UCLA Neuropsychiatric Institute, 300 UCLA Medical
Plaza, Room 3130, Los Angeles, CA 90095-7076; E-mail: mirwin1@ucla.
Study Objectives: The Pittsburgh Sleep Quality Index (PSQI) is widely
used to assess subjective sleep disturbances in psychiatric, medical, and
healthy adult and older adult populations. Yet, validation of the PSQI sin
gle-factor scoring has not been carried out.
Design: The PSQI was administered as a self-report questionnaire. Using
a cross-validation approach, scores from the PSQI were analyzed with
exploratory and confirmatory factor analyses.
Setting: San Diego, Denver, and Los Angeles community-based clinics.
Participants: Community-dwelling depressed and nondepressed adults
older that 60 years of age (N = 417)
Measurements and Results: Results yielded a 3-factor scoring model
that obtained a measure of perfect fit and was significantly better fitted
than either the original single-factor model or a 2-factor model. Compo-
nents of the 3 factors were characterized by the descriptors sleep effi
ciency, perceived sleep quality, and daily disturbances.
Conclusions: These findings validate the factor structure of the PSQI
and demonstrate that a 3-factor score should be used to assess distur
bances in three separate factors of subjective sleep reports.
Keywords: Sleep quality, latent analysis, confirmatory factor analysis,
Citation: Cole JC; Motivala SJ; Buysse DJ et al. Validation of a 3-factor
scoring model for the pittsburgh sleep quality index in older adults.
2006;29(1): 112-116.
Page 1
SLEEP, Vol. 29, No. 1, 2006
geles (UCLA). All procedures were approved by the institutional
review boards of the University of Colorado, UCSD, and UCLA.
A total of 2858 subjects entering the Shingles Prevention Study
underwent screening for entry into the Depression Substudy. De-
pression screening included completion of an abbreviated version
of the Centers for Epidemiological Study of Depression scale
and answering 2 questions as to whether they had a prior episode
of depression or had been treated for a depression. Persons who
scored above the previously validated Centers for Epidemiologi-
cal Study of Depression scale score for depression
or answered
affirmatively for having had or received treatment for a depres-
sion were interviewed using the Structured Clinical Interview for
Diagnostic and Statistical Manual—IV diagnosis (n = 212).
addition, a sample of age- and sex-comparable participants who
did not meet depression-screening criteria were interviewed (n =
219). As part of the Depression Substudy, questionnaire data on
sleep quality, depressive symptom severity, and health function-
ing were obtained along with blood samples for assessment of
varicella zoster virus immunity, to be reported elsewhere. Partici-
pants then received either varicella vaccine or placebo as previ-
ously reported.
Of the 431 older adults who were enrolled into the Depression
Substudy, 14 subjects were excluded due to current or lifetime
history of alcohol dependence or other Axis I psychiatric disor-
ders. The final sample included 417 participants, including 67
persons with current depressive disorder, 143 individuals with
depressive disorder in full remission, and 207 persons who were
never mentally ill. Women comprised 55.2% of the sample, and
participants were predominately Caucasian (97.1%). The sample
ranged in age from 60 to 95 years, with a mean of 68.90 years (SD
= 6.34 years).
The PSQI, administered by questionnaire, includes 19 items
that measure self-reported sleep disturbances, including hours of
sleep, ratings for frequency of problematic sleeping behaviors, and
subjective sleep quality. Items are measured on either an open-
ended format (such as regular bedtime) or a 5-point Likert scale
(with varying anchors depending on the questions). According to
the scoring guidelines provided by Buysse et al,
the 19 items are
recoded with various algorithms to comprise 7 sleep components:
subjective sleep quality, sleep latency, sleep duration, habitual
sleep efficiency, sleep disturbances, use of sleeping medications,
and daytime dysfunction. The PSQI has favorable psychometric
properties, with internal consistency reliability ranging from .80
to .83,
test-retest reliability from .85
to .87,
convergent validity
with other self-report measures of sleep
and sleep logs,
and good
sensitivity and specificity for identifying those with or without
sleep impairments using a PSQI total score cutoff of 5.0 or more.
