Sleep Breath (2015) 19:693–701

DOI 10.1007/s11325-014-1078-6

ORIGINAL ARTICLE

The Epworth Sleepiness Scale in Portuguese adults: from classical

measurement theory to Rasch model analysis
Paulo Sargento & Victoria Perea & Valentina Ladera &
Paulo Lopes & Jorge Oliveira

Received: 20 February 2014 / Revised: 23 May 2014 / Accepted: 6 November 2014 / Published online: 20 November 2014
# Springer-Verlag Berlin Heidelberg 2014

Abstract for both items (.99) and subjects (.78) were considered good.
Background The Epworth Sleepiness Scale (ESS) is a largely The Cronbach’s α coefficient was also satisfactory (.78).
wide used scale for sleepiness assessment. Measurement prop- Conclusions The ESS showed an adequate structural, inter-
erties are studied in a sample of Portuguese adults, using nal, and criterion validity, both in the CMT and the RM,
different statistical procedures. suggesting this as a useful and effective measure for assessing
Methods The sample consisted of 222 Portuguese adults (97 sleepiness in Portuguese adults.
men and 125 women) with a mean age of 42 years old (SD=
12.5), 46 of which had obstructive sleep apnea (OSA) con- Keywords Epworth Sleepiness Scale . Classical measurement
firmed by polysomnography. The participants were assessed theory . Rasch rating scale model . Obstructive sleep apnea
with the ESS, which was tested through a quantitative analysis
based on the classical measurement theory (CMT) or the
Rasch model (RM) conventions. Introduction
Results A principal component factor analysis was performed
according to the CMT, revealing a single factor explaining The Epworth Sleepiness Scale (ESS) is a quantitative method
39.92 % of the total variance of the scale. Internal consistency to evaluate sleepiness or the propensity to fall asleep in eight
measured by Cronbach’s α coefficient was of .77. The mean everyday life situations, through a Likert scale [1–4]. This
of inter-item correlation was of .31 (.05<r>.47), whereas the instrument has been used in different contexts, populations,
item-total correlations were considered good (.46<r>.73). and clinical conditions, with particular interest in respiratory
The ESS total score for OSA patients was significantly higher sleep disorders [5–7]. Several versions of the ESS have been
than healthy participants (p<.05). Overall data from the RM validated for different languages and populations, mainly to
analysis was consistent with the guidelines of Linacre and Portuguese (from Brazil), Chinese, French, German (from
essential unidimensionality was empirically corroborated Switzerland), German (from Germany), Greek, Italian, Japa-
(61 % the percentage of variance explained by the Rasch nese, Korean, English (from New Zealand), Norwegian, En-
analysis). Model fit is adequate and the reliability coefficients glish (from USA), Spanish (from Peru), Serbian, Spanish
(from Spain), Thai, Turkish, and Spanish (from Mexico)
[8–27]. Although the majority of these studies have been
conducted with classical measurement methods, overall re-
Part of the data were presented in a poster session in the 54th Annual
Meeting of Spanish Society of Neurology, in Barcelona, 23 November sults have indicated good measurement properties of previous
2013. versions of the ESS [28–30]. Some studies have also demon-
P. Sargento (*) : P. Lopes : J. Oliveira
strated the suitability of this instrument through confirmatory
Escola de Psicologia e Ciências da Vida, Universidade Lusófona de factor analysis [31] and Rasch analysis [32, 33].
Humanidades e Tecnologias and COPELABS, Campo Grande, 376, Concerning validation studies of measurement health-
1749-024 Lisboa, Portugal related patients-reported outcomes, the relevance to assess
e-mail: p970@ulusofona.pt
its methodological quality has been increasing, mainly in
V. Perea : V. Ladera terms of the instrument’s measurement properties, standards
Universidad de Salamanca, Salamanca, Spain for design requirements, and preferred statistical methods—
694 Sleep Breath (2015) 19:693–701

