![Fit Statistics for Latent Structure Models for HADS and PANAS Data in FM Patients Model 2 df RMSEA [90% CI] TLI CFI](https://www.researchgate.net/profile/Jorge-Osma/publication/258297573/figure/tbl1/AS:613905592229897@1523377907207/Fit-Statistics-for-Latent-Structure-Models-for-HADS-and-PANAS-Data-in-FM-Patients-Model-2_Q320.jpg)
Content uploaded by Jorge Osma
Author content
All content in this area was uploaded by Jorge Osma on Sep 23, 2014
Content may be subject to copyright.
Psychological Assessment
Bifactor Analysis and Construct Validity of the HADS: A
Cross-Sectional and Longitudinal Study in Fibromyalgia
Patients
Juan V. Luciano, Juan R. Barrada, Jaume Aguado, Jorge Osma, and Javier García-Campayo
Online First Publication, December 2, 2013. doi: 10.1037/a0035284
CITATION
Luciano, J. V., Barrada, J. R., Aguado, J., Osma, J., & García-Campayo, J. (2013, December 2).
Bifactor Analysis and Construct Validity of the HADS: A Cross-Sectional and Longitudinal
Study in Fibromyalgia Patients. Psychological Assessment. Advance online publication. doi:
10.1037/a0035284
Bifactor Analysis and Construct Validity of the HADS: A Cross-Sectional
and Longitudinal Study in Fibromyalgia Patients
Juan V. Luciano
Parc Sanitari Sant Joan de Déu, Barcelona, Spain Juan R. Barrada
Universidad de Zaragoza
Jaume Aguado
Parc Sanitari Sant Joan de Déu, Barcelona, Spain Jorge Osma
Universidad de Zaragoza
Javier García-Campayo
Miguel Servet Hospital, Aragon Institute of Health Sciences, Zaragoza, Spain
The dimensionality of the Hospital Anxiety and Depression Scale (HADS) is a current source of controversy
among experts. The present study integrates a solid theoretical framework (Clark & Watson’s, 1991, tripartite
theory) and a fine-grained methodological approach (structural equation modeling; SEM) to examine the
dimensionality and construct validity of the HADS in fibromyalgia (FM) patients. Using the HADS data of
269 Spanish patients with FM, we estimated the cross-sectional and, for the first time, longitudinal fit
(autoregressive model) of 2 competing models (oblique 2-factor vs. bifactor) via confirmatory factor analysis.
The pattern of relationships between the HADS latent dimensions and positive and negative affect (PA and
NA) was analyzed using SEM. HADS reliability was assessed by computing the omega and omega
hierarchical coefficients. The bifactor model, which accounted for the covariance among HADS items with
regard to 1 general factor (psychological distress) and 2 specific factors (depression and anxiety), described
the HADS structure better than the original oblique 2-factor model during both study periods. All latent
dimensions of the bifactor model were temporally stable. The SEM analysis revealed a significant link
between psychological distress and NA as well as between depression and low PA. Only the general factor
of psychological distress showed adequate reliability. In conclusion, the HADS shows a clear bifactor
structure among FM patients. Our results indicate that it is not recommendable to compute anxiety and
depression scores separately because anxiety variance is tapped primarily by the broader construct of
psychological distress, and both specific dimensions show low reliability.
Keywords: Hospital Anxiety and Depression Scale, negative affect, bifactor model, fibromyalgia,
structural equation modeling
Supplemental materials: http://dx.doi.org/10.1037/a0035284.supp
Fibromyalgia (FM) is a prevalent, debilitating, and chronic syn-
drome of unknown etiology that is primarily characterized by chronic
widespread pain, fatigue, disturbed sleep, and psychological distress
(Wolfe et al., 1990, 2010). Among chronic pain conditions, FM is
associated with the highest unemployment rate (6%), claims for
incapacity benefits (up to 29.9%), and the greatest number of days of
absence from work (Leadley, Armstrong, Lee, Allen, & Kleijnen,
2012). Depression is among the nine symptom domains to be assessed
in FM treatment trials (Mease et al., 2009). According to a recent
review (Boomershine, 2012), the Depression subscale of the Hospital
Anxiety and Depression Scale (HADS; Zigmond & Snaith, 1983)
should be considered the “gold standard” assessment of depression in
patients with FM.
The HADS is a 14-item instrument that was originally devel-
oped to quantify the severity of anxiety and depressive symptoms
in (nonpsychiatric) general hospitals or outpatient clinical settings
(Zigmond & Snaith, 1983). One of the key characteristics of the
HADS is that somatic manifestations (e.g., appetite disturbance,
Juan V. Luciano, Research and Development Unit, Parc Sanitari Sant
Joan de Déu, Sant Boi del Llobregat, Barcelona, Spain; Juan R. Bar-
rada, Department of Psychology and Sociology, Faculty of Human and
Social Sciences, Universidad de Zaragoza, Teruel, Spain; Jaume
Aguado, Research and Development Unit, Parc Sanitari Sant Joan de
Déu, Sant Boi del Llobregat, Barcelona, Spain; Jorge Osma, Depart-
ment of Psychology and Sociology, Faculty of Human and Social
Sciences, Universidad de Zaragoza, Teruel, Spain; Javier García-
Campayo, Department of Psychiatry, Miguel Servet Hospital, Aragon
Institute of Health Sciences, Zaragoza, Spain.
This research was supported in part by Institute of Health Carlos III
Grant PI09/90301 (Madrid, Spain). The authors declare that they have no
competing interests.
Correspondence concerning this article should be addressed to Juan V.
Luciano, Research and Development Unit, Parc Sanitari Sant Joan de Déu.
C/ Doctor Antoni Pujadas 42, 08830, Sant Boi de Llobregat, Barcelona,
Spain. E-mail: jvluciano@pssjd.org
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Psychological Assessment © 2013 American Psychological Association
2013, Vol. 26, No. 1, 000 1040-3590/13/$12.00 DOI: 10.1037/a0035284
1
sleep problems, fatigue, and so on) of emotional disorders were
deliberately excluded to avoid an overdiagnosis of anxiety and
depression among patients with physical conditions. The HADS
measures symptoms over the preceding week and includes seven
interrelated items per subscale to assess symptoms of anxiety
(HADS-A) and depression (HADS-D). Each item is answered
using a 4-point (from 0 to 3) Likert-type scale. The HADS contains
both positively and negatively formulated items to reduce acqui-
escent bias. Higher scores indicate more severe symptoms.
Over the last two decades, the psychometric properties and
diagnostic accuracy of the HADS have been extensively examined
in different populations, including patients with musculoskeletal
complaints (Pallant & Bailey, 2005) and those with FM (Vallejo,
Rivera, Esteve-Vives, & Rodríguez-Muñoz, 2012). Specifically,
Pallant and Bailey (2005) found that, with one minor adjustment
(removal of Item 7), the HADS has a two-factor structure and may
be suitable for use as a screening instrument in rehabilitation
settings with musculoskeletal patients. More recently, Vallejo et al.
(2012) found that the HADS possesses adequate reliability and is
a valid instrument for assessing anxiety and depression in FM
patients. In 2002, 71 articles were reviewed to obtain information
about the factor structure, internal consistency, case-finding abil-
ity, and construct validity of the HADS in clinical and nonclinical
samples (Bjelland, Dahl, Haug, & Neckelmann, 2002). With re-
gard to its internal structure, most of the reviewed studies used a
principal component analysis, and the presence of two correlated
factors (mean r⫽.56) was the most common result. In a recent
meta-analysis (Brennan, Worrall-Davies, McMillan, Gilbody, &
House, 2010), the authors concluded that, although the HADS is a
useful screening tool, it is not superior to other instruments (e.g.,
the Beck Depression Inventory) with regard to identifying emo-
tional disorders in populations with physical conditions.
The dimensionality of the HADS is a current source of contro-
versy among experts (Coyne & van Sonderen, 2012a, 2012b;
Norton, Sacker, & Done, 2012). A recent update (Cosco, Doyle,
Ward, & McGee, 2012) of Bjelland et al.’s study systematically
reviewed 50 articles in which the latent structure of the HADS was
examined. These authors concluded that the structure remains
unclear and that the subscale scores of the HADS should be
interpreted with caution.