Data Analysis
Data were entered and cross-checked by research assistants
with ample data-entry experience. PSQI item responses were
scored into 7 different components, which had small amounts of
missing data, with no more than 5.5% missing data for any com-
posite. A single-point multiple imputation procedure for missing
data replacement
was conducted for the missing points. PSQI
component descriptive statistics are in Table 1 for each group in
the cross validation.
A cross-validation approach was undertaken to assess the fac
tor structure of the PSQI.
Given the nature of the variant and
nonlinear transformations from item responses into component
scores, factor analysis was conducted on the component scores.
After randomly splitting the sample into 2 independent subsam-
ples, 1 subsample was analyzed with EFA (EFA sample n = 207:
current depressive disorder n = 36; depressive disorder in full re-
mission, n = 78; never mentally ill, n = 93) followed by CFA on
the second subsample (CFA sample of n = 210: current depressive
disorder n = 31; depressive disorder in full remission, n = 65;
never mentally ill, n = 114)
In the EFA subsample, principal components analysis was em-
ployed to determine the number of factors to retain for the EFA
based on criteria from Preacher and MacCallum.
EFA was carried out using maximum likelihood estimation ex-
traction and direct oblimin rotation. Factor loadings (i.e., the
correlation between each PSQI component to each factor) were
evaluated against criteria from Comrey and Lee
: .71 or greater
signifies excellent loadings, .63 to .70 are very good; .55 to .62
are good; .45 to .54 are fair; and .32 to .44 are deemed poor, while
any values lower than .32 are discarded.
Once the EFA was completed, a CFA was undertaken in the
CFA sample to test the replicability of the EFA results. Models of
the latent structure should be more than well-fitted; they should
also be better fitting than other logical structures or models.
Therefore, along with the results from the EFA, the CFA exam-
ined the single-factor scoring model utilized by the PSQI manu-
al; in the single-factor model, all 7 components load on a single
PSQI score. The resultant models were analyzed to determine
the degree to which each model fit with the CFA subgroup data.
Maximum likelihood extraction was carried out on the covariance
matrix, and multivariate nonnormality was smoothed over using
Per the recommendations of Schumacker and
Validation of Pittsburgh Sleep Quality Index—Cole et al
1—Pittsburgh Sleep Quality Index Component Correlations and
Descriptive Statistics
Exploratory Factor Analysis Sample
1 2 3 4 5 6 7
1. Subjective sleep quality .51 .35 .49 .41 .29 .39
2. Sleep latency .26 .45 .32 .32 .16
3. Sleep duration .60 .10 .04 .04
4. Habitual sleep efficiency .22 .29 .11
5. Sleep disturbances .12 .28
6. Use of sleep medications .17
7. Daytime dysfunction
Mean 0.74 0.76 0.41 0.58 1.22 0.65 0.70
SD 0.73 0.87 0.68 0.89 0.52 1.08 0.65
Confirmatory Factor Analysis Sample
1. Subjective sleep quality .59 .56 .66 .46 .39 .38
2. Sleep latency .40 .52 .34 .37 .21
3. Sleep duration .69 .11 .20 .27
4. Habitual sleep efficiency .25 .31 .25
5. Sleep disturbances .16 .35
6. Use of sleep medications .14
7. Daytime dysfunction
Mean 0.71 0.68 0.52 0.53 1.25 0.57 0.65
SD 0.76 0.86 0.81 0.87 0.62 1.04 0.76
Correlations provided for descriptive purposes and were, therefore, not
analyzed for significance. All data were based upon multiple imputa
tion data replacement database.
Page 2
SLEEP, Vol. 29, No. 1, 2006
multiple fit indexes were used to determine adequate
model fit: goodness of fit and adjusted goodness of fit at .90 or
comparative fit index at .95 or higher,
and root mean
squared error of approximation (RMSEA) at .06 or lower.
information on these different measures of fit can be found in oth-
er applied research.