The COSMIN checklist [34]. Through this checklist, a relative women) were polysomnographically (type 1) diagnosed
consensus from a large group of researchers [35] was reached with untreated OSA, presenting different types of sever-
on the inclusion of the following measurement properties: ity (measured by Apnea-Hypopnea Index (AHI), not
internal consistency, reliability, measurement error, content positional). Patients were recruited from three public
validity (including face validity), construct validity (including hospitals, while the remaining participants were healthy
structural validity, hypotheses testing and cross-cultural valid- volunteers without a clinical diagnosis of OSA (62 men
ity), criterion validity, responsiveness, and interpretability. and 114 women), who were recruited from a sample
Despite the widespread clinical use of the ESS, there are no community based (universities and companies) through
known studies that assess the psychometric properties of the convenience method.
Portuguese version of this scale in Portugal. Furthermore, the Prior to enrollment, for the inclusion criteria, these
existing literature regarding the validation of the ESS is main- subjects completed a short-form sleeping habits ques-
ly based on classical measurement theory, neglecting the tionnaire, in which controls reported not having any
difficulty of scale items and the underlying characteristics of diagnosis or symptoms of sleep disturbances. None of
each individual. Thus, our primary goal is to validate a Por- the subjects assessed (healthy participants and OSA
tuguese version of the ESS in adults by using the classical patients) was a shift worker, had clinical history of
measurement theory methods (CMT) and the Rasch model any neurological or psychiatric disorders, or was doing
(RM) based on the item-response theory. any type of psychotropic medication.
Table 1 shows the basic demographic characteristics
of the sample, as well as the AHI for the clinical
Methods and materials sample of OSA.
The comparisons between healthy subjects and OSA
Design and procedures patients showed statistically significant differences re-
garding age. Tukey HSD revealed statistically signifi-
This study is based on a one-shot design. The required sample cant differences between healthy participants and OSA
size was calculated a priori with Cohen’s d effect size for t tests. patients, but not between the groups of different OSA
A total sample size of 220 participants was required for this severity. As for the gender distribution, standardized
study in order to detect a medium effect size (d=.40; 1−β=.80; residuals showed a difference in the gender distribution,
one-tailed α=.05) for an expected allocation ratio of .30. All particularly in subjects diagnosed with mild to moderate
subjects were volunteers (not paid) that gave their informed OSA.
consent to the study objectives. The study was approved by the
scientific and ethical committee of the clinical institutions
where the subjects were diagnosed and treated for obstructive Materials
sleep apnea (OSA).
Clinical history
Participants
As stated before, a short-form questionnaire was developed to
The sample consisted of 222 Portuguese adults (97 men assess the demographic variables, sleeping habits, and if there
and 125 women). Forty-six of them (35 men and 11 was a previously diagnosed sleeping disorder.

Table 1 Demographical data

Healthy subjects Subjects with mild to moderate OSA Subjects with severe OSA p value

n=176 n=24 n=22

Mean SD Mean SD Mean SD

Age 38.68 10.24 57.17 11.14 56 8.9 .000*

AHI –a –a <30 >30 –
Gender ♂ ♀ ♂ ♀ ♂ ♀
n (sr) n (sr) n (sr) n (sr) n (sr) n (sr)
62 (−1.7) 114 (1.5) 20 (2.9) 4 (−2.6) 15 (1.7) 7 (−1.5) .000*

OSA obstructive sleep apnea, SD standard deviation, AHI Apnea/Hypopnea Index, ♂ masculine, ♀ feminine, sr standardized residuals
a
No polysonographic study has performed
Sleep Breath (2015) 19:693–701 695

Table 2 Descriptives for the eight items and total ESS score chosen because it is used often in Portuguese language to
Mean SD Asymmetry Curtosis Min/max describe a “nap” during the daytime.