Recently, two studies evaluated the internal structure of the
HADS using a confirmatory bifactor analysis ([CBFA] also known
as a nested-factor model; Norton, Cosco, Doyle, Done, & Sacker,
2013; Xie et al., 2012). In CBFA, two types of latent factors are
defined: The first is a general factor in which all items are allowed
to load; the second is composed of specific factors in which the
items are distributed by their content. In the case of the HADS, and
in accordance with the Clark and Watson’s tripartite model (Clark
& Watson, 1991), the general factor represents the shared compo-
nent of anxiety and depression (i.e., a general distress or negative
affect factor), whereas the specific factors (depression and anxiety
after partialing out the general negative affect factor) represent low
positive affect (for depression) and hyperarousal (for anxiety). In
common CBFA, all factors are mutually uncorrelated. CBFA not
only helps to evaluate the internal structure of the instruments,
commonly showing a superior fit to nonnested structures (Simms,
Gras, Watson, & O’Hara, 2008), but also allows assessment of the
reliability of the scores derived from the different factors. As a
result, CBFA is used to determine whether the computation of the
subscale scores is justifiable or whether only the total score should
be computed and reported (Chen, West, & Sousa, 2006).
Xie and colleagues’ CBFA yielded a better fit than other non-
nested confirmatory factor analysis (CFA) models (Xie et al.,
2012). The loadings for the general distress factor were larger than
those for the specific (anxiety/depression) factors with regard to
most items. Interestingly, these authors found that the general
factor (but not anxiety or depression) was significantly correlated
with pain severity. In other words, the general factor completely
accounted for the association between pain and depression/anxiety.
Norton and collaborators re- and meta-analyzed data from 21
previous studies (Norton et al., 2013). Specifically, the eight
commonly reported “best fitting” latent structures were reanalyzed
using CFA, and two bifactor structures, one with two group factors
(anxiety and depression) and another with three group factors
(depression, anxiety, and restlessness), were analyzed. The within-
sample analyses and the subsequent meta-CFA both revealed that
the bifactor model with the general distress factor and two group
factors provided the best fit. The inclusion of a factor that ac-
counted for the wording of items also generally improved model
fit. The substantive weight of the general distress factor in the
HADS was demonstrated by the high amount of common variance
that it explained (ⱖ70% with and without an item-wording factor)
and the moderate to high factor loadings for each item of the
general factor. Therefore, the authors advised against using the
HADS in clinical practice when the objective is to provide a
specific analysis of anxiety or depression.
As noted above, the HADS has enormous clinical relevance in
the assessment of FM patients because it is currently considered
the “gold standard” for evaluating depression among these patients
(Boomershine, 2012). In our opinion, it is of crucial importance to
perform an exhaustive examination of the dimensionality and other
psychometric properties of the HADS in order to confirm that it is
really a good choice to measure psychological distress as well as
to differentiate the symptoms of depression and anxiety in FM
patients. With this goal in mind, we carried out a study with three
main objectives. First, a psychometric perspective was taken to
evaluate the internal structure of the HADS using CBFA. For the
first time, the bifactor approach was fitted by longitudinally mod-
eling the structure of the HADS. Specifically, we evaluated
whether the HADS could be longitudinally modeled using a gen-
eral factor of psychological distress (negative affect), as measured
by all instrument items, and two specific factors (anxiety and
depression), as measured by two item subsets. This autoregressive
structural equation modeling (SEM) analysis was essential to
determine the stability in the general and specific factors over time.
Taking previous findings into account (Norton et al., 2013; Xie et
al., 2012), we expected the bifactor model to be the best fitting
model for both study periods (Hypothesis 1) as well as in the
autoregressive SEM analysis (Hypothesis 2). It was hypothesized
that all latent variables would be highly stable across time (Hy-
pothesis 3). Second, we evaluated the reliability of the anxiety and
depression scores beyond the reliability provided by the general
factor. Third, given the theoretical view of mood and anxiety
disorder research, we further tested Clark and Watson’s (1991)
tripartite model. This model provides clear expectations regarding
the relationships among the latent factors and the measures of
positive and negative affect. To our knowledge, only one previous
study has used a bifactor model to test this theory (Simms et al.,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
2008). However, this study only analyzed the relationships be-
tween the general distress factor and other constructs and not the
correlations between the specific factors and theoretically relevant
variables. In the present work, we examined relationships for each
of the latent HADS factors with positive and negative affect
(Positive and Negative Affect Schedule [PANAS]), considered
usual criterion variables in the domains of anxiety, depression, and
psychological distress. Specifically, we expected that the latent
factor of psychological distress (HADS) and the latent factor of
negative affect (PANAS-NA) would strongly overlap given the
strong theoretical redundancy between these constructs (Hypoth-
esis 4). According to the tripartite theory (Clark & Watson, 1991),
low PA is similar to anhedonia; thus, we expected a stronger
correlation between PA and depression than between PA and
anxiety (Hypothesis 5). The specific factors of depression and
anxiety did not contain a general distress variance because it had
been partialed out in the CBFA. Therefore, we expected that these
specific factors would show nonsignificant or weak correlations
with the PANAS-NA (Hypothesis 6).
Method
Design
In the present work, we used the data set of a 1-year, two-wave
longitudinal multicenter study, whose main aim was to assess the
predictive validity of some pain-related psychological constructs
on pain and quality of life in patients with FM receiving treatment
as usual. A detailed description of the study protocol can be found
elsewhere (Maurel et al., 2011).
Setting and Sample
The potential participants comprised 312 patients with FM
recruited by general practitioners from 24 primary care health
centers in Aragon (Spain). To be included in the study, patients
had to fulfill the 1990 American College of Rheumatology criteria
for FM (Wolfe et al., 1990) according to a diagnosis made by a
rheumatologist and sign an informed consent. The exclusion cri-
teria were the presence of physical/mental conditions that would
impede the patient from accurately answering the battery of mea-
sures, involvement in any compensation claims, and poor knowl-
edge of the Spanish language. Two clinical psychologists admin-
istered a battery of instruments (including the HADS and PANAS)
at Time 1 (T1). The instruments were completed during the visit in
which they were assessed at the hospital to confirm the FM
diagnosis. At the 1-year follow-up assessment (Time 2 [T2]),
another two clinical psychologists administered the HADS to the
same patients in the hospital (the PANAS was not administered at
T2). All patients received treatment-as-usual (standard care). The
treatment provided in Spain is mainly pharmacological and ad-
justed to the symptomatic profile of the fibromyalgic patient. In
addition, doctors received the Consensus for the Treatment of
Fibromyalgia accepted by the Spanish Health Ministry. Data col-
lection was conducted between January 2009 and June 2011.
Before providing informed consent, patients were given a general
overview of the study aims and characteristics. The study followed
Helsinki Convention norms and subsequent updates. The Ethical
Review Board of the regional health authority approved the study
protocol.
Of the 312 initially recruited patients, 11 (3.5%) were not
referred by their general practitioners (GPs) to a rheumatologist;
thus, their FM diagnosis was not confirmed. Furthermore, 17
patients (5.4%) were excluded because their rheumatologist’s di-
agnosis was not FM. In addition, two patients (0.6%) were ex-
cluded for a severe Axis-I psychiatric disorder (opioid-use disor-
der), four patients (1.2%) were excluded for not understanding
Spanish, and nine patients (2.8%) withdrew from the study. There-
fore, at T1, the final sample comprised 269 FM patients. Table 1
displays participant characteristics at T1. One year later (T2), 209
FM patients completed the HADS again, and, as such, the response
rate was 77.7%. This rate is in line with that reported in a recent
longitudinal observational study (Robinson et al., in press) carried
out with FM patients (response rate: 70.9% at 1 year; 1,205/1,700
FM patients).