Finally, the models were compared to each
other to determine which best fit the data. Three statistics were
used to make these comparisons: Δχ
overlap in the RMSEA
confidence intervals,
and the Bayesian information criterion,
for which differences of 10 or more provides near-conclusive
evidence that the model with the lower value is better fitted.
model was determined to be significantly better fitted than an-
other model if at least 2 of the 3 criteria for significant differences
were met. It should be noted that subgroup (e.g., depressed, histo-
ry of depression, and controls) comparisons were not conducted,
as Byrne
has noted that a model should be validated on a general
sample before multigroup latent analyses can be conducted.
PSQI total scores for the 417 participants ranged from 0 to 18,
with a mean of 4.98 (SD = 3.63). Table 1 provides the mean scores,
SDs, ranges, and intercorrelations for each of the 7 components of
the PSQI for the EFA and CFA samples separately. Correlations
among many of the components were small to large,
from the low .10s to the mid .60s. Each of the 7 PSQI component
scores ranges from 0 to 3, with the means and SDs between 0.5
and 1.0 for most components.
Exploratory Factor Analysis
An EFA was performed on a randomly assigned sample to pro-
vide an exploratory analysis of the PSQI latent structure. EFA re-
sults are displayed in Table 2, where each component is given a
loading value. Two factors were identified. Factor 1 was labeled
Sleep Efficiency, given the strong loadings from the PSQI com-
ponents habitual sleep efficiency (.86) and sleep duration (.60).
Factor 2 was labeled Perceived Sleep Quality, given the strong
loadings from subjective sleep quality (.77) and daytime dysfunc-
tion (.55). Six of the 7 components had excellent to fair loadings.
Use of sleeping medications showed similarly poor loading on
both factors, although, for modeling purposes, this component
was determined to be on its most-fitted factor, Perceived Sleep
Quality (.31). Additionally, Table 2 shows that 39.9% of the vari-
ance was accounted for by the factor Sleep Efficiency, and 17.4 %
of the variance was accounted for by the second factor, Perceived
Sleep Quality. Finally, there was a medium-sized effect
for the
correlation between the 2 factors (r = .33).
Confirmatory Factor Analysis
Based on the EFA, the CFA was run on the 2-factor solution on
the other random half of the sample. In addition, a CFA was per-
formed using the original PSQI single-factor model. Fit statistics
for the single-factor model were insufficient for all but goodness
of fit. However, fit statistics for the 2-factor model were more
impressive, with all fit indexes revealing sufficient fit except for
RMSEA (which was .09). Poor RMSEA suggests that either too
many paths or too few latent variables are present in the model.
Finally, the 2-factor model was significantly better fitted than the
single-factor model according to study criteria (on Δχ
and Bayes-
ian information criterion difference but not on RMSEA differ-
The 2-factor model was examined further, given its insuffi-
cient RMSEA yet sufficient fit of all other fit indexes. Lagrange
modifications indexes (which are used to test if any unmodeled
paths will have a marked improvement on mode fit
) indicated
that there was an unmodeled but marked relationship between the
PSQI components of daytime dysfunction and sleep disturbance.