Item 1 1.07 1.02 .50 −0.95 0/3 Statistical procedures

Item 2 1.65 1.02 −.20 −1.07 0/3
Item 3 .49 .85 1.78 2.28 0/3 Classical measurement theory The psychometric properties
Item 4 1.15 1.12 .45 −1.19 0/3 through distribution measures, reliability, and validity were
Item 5 1.63 1.15 −.19 −1.40 0/3 estimated according to the assumptions of the Classical Mea-
Item 6 .15 .39 2.64 6.63 0/2 surement Theory. In order to study data distribution, several
Item 7 1.38 1.05 .14 −1.17 0/3 descriptive measures were used, such as the mean, standard
Item 8 .25 0.64 2.80 7.59 0/3 deviation, skewness, kurtosis, and missing values. The struc-
Total ESS score 7.76 4.71 .60 .04 0/23 tural validity was tested with principal component analysis
(varimax rotation), whereas a Student’s t test for independent
SD standard deviation, Min/max minimum/maximum samples was performed to compare patients and healthy par-
ticipants in criterion validity. Inter-item correlations were also
Epworth Sleepiness Scale conducted to study the internal validity. Reliability was tested
under the Cronbach’s alpha procedure.
The ESS consists of eight items, rated on a scale of 0–3, in
which the total score is computed through the sum of item Rasch model The Rating Scale Model (RSM) is an extension
responses. The total score represents a measure of subjective of the RM for polytomous items. These models transform
daytime sleepiness (ranged 0–24, higher results indicate great- ordinal response data of the subjects on an interval scale [36,
er propensity to fall asleep). 37] and are especially recommended for testing psychological
The English version used of the ESS [3] was translated to assessment instruments [38] because of properties such as
Portuguese language by two different specialists in sleeping conjoint measurement of persons and items responses (i.e.,
disorders, an independent neurologist and a neuropsycholo- parameters for persons and items are expressed in the same
gist, and a native English speaker. The retroversion of the units). Moreover, the patterns of subjects’ responses are ad-
Portuguese version of the ESS was done by two other inde- justed to the model (i.e., the probability of an item response
pendent experts in sleeping disorders and a native in English depends only on the levels of person-item in the measured
language. The final translation of the ESS considered the most attribute) [39]. The RM is empirically useful in determining
consensual designation for “dozing” which was translated to the quality of response categories in particularly in Likert-type
Portuguese as passar pelas brasas. This expression was scales [40].

Fig. 1 ESS scree plot for

variance decomposition
696 Sleep Breath (2015) 19:693–701

Table 3 Rotated component Table 5 Component

matrix Component matrix—unidimensional Component
solution
1 2 Item 1 .74
Item 2 .62
Item 1 ,51 .53
Item 3 .70
Item 2 ,78
Item 4 .66
Item 3 .32 .66
Item 5 .54
Item 4 .41 .53
Item 6 .55
Item 5 .77
Extraction method: principal Extraction method: Item 7 .69
component analysis. Rotation Item 6 .80 principal component Item 8 .52
method: varimax with Kaiser Item 7 .76 analysis
normalization. Rotation Item 8 .77
converged in three iterations
The initial solution was satisfactory [KMO=.806, χ2(28)=
432.760, p=.000]. The communalities ranged from .45 to .64.
Results From this initial solution, two factors were extracted with
eigenvalues greater than 1 according to the Kaiser and
The classical methods to assess the effectiveness of the ESS Guttman rule (eigenvalue: 3.194 and 1.418), explaining
were performed using the SPSS v.20 for Windows. 57.65 % of scale variance (Fig. 1).
The item-response theory through Rasch model was con- Table 3 displays the component matrix loadings for each
ducted testing the Linacre guidelines [41]. This analysis was item (>30) after varimax rotation for a two-factor solution.
performed using the Winsteps 3.80.1 [42]. As shown in Table 3, there are three items of the ESS that
loads on both factors. Furthermore, the data from the compo-
nent transformation matrix also suggests interrelation between
Classical measurement theory
the two extracted factors (see Table 4).
Given the issues raised with a two-factor solution, another
Descriptive and distribution analysis
exploratory principal component analysis was performed with
a forced one-factor solution. The results revealed that com-
Table 2 shows the descriptive statistics for items and the total
monalities ranged from .27 to .55 with a single factor model.
ESS.
The total variance explained by this solution was 39.92 %. As
The descriptive statistics indicate that there are no notable
shown in Table 5, all items have higher loadings (>.50) within
deviations in skewness and kurtosis on the total score of the
a single factor. This unidimensional solution is suitable to
ESS. There are, however, non-normal distributions for some
describe our data since it is appropriate to describe the con-
individual items of the scale, with particular attention to item 6
struct and the underlying factor structure.
(i.e., short range of values) and item 8, both showing moderate
positive skewness and kurtosis. Furthermore, an analysis of
missing values has clarified these results (i.e., three partici-
Internal validity
pants failed to respond to item 8).
To study the internal validity of the ESS, item-total correla-
tions with r Pearson were performed. The results show mod-
Structural validity
erate to strong positive correlations (.46<r>.73) between each
individual item and the total scale (all p levels=.000). Item 6
To study the structural validity of the ESS, a factor analysis
and item 8 are the ones that have the lowest correlation with
with a principal component analysis method was performed
the total scale.
using orthogonal varimax rotation on the eight items of the
ESS. Table 6 Descriptives for ESS (global score)