After applying the common cutoff values (Zigmond & Snaith,
1983) for T1 (and T2), 18.2% (25.8%) of the participants were
classified as possible cases (eight ⱕHADS-A ⱕ10) of anxiety,
and 53.9% (45.5%) were classified as probable cases (11 ⱕ
HADS-A ⱕ21) of anxiety. In addition, 25.7% (22.0%) of the
participants were classified as possible cases (eight ⱕHADS-D ⱕ
10) of depression, and 25.3% (22.5%) were classified as probable
cases (11 ⱕHADS-D ⱕ21) of depression. The prevalence of
possible and probable cases of anxiety and depression in this
Table 1
Demographic and Clinical Characteristics of the Study Sample
(n ⫽269 for T1; n ⫽209 for T2)
Variable Value
Gender, n(females %) 257 (95.5%)
Age, M(SD; range) 52.13 (8.56; 31–70)
Years from FM diagnosis, M(SD; range) 7.73 (5.12; 1–30)
Marital status, n(%)
Married/relationship 199 (74.0%)
Single 24 (8.9%)
Separated/divorced 35 (13%)
Widowed 11 (4.1%)
Educational level, n(%)
No studies 8 (3.0%)
Primary school 126 (46.8%)
Secondary school 102 (37.9%)
University 33 (12.3%)
Employment status, n(%)
Homemaker 34 (12.6%)
Unemployed 42 (15.6%)
Paid employment 68 (25.3%)
Paid employment but on sick leave 34 (12.6%)
Retired/pensioner 36 (13.4%)
Permanent disability 55 (20.4%)
Study measures, M(SD)
HADS-A T1 (0–21) 10.75 (4.97)
HADS-D T1 (0–21) 7.71 (4.66)
HADS-A T2 (0–21) 10.38 (4.64)
HADS-D T2 (0–21) 7.25 (4.34)
PANAS-PA (0–50) 25.15 (8.43)
PANAS-NA (0–50) 24.07 (8.96)
Note.T1⫽Time 1; T2 ⫽Time 2; FM ⫽fibromyalgia; HADS ⫽
Hospital Anxiety and Depression Scale; A ⫽Anxiety; D ⫽Depression;
PANAS ⫽Positive and Negative Affect Schedule; PA ⫽positive affect;
NA ⫽negative affect.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
BIFACTOR ANALYSIS OF THE HADS IN FIBROMYALGIA
primary health care sample is comparable to that reported by
Vallejo et al. (2012) in Spanish patients with FM recruited from
rheumatology clinics (anxiety: possible cases ⫽22.6%, probable
cases ⫽57.1%; depression: possible cases ⫽21.2%, probable
cases ⫽34.9%).
Measures
Participants completed a sociodemographic questionnaire, the
HADS, and the PANAS (Watson, Clark, & Tellegen, 1988) as part
of a paper-and-pencil battery of instruments.
Sociodemographic and clinical data. The following sociode-
mographic information was collected: gender, age, marital status,
education level, and employment status. The clinical variables
were years of FM diagnosis, pharmacological and nonpharmaco-
logical treatments, and referrals to specialized settings (rheuma-
tology or pain clinics).
The HADS (Zigmond & Snaith, 1983). The Spanish version
of the HADS was used (Herrero et al., 2003; Quintana et al., 2003;
Tejero, Guimera, Farré, & Peri, 1986). Herrero et al. (2003) and
Quintana et al. (2003) assessed the psychometric properties of the
Spanish version in 385 adult general hospital outpatients with
severe medical conditions and in 429 patients with five different
diagnoses (and 256 controls), respectively. Both studies conducted
principal component analysis and concluded that the Spanish
HADS contains two factors and has sound psychometric proper-
ties. More recently, the HADS demonstrated to be a reliable (␣of
0.83 and 0.87 for HADS-A and HADS-D, respectively) and useful
tool for assessing the presence of anxiety and depression symp-
toms in a sample of 301 FM patients (Vallejo et al., 2012).
The PANAS (Sandín et al., 1999; Watson et al., 1988). The
PANAS is a 20-item self-report measure that includes two 10-item
mood scales: one for positive affect (PA; i.e., the extent to which
a person experiences pleasurable engagement with the environ-
ment) and one for negative affect (NA; i.e., the extent to which a
person feels distressed, upset, guilty, and so on). Respondents rate
the extent to which they experienced each emotion within a spec-
ified time period (in the present study, e used the “in general” time
frame) using a 5-point scale ranging from 1 (very slightly or not at
all)to5(very much). The PANAS adopts a dimensional approach
to affective states. The PA and NA scales were originally devel-
oped to assess orthogonal (i.e., independent) dimensions of affec-
tive experience rather than opposite poles of the same construct.
The psychometric characteristics of the PANAS have received
much investigative attention (Crawford & Henry, 2004; Sandín et
al., 1999; Tuccitto, Giaccobi, & Leite, 2010). Furlong, Zautra,
Puente, López-López, and Valero (2010) reported an internal con-
sistency (␣) of 0.87 for the PA scale and 0.85 for the NA scale in
a sample of 130 Spanish FM patients recruited in a Spanish pain
clinic.
Data Analyses
SPSS Version 19.0 and Mplus Version 7.0 (Muthén & Muthén,
1998–2012) were used to conduct the data analyses.
Dimensionality analyses. First, with regard to the HADS,
two cross-sectional factor models were tested at T1 and T2: (a) a
CFA using the original two-factor model proposed by Zigmond
and Snaith (1983) and (b) a CBFA model positing that all items are
saturated with a general latent factor of psychological distress and
two specific factors of anxiety and depression that are uncorre-
lated.
Second, autoregressive models allow researchers to examine
relationships over time, where each time point is linearly predicted
by the previous time point. In the present study, two autoregressive
SEM analyses were computed and compared (correlated two-
factor vs. bifactor) to examine the association between each latent
variable at T1 and its T2 counterpart across a 1-year time interval.
The uniqueness of the repeated measures of the items was allowed
to correlate. Only the homotraits from T1 were allowed to predict
traits on T2; that is, for instance, depression at T1 was linked with
depression at T2, but the weights from depression at T1 to anxiety
and the general factor at T2 were fixed to zero. A standardized
weight close to one (which, with single predictors equals a corre-
lation coefficient) indicates that each score can be precisely pre-
dicted from its previous value, meaning that it is highly stable over
time and hence depends on substantive factor. Correlations be-
tween T1 and T2 for each latent factor represent stability coeffi-
cients.
Third, the original two-factor model of the PANAS proposed by
Watson and Clark (Watson et al., 1988), with some modifications
(Crawford & Henry, 2004), was tested. The items drawn from the
same category of Zevon and Tellegen’s mood checklist were
allowed to covary (Zevon & Tellegen, 1982). For PA, these item
groups included (a) attentive, interested, and alert; (b) enthusiastic,
excited, and inspired; (c) proud and determined; and (d) strong and
active. For NA, these items groups included (a) distressed and
upset, (b) hostile and irritable, (c) scared and afraid, (d) ashamed
and guilty, and (e) nervous and jittery. In addition, the PA and NA
factors were allowed to correlate. This model provided the best fit
in a large sample of the general adult population in the United
Kingdom (Crawford & Henry, 2004).
Construct validity analysis. SEM allows researchers to test
the empirical link between all the HADS and PANAS latent
dimensions. Using the T1 data, we examined the correlations
between all the latent variables of the best fitting HADS factor
model and the PANAS two-factor model to examine the pattern of
relationships.
Given that factor loadings and standard errors are underesti-
mated when using maximum likelihood estimation with categori-
cal variables (Wang & Cunningham, 2005), we applied mean- and
variance-corrected weighted least squares (WLSMV) to test the fit
of the alternative factor structures of the HADS and PANAS as
well as for the structural equation models described above. In
addition to the chi-square test, the following fit indices were
analyzed (the values in parentheses denote goodness-of-fit stan-
dards according to Schermelleh-Engel, Moosbrugger, & Müller,
2003): the Tucker–Lewis Index (TLI), the comparative fit index
(TLI and CFI ⱖ.95 indicate an acceptable fit, and ⱖ.97 indicate
agood fit), and the root-mean-square error of approximation
(RMSEA) with 90% confidence intervals (RMSEA ⱕ.08 indi-
cates an acceptable fit and ⱕ.05 indicates a good fit). With regard
to missing data, a total of 269 and 209 FM patients had complete
HADS data at T1 and T2, respectively. For the longitudinal mod-
els, the default approach in Mplus to handle missing data with the
WLSMV estimator was used (Asparouhov & Muthén, 2010).