The modification index indicated that addressing this relation-
ship would improve fit with a reduced χ
from between 33% and
Based on the modification index, a 3-factor model was devel-
oped and tested as shown in Figure 1, with inclusion of a new fac-
tor labeled Daily Disturbances. The 3-factor model met all 4 fit
criteria and was significantly better fitted than either the single-
factor model or the 2-factor model. Indeed, this model obtained
the status of perfect fit, a noteworthy classification for models
that have an RMSEA lower bound of 0. Moreover, to ensure that
the 3-factor model was not sample specific, it was also tested
using CFA on the original EFA subsample. Again, results were
excellent for the 3-factor model, and it was not significantly dif-
ferent on any model comparison criteria than when tested on the
CFA sample. Moreover, the relationship of each PSQI component
score to its respective factor in the 3-factor model was significant
Validation of Pittsburgh Sleep Quality Index—Cole et al
Subjective sleep
Sleep latency
Sleep duration
Habitual sleep
Sleep disturbances
Sleeping medication
Daytime dysfunction
Figure 1—Pittsburgh Sleep Quality Index 3-factor model with stan-
dardized path coefficients between the factor solution and the PSQI
2—Factor Matrix for the 2-Factor Solutions
Pittsburgh Sleep Quality Sleep Perceived
Index Component
Efficiency Sleep Quality
Subjective sleep quality .15
Sleep latency .23
Sleep duration .60
Habitual sleep efficiency .86
Sleep disturbances -.04
Use of sleep medications .14
Daytime dysfunction -.15
Percentage of total variance, % 39.9 17.4
Factor analysis conducted with maximum likelihood estimate extrac
tion and direct oblimin rotation. a = excellent loading, b = very good
loading, c = good loading, d = fair loading, e = poor loading, f = load
ing too low to interpret. Interfactor correlation = .33.
Page 3
SLEEP, Vol. 29, No. 1, 2006
and large, ranging from the standardized path coefficients of .43
(sleeping medication use to Perceived Sleep Quality) to .91 (ha-
bitual sleep efficiency to Sleep Efficiency). Correlations between
the factors ranged from .42 (medium large effect) to .82 (very
large effect).
Based on the original clinical formulation of the PSQI, Buysse
et al
suggested that the 7 components of the PSQI be combined
into a single factor, or the PSQI Total Score. The present find-
ings represent the first empirical examination of the PSQI scoring
system and demonstrate that a 3-factor model is statistically fa-
vored over a single score. In the single-factor model, the average
standardized loading of individual components was .63, whereas,
in the 3-factor model, this jumped to .73. These findings indicate
that the 3-factor model provides a scoring system that is more
reflective of how people respond to the PSQI. In addition, with
the 3-factor model, each PSQI component has a critical role in de-
termining the factor score, which means that the PSQI can assess
severity of sleep impairment in each of 3 separate domains.
The present findings suggest the potential benefit of altering
scoring of the PSQI from a single unitary index of sleep quality
to a 3-dimensional assessment of sleep disturbance with scoring
of the 3 factors: Sleep Efficiency, Perceived Sleep Quality, and
Daily Disturbances. Without such changes in scoring, clinicians
may miss significant sleep impairment that might only reside on
1 of the 3 PSQI factors. In other words, relying solely on the total
score might not identify disturbances in 1 dimension or factor of
the PSQI. Moreover, the 3-factor score has the benefit of obtain-
ing varied assessment of the sleep problems from a single ques-
tionnaire. Knowing more about the type and nature of sleep prob-
lems is necessary to guide the selection of treatment.
these assets, recommendation to change the scoring of the PSQI
requires caution, as these findings were generated using a sample
that was composed exclusively of depressed and nondepressed
older adults. Hence, these data may not generalize to middle-
aged adults or other clinical samples. Moreover, future studies are
needed to address whether these 3 factors have clinical utility in
identifying persons with and without insomnia.
Both the 2- and 3-factor models suggest subtle distinctions
in the relationships between sleep difficulties. For example, 2
prominent sleep complaints in older adults—sleep duration and
latency—loaded on separate factors. Furthermore, the PSQI com-
ponents of sleep latency and sleeping medication use were more
closely associated with Perceived Sleep Quality than were mea-
sures of sleep duration or habitual sleep efficiency. In addition,
future studies are needed to define the factor-score cutpoint that
will optimally identify sleep impairment.
One general limitation to the current study is the matter of as-
sumed structural invariance. Structural invariance exists when the
factor structure of a model remains constant between different
groups. Herein, structural invariance was assumed for sex, de-
pression group (current depressive disorder, depressive disorder
in remission, and never mentally ill), and age cohort (all partic-
ipants were over 60 years of age). Given that different scoring
procedures do not exist for any subgroups on the PSQI,
the as-
sumption of invariance is consistent with the PSQI scoring sys-
tem. Nevertheless, future investigation of the PSQI could benefit
from multigroup CFA.