Table 4 Component transformation matrix Healthy subjects Subjects with OSA

1 2 n=176 n=46

1 .714 .700 Mean SD Mean SD

2 −.700 .714
ESS 7.4 4.14 9.13 6.3
Extraction method: principal component analysis. Rotation method:
varimax with Kaiser normalization SD standard deviation
Sleep Breath (2015) 19:693–701 697

Table 7 Effectiveness of rating scale categories of the Epworth Sleepiness Scale

Stage Guideline (Linacre, 2002) Category Epworth

Pre. Scale oriented with latent variable – .43 to .72

1 At least 10 observation of each category Category 0 845 (48 %)
Category 1 381 (21 %)
Category 2 309 (17 %)
Category 3 241 (14 %)
2 Regular observation distribution – Yes
3 Average measures advance monotonically Category 0 −2.39
with category Category 1 −.84
Category 2 .22
Category 3 1.11
Infit Outfit
4 Outfit mean-square less than 2.0 Category 0 1.05 1.03
Category 1 .91 .85
Category 2 .93 1.01
Category 3 1.05 1.40
5 Step calibration advance Category 0 –
Category 1 −.78
Category 2 −.09
Category 3 .88
M→C C→M
6 Ratings imply measures (C→M) and measures Category 0 86 % 74 %
imply ratings (M→C) Category 1 41 % 59 %
Category 2 41 % 50 %
Category 3 67 % 34 %
7 Step difficulties advance by at least 1.4 logits Category 0 −2.16
Category 1 −.62
8 Step difficulties advance by at less than 5 logits Category 2 .60
Category 3 2.20

Note: categories of the ESS: 0—would never doze; 1—slight chance of dozing; 2—moderate chance of dozing; 3—high chance of dozing

Criterion-related validity alpha level when items were removed was studied. The aver-
age inter-item correlation was r=.31 (.05<r>.47), in which
The criterion-related validity of the ESS was tested with an item 6 and item 8 were the most problematic items according
independent t test that was performed to compare patients to this analysis.
diagnosed with OSA vs. healthy participants in ESS scores.
Table 6 depicts mean scores and standard deviations of the
ESS total score in OSA patients and healthy participants. Table 8 Statistics of ESE
The results presented in Table 6 indicate that subjects
M (SD)
diagnosed with OSA have higher levels of sleepiness com-
pared to healthy participants [t(220)=−2.234, p=.026]. Item outfit 1.04 (.29)
Person outfit 1.03 (1.30)
Reliability Item separation reliability .99
Person separation reliability .78
The internal consistency was estimated using Cronbach’s Cronbach’s α .78
alpha method that was performed to study the reliability of % items with outfit >2 0%
the ESS in evaluating sleeping disturbances. The Cronbach’s % persons with outfit >2 9.00 %
alpha (.77) was acceptable for the version of the ESS with the
original eight items, even after the possibility of increasing the M mean, SD standard deviation
698 Sleep Breath (2015) 19:693–701