Reliability estimates. When computing CBFA models, two
types of reliability estimates can be computed: omega and omega
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
hierarchical (Brunner, Nagy, & Wilhelm, 2012). Omega refers to
the reliability of a summed score given all the factors that comprise
that score. Omega hierarchical, which can be equal to or smaller
than omega, refers to the reliability of a summed score composed
of only one construct. With regard to the general factor, the
difference between omega and omega hierarchical provides infor-
mation regarding the reliability of the total score derived from its
specific factors. For the specific factors, omega hierarchical pro-
vides information regarding the capacity of the subscale scores to
reliably measure the variance due to the specific factors by them-
selves, beyond reliability provided by the general factor. Low
omega hierarchical values advise against the use of subscale
scores.
Results
Descriptive Statistics
Descriptive statistics were computed for all the HADS items in
both study periods (see Table 2).
Dimensionality Analyses
Cross-sectional HADS models. The chi-square test revealed
that the tested models did not fit the data. As expected, however,
the other fit indices for the bifactor model (Model 2) were better
than those obtained for the original oblique two-factor model
Table 2
Mean (M), Standard Deviation (SD), and Factor Loadings () for All HADS Items in T1 and T2
HADS Spanish version
a
Time 1 (n⫽269) Time 2 (n⫽209)
M1 M2 M1 M2
M(SD)general specific M(SD)general specific
HADS-Anxiety
1. Me siento tenso/a o nervioso/a [I feel tense
or “wound up”] 1.84 (0.87) .67 .65 .18 1.72 (0.84) .78 .70 .31
3. Siento una especie de temor como si algo
malo fuera a suceder [I get a sort of
frightened feeling as if something awful is
about to happen] 1.23 (1.14) .73 .75 ⫺.10 1.22 (1.06) .72 .68 .21
5. Tengo la cabeza llena de preocupaciones
[Worrying thoughts go through my mind] 1.71 (1.01) .78 .78 .08 1.60 (0.97) .74 .76 .07
7. Soy capaz de permanecer sentado/a
tranquilo/a y relajado/a [I can sit at ease
and feel relaxed] 1.51 (0.94) .61 .55 .46 1.45 (0.87) .56 .46 .36
9. Experimento una desagradable sensación de
« nervios y hormigueos » en el estómago [I
get a sort of frightened feeling like
“butterflies” in my stomach] 1.43 (0.92) .70 .68 .14 1.51 (0.95) .67 .62 .25
11. Me siento inquieto/a como si no pudiera
parar de moverme [I feel restless as if I
have to be on the move] 1.61 (0.91) .68 .62 .52 1.45 (0.82) .59 .38 .92
13. Experimento de repente sensaciones de
gran angustia o temor [I get sudden feelings
of panic] 1.48 (0.97) .89 .91 ⫺.11 1.44 (0.98) .88 .85 .21
HADS-Depression
2. Sigo disfrutando de las cosas como siempre
[I still enjoy the things I used to enjoy] 1.19 (0.82) .82 .58 .60 1.02 (0.72) .84 .49 .74
4. Soy capaz de reírme y ver el lado gracioso
de las cosas [I can laugh and see the funny
side of things] 0.85 (0.87) .83 .63 .54 0.88 (0.84) .88 .62 .61
6. Me siento alegre [I feel cheerful] 1.04 (0.85) .82 .58 .60 0.93 (0.81) .85 .56 .66
8. Me siento lento/a y torpe [I feel as if I’m
slowed down] 2.07 (0.92) .66 .56 .24 1.92 (0.87) .54 .37 .38
10. He perdido el interés por mi aspecto
personal [I have lost interest in my
appearance] 0.88 (0.93) .70 .54 .43 0.93 (0.90) .67 .55 .36
12. Espero las cosas con ilusión [I look
forward to enjoyment to things] 1.02 (0.95) .87 .61 .65 1.01 (0.94) .78 .43 .71
14. Soy capaz de disfrutar con un buen libro o
con un buen programa de radio o televisión
[I can enjoy a good book or radio or TV
programme] 0.75 (0.97) .71 .58 .36 0.55 (0.83) .76 .63 .36
Note.T1⫽Time 1; T2 ⫽Time 2; HADS ⫽Hospital Anxiety and Depression Scale; M1 ⫽Model 1; M2 ⫽Model 2. Nonsignificant (p⬎.05) factor
loadings are shown in italics.
a
The original wording of items in English is shown in brackets.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
BIFACTOR ANALYSIS OF THE HADS IN FIBROMYALGIA
(Model 1) over both T1 and T2, which provides strong support for
the adequacy of the bifactor model with regard to the sample data
and supports Hypothesis 1. As can be seen in Table 3, unlike the
correlated two-factor model, the bifactor structure exhibited
“good” goodness-of-fit values (CFI and TLI ⱖ.97 and RMSEA ⱕ
.05) in both study periods.
The standardized factor loadings obtained in the bifactor and the
two-factor models during each study period are displayed in Table
2. The interfactor correlations for the CFA models were .74 and
.63 for T1 and T2, respectively. The item loadings on the general
factor of the bifactor model were all moderate to large at T1 (M⫽
.64, range ⫽.54–.91) and T2 (M⫽.58, range ⫽.37–.85). The
results notably differed between anxiety and depression with re-
gard to the specific factors. For anxiety, five of the seven items (1,
3, 5, 9, and 13) at T1 presented nonsignificant loadings on their
specific factor, and the factor loadings ranged from –.11 to .17
(unsigned M⫽.17); Item 5 at T2 did not significantly load onto
the anxiety specific factor (p⫽.44) with factor loadings within the
.07–.93 range (M⫽.33). The loading on the general factor was
higher than that on the specific factor for 13 of the 14 anxiety items
(seven items each at T1 and T2). In contrast, all factor loadings on
the specific depression factor were moderate to large and reached
significance at T1 (M⫽.49, range ⫽.24–.65) and T2 (M⫽.55,
range ⫽.36–.74). The loadings of the 14 depression items were
greater for the specific factor than the general factor in half of the
comparisons.
For T1 (and T2), the general factor explained 71.5% (58.6%) of
the common variance, whereas the specific factors explained
28.5% (41.4%). When only the anxiety items were considered, the
specific factor accounted for 13.6% (29.3%) of the common vari-
ance, whereas the specific factor explained 36.5% (47.5%) of this
common variance with regard to the depression items. These
results suggest that the depression-specific factor reflects residual
variance not accounted for by the general factor, whereas the
general factor mostly explains the anxiety variance.
Autoregressive HADS models. Again, the chi-square test did
not support the models. As a whole, however, the SEM analysis
indicated that the bifactor model provided a better fit to the data
than the two-factor model (see Table 3). Both models had fit
indices over the recommended thresholds. Model comparisons
based on a practical improvement to the model-fit approach (TLI
difference ⱖ.01; Gignac, 2007) provided strong support for the
bifactor model (TLI ⫽.99 vs. .97). The superiority of the bifactor
model provided support to Hypothesis 2.
The standardized factor loadings and the relationships among
the latent factors for the autoregressive bifactor model are pre-
sented in Figure 1. These factor loadings were consistent with
those obtained in the cross-sectional models, and the same com-
ments also apply here. The lagged relationships were high for the
general factor and the specific depression factor (.82 and .80,
respectively) but moderate for the anxiety-specific factor (.64).
Our results partially support Hypothesis 3 because the specific
anxiety dimension of the HADS had adequate (not high) stability
over time.
Construct Validity using Clark and Watson’s (1991)
Tripartite Theory as a Framework
Although the PANAS two-factor model had a poor fit according
to the chi-square test, the other fit indices were acceptable (RMSEA ⫽
.07, TLI ⫽.96, and CFI ⫽.97). The item loadings with regard to
their respective factors were high (PA and NA ranges from .66 to
.88 and from .71 to .81, respectively). Although the correlation
between the NA and PA factors was moderate (r⫽–.56, p⬍
.001), this value is clearly greater than the commonly reported
correlation between affectivity dimensions and the theoretically
expected relation.