In summary, this study provides the first examination of the
factor structure of the PSQI—a crucial aspect of validity. Three
factors—Sleep Efficiency, Perceived Sleep Quality, and Daily
Disturbances—are derived from the PSQI. Although these fac-
tors require further validation, especially in other populations,
multidimensional 3-factor scoring of the PSQI is favored over the
single-factor PSQI total score.
This work was supported in part by grants MH55253, T32-
MH19925, AG18367. The authors thank Jessica Lee for her as-
sisting in preparation of the manuscript.
1. Ancoli-Israel S. Insomnia in the elderly: A review for the primary
care practitioners. Sleep 2000;23:S23-30.
2. L
éger D, Scheuermaier K, Philip P, Paillard M, Guilleminault C.
SF-36: Evaluation of quality of life in severe and mild insomniacs
compared with good sleepers. Psychosom Med 2001;63:49-55.
3. Dew MA, Hoch CC, Buysse DJ, et al Healthy older adults’ sleep
predicts all-cause mortality at 4 to 19 years of follow-up. Psycho
som Med 2003;65:63-73.
4. Buysse DJ, Reynolds CF, Monk TH, Berman SR, Kupfer DJ. The
Pittsburgh Sleep Quality Index: A new instrument for psychiatric
practice and research. Psychiatry Res 1989;28:193-213.
5. 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 Psychosom Res 2002;53:737-40.
6. Carpenter JS, Andrykowski MA. Psychometric evaluation of the
Pittsburgh Sleep Quality Index. J Psychosom Res 1998;45:5-13.
7. Doi Y, Minowa M, Uchiyama M, et al Psychometric assessment of
subjective sleep quality using the Japanese version of the Pittsburgh
Sleep Quality Index (PSQI-J) in psychiatric disordered and control
subjects. Psychiatry Res 2000;97:165-72.
8. Messick S. Validity in psychological assessment: Validation of in
ferences from persons’ responses and performances as scientific
inquiry into score meaning. Am Psychol 1995;50:741-9.
9. Irwin M, Cole JC. Depression and psychoneuroimmunology. In:
Vedhara K, Irwin M, eds. Human Psychoneuroimmunology. New
York: Oxford; 2005:243-64.
10. Morin CM, Hauri PJ, Espie CA, Spielman AJ, Buysse DJ, Bootzin
RR. Nonpharmacologic treatment of chronic insomnia. Sleep
11. Oxman MN, Levin MJ, Johnson GR, et al A vaccine to prevent her
pes zoster and postherpetic neuralgia in older adults. N Engl J Med
12. Irwin M, Artin KH, Oxman MN. Screening for depression in the
older adult: criterion validity of the 10-item Center for Epide
miological Studies Depression Scale (CES-D). Arch Intern Med
13. Spitzer RL, Williams JBW, Gibbons M, First MD. Structured Clini-
cal Interview of the DSM-IV. Washington: American Psychiatric
Press; 1994.
14. Schafer JL, Graham JW. Missing data: our view of the state of the
art. Psychol Methods 2002;7:147-77.
15. Cole JC, Oliver TM, McLeod JS, Ouchi BO. Cross validating the
latent structure of Accuplacer: a factor analytic approach. Research
in the Schools 2003;10:63-70.
16. Preacher KJ, MacCallum RC. Repairing Tom Swift’s electric factor
analysis machine. Understanding Statistics 2003;2:13-43.
17. Comrey AL, Lee HB. A First Course in Factor Analysis. 2
ed Hill-
sdale, NJ: Lawrence Erlbaum; 1992.