Rasch model no more than 10 % of the variance (exact value was of 9.2 %)
and the percentage of variance explained by the RM is over
In line with the recommendations of Linacre [41], the category 20 % (exact value of 61 %).
effectiveness of rating scale was tested with a four-category The score statistics is presented in Table 8. Model fit is
system. The item polarity in the actual scale shows a range adequate: no item outfit is over 2 (severe misfit), the percent-
from .43 to .72, in which all items are aligned in the same age of people with outfit over 2 is small, and average outfit
direction in the latent variable (Table 7). values, for items and people, are close to 1 (perfect fit).
The category statistics is shown in Table 8. The results for Furthermore, the score reliability through Item Separation
categories are consistent with the guidelines of Linacre [41], Reliability value (.99), Person Separation Reliability value
indicating the following: at least 10 observations for each (.85), and Cronbach’s alpha (.78) was considered high.
category; an unimodal distribution (peak at category 0 and a The item-person map is depicted in Fig. 3, which shows the
negative skew to category 3); the average measure advances “ability” of the people and “difficulty” of the items on the
monotonically with category, with an increase from .89 to same pathway. A visual inspection of Fig. 3 suggests that item
1.55 logit between consecutive categories; the Outfit mean- 6 (sitting and talking to someone), item 8 (in a car, while
square values of the four categories are around 1.0 logit, all stopped for a few minutes in traffic), and item 3 (sitting
categories are less than 2.0 logit, suggesting a reasonably inactive in a public place; e.g., a theater or meeting) are those
uniform level of randomness in the data; the step calibration considered as more difficult according to these results. On the
advances about .69 to .79 per step, suggesting that people with other hand, item 2 (watching television) and item 5 (lying
higher levels of sleepiness are more prone to choose higher down to rest in the afternoon when circumstances allow it) are
categories, which were modal along the variable. the less difficult items according to RM.
As shown in Fig. 2, each category has a real probability of
being selected by the sample; the relationship between mea-
sure and ratings for each category is considered adequate; the
small and large advancements of steps’ difficulties between Discussion
consecutive categories show about 1.2 to 1.6 logit. Overall,
these results suggest that all the guidelines of Linacre have In order to increase variability and to prevent floor effects, the
been satisfied. statistical analysis was performed for the total sample, includ-
The principal component analysis of residuals from the RM ing both healthy participants and patients with OSA. Howev-
indicates that the assumption of unidimensionality is empiri- er, it is worth noting that the proportion of gender is different
cally corroborated given that the analysis of residuals explains between OSA and healthy participants. In this regard, one

Fig. 2 Probability curves of the

categories of ESS
Sleep Breath (2015) 19:693–701 699

normative study using the ESS [43] does not show any effect
of gender on the ESS total score, although there is evidence
[44] for an impact of respiratory sleep disorders on daytime
sleepiness, especially under the age of 40. This becomes,
however, much less likely for the overall sample, in which
the main statistical procedures were carried out.
Item analysis showed that item 6 and item 8 have relevant
deviations from normal distribution (as regard to kurtosis,
skewness, and range of values, especially in item 6). In our
view, this result may describe a floor effect which is often
observed in healthy individuals [13], but it is also possible that
these items are those that are most affected by social
desirability. One possible explanation for this is that
situations as described in item 6 (during a conversation
with another person) and item 8 (driving a car) are
difficult to assume as possible situations where one
can fall asleep compared to other contexts in which
attention is not socially expected.
The structural validity, evaluated through a principal com-
ponent factor analysis, initially showed two strongly interde-
pendent factors. However, a forced factor analysis shows an
interpretable unidimensional structure, with item loadings (.52
to .74) close to those found in a clinical sample with the
original version from English language [28]. The analysis of
the item-total correlations supports the internal validity of the
scale, with exception for item 6 and item 8. These results are
also consistent with a previous data, despite being more sim-
ilar to clinical than the control samples of Bloch and col-
leagues [13].
The comparisons of the ESS total score among patients
diagnosed with OSA and healthy participants reveal adequate
criterion validity. It is important to consider that the scores
found in our study are lower than those reported for the
original version [28], either for patients with OSA or healthy
subjects, being more adjusted to other versions of the ESS
found in the literature [9, 13, 27].
One important issue with previous research on this topic is
that most studies include convenience samples of college
students as controls that are younger than most of the patients
with OSA. In these versions [9, 13], the mean age of the
control group is higher (about 10 years higher), being more
similar to our control sample. Another important confounder
in these samples of college students is that the need for sleep
may vary depending on the time of school year. During exam
periods, students have less sleeping hours, which may in-
crease the propensity to fall asleep during some of the situa-
tions described in these measures.
The internal consistency of the total scale, estimated
through the Cronbach’s alpha, suggests minimal adequacy of
the scale according to Nunnaly [45]. In addition, the
Cronbach’s alpha coefficient found in our study is in the range
of values found for clinical and control samples [9, 11, 13, 18,
Fig. 3 Item-person map
20, 22, 27].
700 Sleep Breath (2015) 19:693–701