Second, an SEM analysis was computed to assess the construct
validity of the HADS bifactor structure. As shown in Table 3, all
fit indices were within acceptable limits (CFI and TLI ⱖ.95 and
RMSEA ⱕ.08). The pattern of the relationships between the latent
factors of the HADS and PANAS is shown in Figure 2. Both PA
and NA were significantly associated with the general psycholog-
ical distress factor of the HADS but in opposite directions (–.63
and .81, respectively). As Hypothesis 4 predicted, the highest
correlation was found between NA and the HADS general factor.
Table 3
Fit Statistics for Latent Structure Models for HADS and PANAS Data in FM Patients
Model
2
df RMSEA [90% CI] TLI CFI
HADS (Cross-sectional - T1)
M1 Correlated two-factor model (Zigmond & Snaith, 1983) 167.129 76 .07 [.05, .08] .97 .98
M2 Bifactor model (Norton et al., 2013) 90.437 63 .04 [.02, .06] .99 .99
HADS (Cross-sectional - T2)
M1 Correlated two-factor model (Zigmond & Snaith, 1983) 161.501 76 .07 [.06, .09] .96 .97
M2 Bifactor model (Norton et al., 2013) 81.242 63 .04 [.00, .06] .99 .99
HADS (Autoregressive - T1 & T2)
M1 Correlated two-factor model (Zigmond & Snaith, 1983) 498.598 330 .04 [.04, .05] .97 .98
M2 Bifactor model (Norton et al., 2013) 363.520 305 .03 [.01, .04] .99 .99
PANAS (Cross-sectional - T1)
Correlated two-factor model ⫹covariated errors (Crawford & Henry, 2004) 381.973 156 .07 [.07, .08] .96 .97
Construct validity
HADS (M2) & PANAS 905.318 493 .06 [.05, .06] .96 .96
Note. HADS ⫽Hospital Anxiety and Depression Scale; PANAS ⫽Positive and Negative Affect Schedule; FM ⫽fibromyalgia; RMSEA ⫽
root-mean-square error of approximation; 90% CI ⫽90% confidence interval of the RMSEA; TLI ⫽Tucker–Lewis index; CFI ⫽comparative fit index;
T1 ⫽Time 1; M ⫽model; T2 ⫽Time 2.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
PA was most highly correlated with the depression factor (–.54),
and none of the correlations between NA and the specific factors
were significant, which supports Hypotheses 5 and 6, respectively.
Reliability
According to the CBFA results, the omega estimates for the
total, anxiety, and depression scores of the HADS were the same
for T1 and T2: .94, .90, and .91, respectively. The omega hierar-
chical revealed the following values for the same components
above: .81, .05, and .38 for T1, and .72, .19, and .47 for T2. The
difference between omega and omega hierarchical in both time
periods suggests that there is not a substantial influence of the
specific factors on the reliability of the HADS total score. Fur-
thermore, when the general factor is partialed out, the capacity of
the subscales to reliably measure the variance due to the specific
factors of anxiety and depression is considerably low.
Discussion
As we expected, the HADS bifactor model provided better fit to
the sample data than the original oblique two-factor solution at T1
and T2, which supports Hypothesis 1. Our results are in line with
those obtained from other anxiety/depression instruments, such as
the Inventory of Depression and Anxiety Symptoms (Simms et al.,
2008), the State-Trait Anxiety Inventory-trait version (STAI; Ba-
dos, Gómez-Benito, & Balaguer, 2010), the Revised Children’s
Manifest Anxiety Scale (Brodbeck, Abbott, Goodyer, & Croudace,
2011), the Hopkins Symptom Checklist 25 (Al-Turkait, Ohaeri,
El-Abbasi, & Nagury, 2011), the Depression Anxiety Stress
Scales-21 (Osman et al., 2012), and the Composite International
Diagnostic Interview (Simms, Prisciandaro, Krueger, & Goldberg,
2012). In all cases, the bifactor model fit the data well and
outperformed the traditional first-order and second-order factor
structures.
The dearth of scientific literature concerning the stability of the
latent dimensions of the HADS prioritized this research topic. The
time between measures (1 year) must be considered when evalu-
ating the findings reported above. Our study is the first to dem-
onstrate that the HADS general and depression latent factors have
high stability over time, whereas the stability of the anxiety factor
is moderate over time. As expected (Hypothesis 2), the fit indices
for the autoregressive bifactor model were better than those ob-
tained for the autoregressive two-factor model. The difference in
temporal stability between the specific factors is very difficult to
interpret due to the absence of previous studies addressing this
issue in clinical or nonclinical samples. Any clinical interpretation
(e.g., greater temporal oscillation of anxiety symptoms compared
with depression symptoms in FM patients) of these findings might
be considered speculative.
Recently, several authors have recommended abandoning the
use of the HADS (Coyne & van Sonderen 2012a, 2012b; Zakrze-
wska, 2012). Our results offer some insight concerning the valid
inferences about the HADS scores, both for the total and subscale
scores. The total score is a reliable measure of negative affect,
although a small portion of variance can be attributed to specific
factors. Negative affect is a core component of both mood and
Figure 1. Stability over time (1 year) of the Hospital Anxiety and Depression Scale bifactor structure.
Nonsignificant factor loadings are given in italics and indicated by dashed lines.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
BIFACTOR ANALYSIS OF THE HADS IN FIBROMYALGIA
anxiety disorders. Due to the omega hierarchical values of this
study, clinicians and researchers should understand that they are
interpreting measures of negative affect when anxiety and depres-
sion scores are reported and interpreted. Given our results, it is not
surprising that the total score can be more useful in detecting
specific disorders compared with the subscale scores because
increasing test length improves reliability. If the correct detection
of anxiety and depression is required, then longer tests with high-
loading items on the general and specific factors are needed
(Watson, 2009). The HADS is a 30-year-old measure, and we
cannot expect it to contain all the theoretical and assessment
updates regarding depression and anxiety that have been made
since 1983.
The present work also provided support for the construct
validity of the bifactor solution of the HADS in patients with
FM by exploring relationships between its latent factors and PA
and NA as measured by the PANAS. Our analysis confirmed
the HADS bifactor structure using Clark and Watson’s (1991)
tripartite theory of anxiety and depression as a reference. We
demonstrated that the variance of the HADS items can be
divided into three components: specific hyperarousal (anxiety),
specific low-positive affect (depression), and a shared compo-
nent of general negative affect, which is considered a “some-
what ‘muddy’ area in the middle where symptoms common to
both syndromes overlap” (Roberts, Bonnici, MacKinnon, &
Worcester, 2001, p. 380). Therefore, the present SEM analysis
was the first to test this theory by considering a CBFA model
with specific factors. We found the expected strong positive
association between general psychological distress and NA
(Hypothesis 4). Although these instruments differ in response
format and content, the general factor from the HADS and the
NA component from the PANAS are equivalent constructs. The
expected correlation between depression and low PA was ob-
tained (Hypothesis 5). In addition, we found that specific fac-
tors were not correlated with NA (Hypothesis 6). Finally, we
have found, for the first time, that PA and anxiety are signifi-
cantly and positively associated (.13) when psychological dis-
tress is disentangled from anxiety. In other words, PA and
anxiety usually correlate negatively (see, e.g., Crawford &
Henry, 2004) because of the influence of the general factor of
psychological distress. In our opinion, this counterintuitive
finding deserves to be replicated in future studies.
Figure 2. Construct validity of the Hospital Anxiety and Depression Scale (HADS) bifactor structure.