18. Schumacker RE, Lomax RG. A Beginner’s Guide to Structural
Equation Modeling. 2
ed. Mahwah, NJ: Lawrence Erlbaum;
Validation of Pittsburgh Sleep Quality Index—Cole et al
Page 4
SLEEP, Vol. 29, No. 1, 2006
19. Bollen K, Stine RA. Bootstrapping goodness-of-fit measures in
structural equation models. Sociol Methods Res 1992;21:205-29.
20. Nevitt J, Hancock GR. Improving the root mean square error of ap
proximation for nonnormal conditions in structural equation model-
ing. J Exp Education 2000;68:251-68.
21. Schumacker RE, Lomax RG. A Beginner‘s Guide to Structural
Equation Modeling. Mahwah, NJ: Lawrence Erlbaum; 1996.
22. Bentler PM, Bonett DG. Significance tests and goodness-of-fit in
the analysis of covariance structures. Psychol Bull 1980;88:588-
23. Hu L-t, Bentler PM. Fit indices in covariance structure modeling:
Sensitivity to underparameterized model misspecification. Psychol
Methods 1998;3:424-53.
24. Hu L-t, Bentler PM. Cutoff criteria for fit indexes in covariance
structure analysis: Conventional criteria versus new alternatives.
Structural Equation Modeling 1999;6:1-55.
25. Cole JC, Motivala SJ, Dang J, et al Structural validation of the
Hamilton Depression Rating Scale. J Psychopath Behav Assess
26. Cole JC, Motivala SJ, Khanna D, Lee JY, Paulus HE, Irwin MR.
Validation of a single-factor structure and the scoring protocol for
the Health Assessment Questionnaire-Disability Index (HAQ-DI).
Arthritis Care Res 2005;53:536-42.
27. Long JS. Confirmatory factor analysis Newbury Park, CA: Sage;
28. Browne MW, Cudeck R. Alternative ways of assessing model fit.
In: Bollen KA, Long JS, eds. Testing Structural Equation Models.
Newbury Park, CA: Sage; 1993:136-62.
29. Raftery AE. Bayesian model selection in structural equation models.
In: Bollen KA, Long JS, eds. Testing Structural Equation Models.
Newbury Park, CA: Sage; 1993:163-80.
30. Byrne BM. Structural Equation Modeling With AMOS: Basic CON-
CEPTS, Applications, and Programming. Mahwah, NJ: Lawrence
Erlbaum; 2001.
31. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2
ed Hillsdale, NJ: Lawrence Erlbaum; 1988.
32. Irwin M, Cole JC, Nicassio PM. Comparative meta-analysis of be-
havioral interventions for insomnia in adults and in older adults 55
+ years. Health Psychology 2006: In Press.
33. Morin CM, Colecchi C, Stone J, Sood R, Brink D. Behavioral and
pharmacological therapies for late-life insomnia: a randomized con
trolled trial. JAMA 1999;281:991-9.
Validation of Pittsburgh Sleep Quality Index
—Cole et al
Page 5
    • "According to this criterion, 32% of the Austrian general population [18] and 39% of the general population of Hong Kong [16] are bad sleepers. Psychometric properties of the PSQI have been examined in multiple studies: internal consistency [19], test-retest reliability [20], validity [8,[21][22][23], and factorial structure [24][25][26][27]. The PSQI has been used in different cultures such as in Japan [7], China and Hong Kong [16,28], Nigeria [29], and Brazil [30]. "
    [Show abstract] [Hide abstract] ABSTRACT: • The prevalence of sleep problems in the general population was 36%.