The RM analysis suggests that the standard version of the 17. Takegami M, Suzukamo Y, Wakita T et al (2009) Development of a
Japanese version of the Epworth Sleepiness Scale (JESS) based on
scale with four-category system is consistent with the guide-
item response theory. Sleep Med 5:556–565
lines [41], and the data fit the Rasch model fairly well. Con- 18. Cho YW, Lee JH, Son HK et al (2011) The reliability and validity of
sistent with previous reports [32, 33], no items with high the Korean version of the Epworth Sleepiness Scale. Sleep Breathing
misfit were found through the RM, which also supported the 15:377–384
19. Gander PHH, Marshall NS, Harris R, Reid P (2005) The Epworth
unidimensionality of the ESS.
Sleepiness Scale: influence of age, ethnicity, and socioeconomic
Overall results from the CMT and RM showed ade- deprivation. Epworth Sleepiness Scores of adults in New Zealand.
quate structural, internal, and criterion-related validity of Sleep 28:249–253
the ESS, and reliability as well, suggesting this as a 20. Beiske KK, Kjelsberg FN, Ruud EA, Stavem K (2009) Reliability
and validity of a Norwegian version of the Epworth Sleepiness Scale.
useful and effective measure for assessing sleepiness in
Sleep Breathing 13:65–72
Portuguese adults. 21. Sanford SD, Lichstein KL, Durrence HH et al (2006) The influence
of age, gender, ethnicity, and insomnia on Epworth Sleepiness
Scores: a normative US population. Sleep Med 7:319–326
22. Rosales-Mayor E, Castro JR, Huaynay L, Zagaceta K (2012)
Conflict of interest None, for all authors. Validation and modification of the Epworth Sleepiness Scale in
Peruvian population. Sleep Breathing 16:59–69
23. Kopitovic I, Trajanovic N, Prodic S et al (2011) The Serbian
version of the Epworth Sleepiness Scale. Sleep Breathing 15:
References 775–780
24. Chiner E, Arriero JM, Signes-Costa J, Marco J, Fuentes I (1999)
Validation of the Spanish version of the Epworth Sleepiness Scale in
1. Johns M (1991) A new method for measuring daytime sleepiness: the patients with a sleep apnea syndrome [article in Spanish]. Arch
Epworth Sleepiness Scale. Sleep 14:540–545 Bronconeumol 35:422–427
2. Johns M (1994) Sleepiness in different situations measured by the 25. Banhiran W, Assanasen P, Nopmaneejumruslers C, Metheetrairut C
Epworth Sleepiness Scale. Sleep 17:703–710 (2011) Epworth Sleepiness Scale in obstructive sleep disordered
3. Johns M (1998) Rethinking the assessment of sleepiness. Sleep Med breathing: the reliability and validity of the Thai version. Sleep
Rev 2:3–15 Breathing 15:571–577
4. Johns M (2000) Sensitivity and specificity of the multiple sleep 26. Izci B, Ardic S, Firat H et al (2008) Reliability and validity studies of
latency test (MSLT), the maintenance of wakefulness test and the the Turkish version of the Epworth Sleepiness Scale. Sleep Breathing
Epworth Sleepiness Scale: failure of the MSLT as a gold standard. J 12:161–168
Sleep Res 9:5–11 27. Jiménez-Correa U, Haro R, Poblano A et al (2009) Mexican version
5. Johns M (1993) Daytime sleepiness, snoring, and obstructive sleep of the Epworth Sleepiness Scale. Open Sleep J 2:6–10
apnea. The Epworth Sleepiness Scale. Chest 103:30–36 28. Johns M (1992) Reliability and factor analysis of the Epworth
6. Furuta I, Kaneda R, Kosaka K et al (1999) Epworth Sleepiness Scale Sleepiness Scale. Sleep 15:376–381
and sleep studies in patients with obstructive sleep apnea syndrome. 29. Knutson KL, Rathouz PJ, Yan LL, Liu K, Lauderdale D (2006)
Psychiatry Clin Neurosci 51:301–302 Stability of the Pittsburgh sleep quality index and the Epworth
7. Silva GE, Vana KD, Goodwin JL, Sherill DL, Quan SF (2011) sleepiness questionnaires over 1 year in early middle-aged adults:
Identification of patients with sleep disordered breathing: comparing the CARDIA study. Sleep 11:1503–1506
the four-variable screening tool, STOP, STOP-bang, and Epworth 30. Miletin MS, Hanly PJ (2003) Measurement properties of the Epworth