Nonsignificant correlations, factor loadings, and covariated errors are given in italics and indicated by dashed
lines. The left part of the figure shows Positive and Negative Affect Schedule (PANAS) items, and the right part
shows HADS items. PA ⫽positive affect; NA ⫽negative affect.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
Although we discussed our data with regard to Clark and Wat-
son’s (1991) theoretical framework, other theoretical models might
also explain the results of the SEM analysis, such as Brown and
Barlow’s (2009) classification of emotional disorders. These au-
thors emphasized the similarities of anxiety and mood disorder
psychopathologies and gave them a common classification: emo-
tional disorders (see also Brown, 2007). Furthermore, they stated
that the similar psychopathologies are due to two genetic temper-
ament dimensions that determine the etiology and course of emo-
tional disorders: neuroticism/negative affect and extraversion/pos-
itive affect (Brown, 2007; Brown, Chorpita, & Barlow, 1998).
Although the sample sizes for both study periods were rela-
tively small, our results might be generalizable because the
sociodemographic characteristics of the study sample were very
similar to those of FM individuals from the general population
in Spain (Mas, Carmona, Valverde, Ribas, & EPISER Study
Group, 2008), which provides a certain degree of external
validity. Moreover, the findings reported here have theoretical
and clinical implications for the assessment of patients with FM. In
the upcoming primary health care version of the International Clas-
sification of Diseases-11 (the ICD-11-PHC), FM will be classified as
part of bodily stress syndrome (BSS; Lam et al., 2013). This new
diagnosis will group patients who might have previously been
considered different (e.g., those with FM, chronic fatigue syn-
drome, irritable bowel syndrome, and so on). Frontline clinicians
(e.g., GPs) will need reliable tools to identify possible/probable
clinical cases of anxiety (i.e., cognitive overarousal) among pa-
tients with BSS who are characterized by elevated somatic over-
arousal. Unless the HADS is refined in the future, this instrument
does not seem suitable to this task, given that its Anxiety subscale,
which was created to measure cognitive overarousal, is unspecific
and almost completely accounted for within the broader construct
of general distress. In our opinion, the anxiety and depression
items recently developed by the Patient-Reported Outcomes Mea-
sure Information System (PROMIS) Cooperative group (Pilkonis
et al., 2011), particularly their short-form scales, are a better option
due to their excellent internal consistency and construct validity.
Importantly, the rationale-derived option of excluding somatic
items (Zigmond & Snaith, 1983) was supported for these scales
after a careful psychometric analysis with hundreds of available
items. As far as we know, the U.S. National Institutes of Health is
currently performing a field test of PROMIS item banks on ⬎
3,500 FM patients. If PROMIS item banks demonstrate to be valid
and reliable in FM patients, it is possible that they will progres-
sively replace the “legacy” instruments—the Beck Anxiety Inven-
tory (Beck, Epstein, Brown, & Steer, 1988), Beck Depression
Inventory-II (Beck, Steer, & Brown, 1996), Center for Epidemio-
logical Studies Depression Scale (Radloff, 1977), HADS, Patient
Health Questionnaire (Kroenke & Spitzer, 2002), and STAI—that
are currently used worldwide to assess emotional symptoms in
these patients (D. A. Williams & Schilling, 2009).
The following limitations should be considered when inter-
preting our findings. The conclusions based on the present data
must be considered preliminary until more factor analytic stud-
ies of the HADS in patients with FM are conducted. The
reported correlation between PA and NA is much stronger than
the typical value, and the expected relationship between these
constructs is approximately orthogonal. Nevertheless, this re-
sult might be explained by the high presence of probable cases
of anxiety among our FM patients at both study periods (53.9%
at T1 and 45.5% at T2). According to J. Williams, Peeters, and
Zautra (2004), people suffering from anxiety are characterized
by their hypervigilance against possible threat, which imposes
increased cognitive demands. These attentional demands on
patients suffering from anxiety reduce the availability of cog-
nitive resources needed to evaluate PA and NA separately,
leading to a higher inverse correlation. As these authors point
out, “One would expect, therefore, to find a significant inverse
correlation between PA and NA factors in anxious populations”
(Williams et al., 2004, p. 323). Another limitation that is
important to acknowledge is the 23.3% of participants missing
at T2. Missing data are very usual in prospective research
among FM patients (e.g., Robinson et al., in press) and poses a
threat to internal (loss of statistical power) and external (gen-
eralizability) validity. In addition, certain important psychomet-
ric aspects of the HADS with regard to patients with FM remain
unknown and should be examined in the future—first, whether
the HADS dimensionality varies across male and female FM
patients. In our study, most participants were women (95.5%),
so dimensionality and construct validity results may not fully
apply to men. The second aspect is whether differences in
HADS language translation might introduce variation regarding
its factor structure and item loadings (Maters, Sanderman, Kim,
& Coyne, 2013). Therefore, the HADS bifactor model should be
replicated in a large, multinational sample of patients with FM.
Third, we addressed the potential effect of the methods associ-
ated with the mixture of negatively and positively worded
items. As expected, adding an item wording method factor
slightly improved model fit, compared with the same model
without a method factor.
1
We decided not to include this
analysis as part of our objectives because of the existence of
solid evidence that all viable HADS factor models improve
their fit when a wording factor is incorporated (Norton et al.,
2013; Schönberger & Ponsford, 2010; Wouters, Booysen, Pon-
net, & Baron Van Loon, 2012). An interesting issue that Wout-
ers et al. (2012) highlighted is that not all individuals are
equally affected by item wording and that “different sociocul-
tural populations may respond differently to negatively (or
positively) worded items with poorer, younger and less-
educated people being more susceptible to wording effects” (p.
2). In our opinion, those FM patients with “fibrofog” symptoms
(e.g., forgetfulness, losing the train of thought, mixing up
words, etc.) might constitute a particularly vulnerable group to
item phrasing. However, we were unable to examine the spe-
cific relationship between fibrofog symptoms and method ef-
fects due to sample size limitations.
In conclusion, the present work supports the existence of a
bifactor structure in the HADS. On the one hand, despite the
multidimensionality of the HADS items, the total score reliably
reflects variation on psychological distress if we bear in mind
the omega hierarchical values of the general factor in the
bifactor model (Reise, Bonifay, & Haviland, 2013). Moreover,
1
Correlated two-factor model ⫹Item wording method factor (T1:
2
⫽
141.634; RMSEA ⫽.06 [.05, .08], TLI ⫽.97, CFI ⫽.98; T2: The
Mplusmodel did not converge). Bifactor model ⫹Item wording method
factor (T1:
2
⫽68.537; RMSEA ⫽.03 [.00, .05], TLI ⫽.99, CFI ⫽.99;
T2:
2
⫽49.414; RMSEA ⫽.00 [.00, .03], TLI ⫽1.00, CFI ⫽1.00).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
BIFACTOR ANALYSIS OF THE HADS IN FIBROMYALGIA
it seems reasonable to compute the total score by summing the
ratings on the 14 items given the moderate to large item
loadings on the general factor at both assessment periods. In
other words, no item removal seems necessary. On the other
hand, it does not seem recommendable to compute anxiety and
depression scores separately because anxiety variance is tapped
primarily by the broader construct of psychological distress and
both specific dimensions show low reliability. Notwithstanding,
as Reise et al. (2013) pointed out, even when a unidimensional
measurement model in an instrument might be good enough for
practical purposes, there are occasions in which complex mea-
surement models should be considered; for example, the use of
a complex representation of the HADS structure was fully
justified when the Clark and Watson’s (1991) tripartite theory
was tested owing to the unique association of general (psycho-
logical distress) and group factors (anxiety and depression) with
relevant external criteria (positive and negative affect). Thus,
we have demonstrated that CBFA is a useful technique not only
for analyzing the psychometric properties of instruments but
also for testing relevant theoretically derived hypotheses.
References
Al-Turkait, F. A., Ohaeri, J. U., El-Abbasi, A. H., & Naguy, A. (2011).
Relationship between symptoms of anxiety and depression in a sample
of Arab college students using the Hopkins Symptom Checklist 25.
Psychopathology, 44, 230–241. doi:10.1159/000322797
Asparouhov, T., & Muthén, B. (2010). Weighted least squares estimation
with missing data (Technical report). Retrieved from http://www
.statmodel.com/download/GstrucMissingRevision.pdf
Bados, A., Gómez-Benito, J., & Balaguer, G. (2010). The State-Trait
Anxiety Inventory, trait version: Does it really measure anxiety? Journal
of Personality Assessment, 92, 560–567. doi:10.1080/00223891.2010
.513295
Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory
for measuring clinical anxiety: Psychometric properties. Journal of
Consulting and Clinical Psychology, 56, 893–897.