    No preview · Article · May 2016
    • "Although the total score of PSQI is typically used to identify sleep quality, the results of our exploratory factor analysis yielded a two-factor structure, including a sleep quality factor and sleep disturbance factor. Based on the results of exploratory and confirmatory factor analysis, a two-factor model demonstrated a better fit than the one-factor model proposed by Buysse [17] , which was consistent with reports from several previous studies [12, 14, 18, 30]. Our study findings and those of others suggest that the use of a single summed global score of all the six subscales of PSQI might not best capture the multidimensional nature of poor sleep quality. "
    [Show abstract] [Hide abstract] ABSTRACT: Purpose Poor sleep quality during pregnancy is associated with adverse obstetric and neuropsychiatric outcomes. Despite its routine use as a sleep quality assessment scale among men and non-pregnant women, the psychometric properties of the Pittsburgh Sleep Quality Index (PSQI) have not been assessed among US pregnant women. We sought to evaluate the construct validity and factor structure of the PSQI among 1488 pregnant women. Methods A structured interview was used to collect information about demographics and sleep characteristics in early pregnancy. The Patient Health Questionnaire-9 (PHQ-9) and the Depression, Anxiety, and Stress Scale-21 (DASS-21) were used to assess symptoms of depression, anxiety, and stress. Consistency indices, exploratory and confirmatory factor analyses (EFA and CFA), correlations, and logistic regression procedures were used. Results The reliability coefficient, Cronbach’s alpha for the PSQI items was 0.74. Results of the EFA showed that a rotated factor solution for the PSQI contained two factors with eigenvalues >1.0 accounting for 52.8 % of the variance. The PSQI was significantly positively correlated with the PHQ-9 (r s = 0.48) and DASS-21 (r s = 0.42) total scores. Poor sleepers (PSQI global score >5) had increased odds of experiencing depression (OR = 6.47; 95 % CI = 4.56–9.18), anxiety (OR = 3.59; 95 % CI = 2.45–5.26), and stress (OR = 4.37; 95 % CI = 2.88–6.65) demonstrating evidence of good construct validity. CFA results corroborated the two-factor structure finding from the EFA and yielded reassuring measures indicating goodness of fit (comparative fit index = 0.975) and accuracy (root mean square error of approximation = 0.035). Conclusions The PSQI has good construct validity and reliability for assessing sleep quality among pregnant women.
    No preview · Article · Jan 2016 · Sleep And Breathing
  • Source
    • "A semi-structured interview was used to obtain consumption histories for alcohol, cocaine, other substances, and smoking [26]. The Pittsburgh Sleep Quality Index (PSQI) and the Hamilton rating scale for depression-24 items (HAMD-24) assessed sleep quality [27] and depressive symptoms respectively [28]. Body mass index (BMI) was calculated from height and weight. "
    [Show abstract] [Hide abstract] ABSTRACT: Background and aims: Sleep disturbance is a prominent complaint in cocaine and alcohol dependence. This controlled study evaluated differences of polysomnographic (PSG) sleep in cocaine dependent and alcohol dependent subjects, and examined whether substance dependence interacts with age to alter slow wave sleep and rapid eye movement (REM) sleep. Design: Cross-sectional comparison SETTING: Los Angeles and San Diego, California, USA. Participants: Abstinent cocaine dependent subjects (n = 32), abstinent alcohol dependent subjects (n = 73), and controls (n = 108); mean age 40.3 years, 91% male; recruited 2005-2012. Measurements: PSG measures of sleep continuity and sleep architecture primary outcomes of Stage 3 sleep and REM sleep. Covariates included age, ethnicity, education, smoking, body mass index, and depressive symptoms. Findings: Compared with controls, both groups of substance dependent subjects showed loss of Stage 3 sleep (p < 0.001). A substance dependence by age interaction was found in which both cocaine- and alcohol dependent groups showed loss of Stage 3 at an earlier age than controls (p < 0.05 for all), and cocaine dependent subjects showed loss of Stage 3 at an earlier age than alcoholics (p < 0.05). Compared with controls, REM sleep was increased in both substance dependent groups (p < 0.001), and cocaine and alcohol dependence were associated with earlier age-related increase in REM sleep (p < 0.05 for all). Conclusions: Cocaine and alcohol dependence appear to be associated with marked disturbances of sleep architecture, including increased rapid eye movement sleep and accelerated age-related loss of slow wave, Stage 3 sleep.
    Full-text · Article · Jan 2016 · Addiction
Show more