Sleepiness Scales. J Clin Sleep Med 7:467–472 Sleepiness Scale. Sleep Med 4:195–201
8. Aloé F, Pedro A, Tavares SM (1997) Epworth Sleepiness Scale 31. Smith SS, Oei TP, Douglas JA et al (2008) Confirmatory factor
outcome in 616 Brazilian medical students. [Article in Portuguese]. analysis of the Epworth Sleepiness Scale (ESS) in patients with
Arq Neuropsiquiatr 55:220–226 obstructive sleep apnea. Sleep Med 9:739–744
9. Bertolazi AN, Fagondes SCH, Hoff LS et al (2009) Portuguese- 32. Martinez D, Breitenbach TC, Lumertz MS et al (2011) Repeating
language version of the Epworth Sleepiness Scale: validation for administration of the Epworth Sleepiness Scale in clinical useful.
use in Brazil. [Article in Portuguese]. J Bras Pneumol 35:877–833 Sleep Breathing 15:763–773
10. Chung KF (2000) Use of the Epworth Sleepiness Scale in Chinese 33. Ulander M, Arested K, Svanborg E, Johansson B, Broström A (2012)
patients with obstructive sleep apnea and normal hospital employees. The fairness of the Epworth Sleepiness Scale: two approaches to
J Psychosom Res 49:367–372 differential item functioning. Sleep Breathing 17:157–165
11. Chen N-H, Johns MD, Li H-W et al (2002) Validation of a Chinese 34. Mokkink LB, Terwee CB, Knol DL et al (2010) The COSMIN
version of the Epworth Sleepiness Scale. Qual Life Res 11:817–821 checklist for evaluating the methodological quality of studies on
12. Kaminska M, Jobin V, Mayer P et al (2010) The Epworth Sleepiness measurement properties: a clarification of its content. BMC Med
Scale: self -administration versus administration by the physician, Res Methodol 10:22
and validation of a French version. Can Respir J 17:e23–e34 35. Mokkink LB, Terwee CB, Patrick LD et al (2010) The COSMIN
13. Bloch KE, Schoch OD, Zhang JN, Russi EW (1999) German version checklist for assessing the methodological quality of studies on
of the Epworth Sleepiness Scale. Respiration 66:440–447 measurement properties of health status measurement instruments:
14. Sauter C, Popp R, Danher-Hopfe H et al (2007) Normative values of an international Delphi study. Qual Life Res 19:539–549
German Epworth Sleepiness Scale. Somnolgie 11:272–278 36. Andrich D (1988) Rasch models for measurement. Sage, London
15. Tsara V, Serasli E, Amfilochiou A, Constantinidis T, Christaki P 37. Wright BD, Mok M (2004) An overview of the family of Rasch
(2004) Greek version of the Epworth Sleepiness Scale. Sleep measurement models. In: Smith EV, Smith RM (eds) Introduction to
Breathing 8:91–95 Rasch measurement. JAM Press, Maple Grove
16. Vignatelli L, Plazzi G, Barbato A et al (2003) Italian version of the 38. Prieto G, Delgado AR, Perea MV, Ladera V (2010) Scoring neuro-
Epworth Sleepiness Scale: external validity. Neurol Sci 23:295–300 psychological tests using the Rasch model: an illustrative example
Sleep Breath (2015) 19:693–701 701

with the Rey-Osterrieth complex figure. Clin Neuropsychol 24:45– 43. Sanford SD, Lichstein KL, Durrence HH et al (2011) The
56 influence of age, gender, ethnicity, and insomnia on Epworth
39. Wilson M (2005) Constructing measures. LEA, Mahwah Sleepiness Scores: a normative US population. Sleep Med 7:
40. Bond TG, Fox CM (2007) Applying the Rasch model, 2ªth edn. 319–326
LEA, Mahwah 44. Morrell MJ, Finn L, McMillan A, Peppard P (2012) The impact of
41. Linacre JM (2002) Optimizing rating scale category effectiveness. J ageing and sex on the association between sleepiness and sleep
Appl Meas 3:85–106 disordered breathing. Eur Respir J 40:386–393
42. Linacre JM (2013) A user’s guide to Winsteps Ministep—Rasch- 45. Nunnally JC, Bernstein IH (1994) Psychometric theory, 3rd edn.
model computer programs. Winsteps.com, Chicago McGraw-Hill, USA
Copyright of Sleep & Breathing is the property of Springer Science & Business Media B.V.
and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.