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). BDI-II: Beck Depression
Inventory–Second Edition manual. San Antonio, TX: Psychological
Corporation.
Bjelland, I., Dahl, A. A., Haug, T. T., & Neckelmann, D. (2002). The
validity of the Hospital Anxiety and Depression Scale: An updated
literature review. Journal of Psychosomatic Research, 52, 69–77. doi:
10.1016/S0022-3999(01)00296-3
Boomershine, C. S. (2012). A comprehensive evaluation of standardized
assessment tools in the diagnosis of fibromyalgia and in the assessment
of fibromyalgia severity. Pain Research and Treatment, 2012, 1–11.
doi:10.1155/2012/653714
Brennan, C., Worrall-Davies, A., McMillan, D., Gilbody, S., & House, A.
(2010). The Hospital Anxiety and Depression Scale: A diagnostic meta-
analysis of case-finding ability. Journal of Psychosomatic Research, 69,
371–378. doi:10.1016/j.jpsychores.2010.04.006
Brodbeck, J., Abbott, R. A., Goodyer, I. M., & Croudace, T. J. (2011).
General and specific components of depression and anxiety in an ado-
lescent population. BMC Psychiatry, 11, 191. doi:10.1186/1471-244X-
11-191
Brown, T. A. (2007). Temporal and structural relationships among dimen-
sions of temperament and DSM-IV anxiety and mood disorder con-
structs. Journal of Abnormal Psychology, 116, 313–328. doi:10.1037/
0021-843X.116.2.313
Brown, T. A., & Barlow, D. H. (2009). A proposal for a dimensional
classification system based on the shared features of the DSM-IV
anxiety and mood disorders: Implications for assessment and treatment.
Psychological Assessment, 21, 256–271. doi:10.1037/a0016608
Brown, T. A., Chorpita, B. F., & Barlow, D. H. (1998). Structural rela-
tionships among dimensions of the DSM-IV anxiety and mood disorders
and dimensions of negative affect, positive affect, and autonomic arous-
al. Journal of Abnormal Psychology, 107, 179–192. doi:10.1037/0021-
843X.107.2.179
Brunner, M., Nagy, G., & Wilhelm, O. (2012). A tutorial on hierarchically
structured constructs. Journal of Personality, 80, 796–846. doi:10.1111/
j.1467-6494.2011.00749.x
Chen, F. F., West, S. G., & Sousa, K. H. (2006). A comparison of bifactor
and second-order models of quality of life. Multivariate Behavioral
Research, 41, 189–225. doi:10.1207/s15327906mbr4102_5
Clark, L. A., & Watson, D. (1991). Tripartite model of anxiety and
depression: Psychometric evidence and taxonomic implications. Journal
of Abnormal Psychology, 100, 316–336. doi:10.1037/0021-843X.100.3
.316
Cosco, T. D., Doyle, F., Ward, M., & McGee, H. (2012). Latent structure
of the Hospital Anxiety and Depression Scale: A 10-year systematic
review. Journal of Psychosomatic Research, 72, 180–184. doi:10.1016/
j.jpsychores.2011.06.008
Coyne, J. C., & van Sonderen, E. (2012a). The Hospital Anxiety and
Depression Scale (HADS) is dead, but like Elvis, there will still be
citings. Journal of Psychosomatic Research, 73, 77–78. doi:10.1016/j
.jpsychores.2012.04.002
Coyne, J. C., & van Sonderen, E. (2012b). No further research needed:
Abandoning the Hospital and Anxiety Depression Scale (HADS). Jour-
nal of Psychosomatic Research, 72, 173–174. doi:10.1016/j.jpsychores
.2011.12.003
Crawford, J. R., & Henry, J. D. (2004). The Positive and Negative Affect
Schedule (PANAS): Construct validity, measurement properties and
normative data in a large non-clinical sample. British Journal of Clinical
Psychology, 43, 245–265. doi:10.1348/0144665031752934
Furlong, L. V., Zautra, A., Puente, C. P., López-López, A., & Valero, P. B.
(2010). Cognitive-affective assets and vulnerabilities: Two factors in-
fluencing adaptation to fibromyalgia. Psychology & Health, 25, 197–
212. doi:10.1080/08870440802074656
Gignac, G. E. (2007). Multifactor modeling in individual differences
research: Some recommendations and suggestions. Personality and In-
dividual Differences, 42, 37–48. doi:10.1016/j.paid.2006.06.019
Herrero, M. J., Blanch, J., Peri, J. M., De Pablo, J., Pintor, A., & Bulbena,
A. (2003). A validation study of the Hospital Anxiety and Depression
Scale (HADS) in the Spanish population. General Hospital Psychiatry,
25, 277–283. doi:10.1016/S0163-8343(03)00043-4
Kroenke, K., & Spitzer, R. L. (2002). The PHQ-9: A new depression
diagnostic and severity measure. Psychiatric Annals, 32, 509–521.
Lam, T. P., Goldberg, D. P., Dowell, A. C., Fortes, S., Mbatia, J. K.,
Minhas, F. A., & Klinkman, M. S. (2013). Proposed new diagnoses of
anxious depression and bodily stress syndrome in ICD-11-PHC: An
international focus group study. Family Practice, 30, 76–87. doi:
10.1093/fampra/cms037
Leadley, R. M., Armstrong, N., Lee, Y. C., Allen, A., & Kleijnen, J.
(2012). Chronic diseases in the European Union: The prevalence and
health cost implications of chronic pain. Journal of Pain & Palliative
Care Pharmacotheraphy, 26, 310–325. doi:10.3109/15360288.2012
.736933
Mas, A. J., Carmona, L., Valverde, M., Ribas, B., & EPISER Study Group.
(2008). Prevalence and impact of fibromyalgia on function and quality
of life in individuals from the general population: Results from a
nationwide study in Spain. Clinical and Experimental Rheumatology,
26, 519–526.
Maters, G. A., Sanderman, R., Kim, A. Y., & Coyne, J. C. (2013).
Problems in cross-cultural use of the Hospital Anxiety and Depression
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
Scale: “No butterflies in the desert.” PLoS ONE, 8(8): e70975. doi:
10.1371/journal.pone.0070975
Maurel, S., Rodero, B., López del Hoyo, Y., Luciano, J. V., Andrés, E.,
Roca, M., & García-Campayo, J. (2011). Correlational analysis and
predictive validity of psychological constructs related with pain in
fibromyalgia. BMC Musculoskeletal Disorders, 12, 4. doi:10.1186/
1471-2474-12-4
Mease, P., Arnold, L. M., Choy, E. H., Clauw, D. J., Crofford, L. J., Glass,
J.M.,...Williams, D. A. (2009). Fibromyalgia syndrome module at
OMERACT 9: Domain construct. Journal of Rheumatology, 36, 2318–
2329. doi:10.3899/jrheum.090367
Muthén, L. K., & Muthén, B. O. (1998–2012). Mplus users guide (7th ed.).
Los Angeles, CA: Muthén & Muthén.
Norton, S., Cosco, T. D., Doyle, F., Done, J., & Sacker, A. (2013). The
Hospital Anxiety and Depression Scale: A meta confirmatory factor
analysis. Journal of Psychosomatic Research, 74, 74–81. doi:10.1016/
j.jpsychores.2012.10.010
Norton, S., Sacker, A., & Done, D. J. (2012). Further research needed: A
comment on Coyne and van Sonderen’s call to abandon the Hospital
Anxiety and Depression Scale. Journal of Psychosomatic Research, 73,
75–76. doi:10.1016/j.jpsychores.2012.04.005
Osman, A., Wong, J. L., Bagge, C. L., Freedenthal, S., Gutierrez, P. M.,
& Lozano, G. (2012). The Depression Anxiety Stress Scales–21
(DASS-21): Further examination of dimensions, scale reliability, and
correlates. Journal of Clinical Psychology, 68, 1322–1338. doi:10
.1002/jclp.21908
Pallant, J. F., & Bailey, C. M. (2005). Assessment of the structure of the
Hospital Anxiety and Depression Scale in musculoskeletal patients.
Health and Quality of Life Outcomes, 3, 82–90. doi:10.1186/1477-7525-
3-82
Pilkonis, P. A., Choi, S. W., Reise, S. P., Stover, A. M., Riley, W. T., &
Cella, D. (2011). Item banks for measuring emotional distress from
the Patient-Reported Outcomes Measurement Information System
(PROMIS®): Depression, anxiety, and anger. Assessment, 18, 263–
283. doi:10.1177/1073191111411667
Quintana, J. M., Padierna, A., Esteban, C., Arostegui, I., Bilbao, A., &
Ruiz, I. (2003). Evaluation of the psychometric characteristics of the
Spanish version of the Hospital Anxiety and Depression Scale. Acta
Psychiatrica Scandinavica, 107, 216–221. doi:10.1034/j.1600-0447
.2003.00062.x
Radloff, L. (1977). The CES-D Scale: A self-report depression scale for
research in the general population. Applied Psychological Measurement,
1, 385–401.
Reise, S. P., Bonifay, W. E., & Haviland, M. G. (2013). Scoring and
modeling psychological measures in the presence of multidimensional-
ity. Journal of Personality Assessment, 95, 129–140. doi:10.1080/
00223891.2012.725437
Roberts, S. B., Bonnici, D. M., MacKinnon, A. J., & Worcester, M. C.
(2001). Psychometric evaluation of the Hospital Anxiety and Depres-
sion Scale (HADS) among female cardiac patients. British Journal of
Clinical Psychology, 6, 373–383. doi:10.1348/135910701169278
Robinson, R. L., Kroenke, K., Williams, D. A., Mease, P., Chen, Y.,
Faries, D.,...McCarberg, B. (in press). Longitudinal observation of
treatment patterns and outcomes for patients with fibromyalgia:
12-month findings from the REFLECTIONS study. Pain Medicine.
doi:10.1111/pme.12168
Sandín, B., Chorot, P., Lostao, L., Joiner, T. E., Santed, M. A., & Valiente,
R. M. (1999). Escalas PANAS de afecto positivo y negativo: Validación
factorial y convergencia transcultural [The PANAS scales of positive
and negative affect: Factor analytic validation and cross-cultural con-
vergence]. Psicothema, 11, 37–51.
Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating
the fit of structural equation models: Tests of significance and descrip-
tive goodness-of-fit measures. Methods of Psychological Research On-
line, 8, 23–74.
Schönberger, M., & Ponsford, J. (2010). The factor structure of the Hos-
pital Anxiety and Depression Scale in individuals with traumatic brain
injury. Psychiatry Research, 179, 342–349. doi:10.1016/j.psychres.2009
.07.003
Simms, L. J., Gras, D. F., Watson, D., & O’Hara, M. W. (2008). Parsing
the general and specific components of depression and anxiety with
bifactor modeling. Depression and Anxiety, 25, E34–E46. doi:10.1002/
da.20432
Simms, L. J., Prisciandaro, J. J., Krueger, R. F., & Goldberg, D. P.
(2012). The structure of depression, anxiety and somatic symptoms in
primary care. Psychological Medicine, 42, 15–28. doi:10.1017/
S0033291711000985
Tejero, A., Guimera, E. M., Farré, J. M., & Peri, J. M. (1986). Uso clínico
del HADS (Hospital Anxiety and Depression Scale) en población
psiquiátrica: Un estudio de fiabilidad, sensibilidad y validez [Clinical
use of the HADS (Hospital Anxiety and Depression Scale) in a psychi-
atric population: A reliability, sensitivity, and validity study]. Revista del
Departamento de Psiquiatría de la Facultad de Medicina de Barcelona,
13, 233–238.
Tuccitto, D. E., Giaccobi, P. R., & Leite, W. L. (2010). The internal
structure of positive and negative affect: A confirmatory factor analysis
of the PANAS. Educational and Psychological Measurement, 70, 125–
141. doi:10.1177/0013164409344522
Vallejo, M. A., Rivera, J., Esteve-Vives, J., & Rodríguez-Muñoz, M. F.
(2012). Uso del cuestionario Hospital Anxiety and Depression Scale
(HADS) para evaluar la ansiedad y la depresión en pacientes con
fibromialgia [Use of the Hospital Anxiety and Depression Scale
(HADS) to evaluate anxiety and depression in fibromyalgia patients].
Revista de Psiquiatría y Salud Mental, 5, 107–114. doi:10.1016/j
.rpsm.2012.01.003
Wang, W. C., & Cunningham, E. G. (2005). Comparison of alternative
estimation methods in confirmatory factor analyses of the General
Health Questionnaire. Psychological Reports, 97, 3–10. doi:10.2466/pr0
.97.1.3-10
Watson, D. (2009). Differentiating the mood and anxiety disorders: A
quadripartite model. Annual Review of Clinical Psychology, 5, 221–247.
doi:10.1146/annurev.clinpsy.032408.153510
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali-
dation of brief measures of positive and negative affect: The PANAS
scales. Journal of Personality and Social Psychology, 54, 1063–1070.
doi:10.1037/0022-3514.54.6.1063
Williams, D. A., & Schilling, S. (2009). Advances in the assessment of
fibromyalgia. Rheumatic Diseases Clinics of North America, 35, 339–
357. doi:10.1016/j.rdc.2009.05.007
Williams, J., Peeters, F., & Zautra, A. (2004). Differential affect structure
in depressive and anxiety disorders. Anxiety, Stress, & Coping, 17,
321–330. doi:10.1080/10615800412331318634
Wolfe, F., Clauw, D. J., Fitzcharles, M. A., Goldenberg, D. L., Katz, R. S.,
Mease, P.,...Yunus, M. B. (2010). The American College of Rheu-
matology preliminary diagnostic criteria for fibromyalgia and measure-
ment of symptom severity. Arthritis Care and Research, 62, 600–610.
doi:10.1002/acr.20140
Wolfe, F., Smythe, H. A., Yunus, M. B., Bennett, R. M., Bombardier, C.,
Goldenberg, D. L.,...Sheon, R. P. (1990). The American College of
Rheumatology 1990 criteria for the classification of fibromyalgia: Re-
port of the Multicenter Criteria Committee. Arthritis and Rheumatism,
33, 160–172. doi:10.1002/art.1780330203
Wouters, E., Booysen Fle, R., Ponnet, K., & Baron Van Loon, F.
(2012). Wording effects and the factor structure of the Hospital
Anxiety and Depression Scale in HIV/AIDS patients on antiretroviral
treatment in South Africa. PLoS One, 7, e34881. doi:10.1371/journal
.pone.0034881
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
BIFACTOR ANALYSIS OF THE HADS IN FIBROMYALGIA
Xie, J., Bi, Q., Li, W., Shang, W., Yan, M., Yang, Y., Miao, D., &
Zhang, H. (2012). Positive and negative relationship between anxiety
and depression of patients in pain: A bifactor model analysis. PLoS
One, 7, e47577. doi:10.1371/journal.pone.0047577
Zakrzewska, J. M. (2012). Should we still use the Hospital Anxiety and
Depression Scale?. Pain, 153, 1332. doi:10.1016/j.pain.2012.03.016
Zevon, M. A., & Tellegen, A. (1982). The structure of mood change: An
idiographic/nomothetic analysis. Journal of Personality and Social Psy-
chology, 43, 111–122. doi:10.1037/0022-3514.43.1.111
Zigmond, A. S., & Snaith, R. P. (1983). The Hospital Anxiety and De-
pression Scale. Acta Psychiatrica Scandinavica, 67, 361–370. doi:
10.1111/j.1600-0447.1983.tb09716.x
Received April 22, 2013
Revision received September 5, 2013
Accepted November 4, 2013 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 LUCIANO, BARRADA, AGUADO, OSMA, AND GARCÍA-CAMPAYO
