Journal of Anxiety Disorders 25 (2011) 604–611
Contents lists available at ScienceDirect
Journal of Anxiety Disorders
Gender differences in the factor structure of posttraumatic stress disorder
symptoms in war-exposed adolescents
Cherie Armoura,b,∗, Jon D. Elhaic, Christopher M. Layned, Mark Shevlinb, Elvira Durakovi´ c-Belkoe,
Nermin Djapoe, Robert S. Pynoosd
aNational Centre of Psychotraumatology, University of Southern Denmark, Odense, Denmark
bSchool of Psychology, Faculty of Life and Health Sciences, University of Ulster at Magee Campus, Londonderry, Northern Ireland BT48 7JL, United Kingdom
cDepartment of Psychology, University of Toledo, Toledo, OH, USA
dUCLA National Center for Child Traumatic Stress, Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, USA
eUniversity of Sarajevo, Sarajevo, Bosnia and Herzegovina
a r t i c l ei n f o
Received 7 June 2010
Received in revised form 28 January 2011
Accepted 28 January 2011
a b s t r a c t
DSM-IV’s three-factor model of posttraumatic stress disorder (PTSD) is rarely empirically supported,
whereas other four-factor models (King et al., 1998; Simms, Watson, & Doebbeling, 2002) have proven
to be better representations of PTSD’s latent structure. To date, a clear consensus as to which model
investigated whether gender is associated with factor structure differences using the King et al. (1998)
exposed Bosnian secondary/high school boys and girls (N=1572) assessed nearly two years after the
1992–1995 Bosnian conflict. Confirmatory factor analytic tests of measurement invariance across PTSD
model parameters revealed many significant sex-linked differences. Implications regarding the potential
role of gender as a moderator of the King et al. (1998) model’s factor structure are discussed.
© 2011 Elsevier Ltd. All rights reserved.
Posttraumatic stress disorder (PTSD) was first introduced into
the Diagnostic and Statistical Manual (DSM) nomenclature in 1980
(American Psychological Association [APA], 1980). Since its initial
inclusion, the diagnosis of PTSD has been visited with controversy
latent factor structure. Namely, although PTSD is currently repre-
grouped into three symptom clusters, this three-factor model has
rarely received empirical support (cf. Shevlin, McBride, Armour, &
Adamson, 2009). Indeed, the factor analytic literature to date has
overwhelmingly supported a four-factor structure for PTSD (King,
Leskin, King, & Weathers, 1998; Simms et al., 2002). Recent empiri-
focused increasingly on evaluating whether the goodness of fit of
PTSD models is contingent on key moderating variables. In particu-
lar, gender is strongly related to the likelihood of a PTSD diagnosis
(Tolin & Foa, 2006). However, the influence of gender as a potential
∗Corresponding author at: National Centre of Psychotraumatology, University of
Southern Denmark, Odense, Denmark. Tel.: +44 7876142643.
E-mail addresses: firstname.lastname@example.org, email@example.com
moderator of PTSD’s factor structure has yet to be systematically
The underlying dimensions of PTSD are currently represented
in the DSM-IV by three factors: Intrusion (Criteria B1–B5), Effort-
ful Avoidance/Emotional Numbing (C1–C7), and Arousal (D1–D5).
Empirical support for this tripartite latent structure is rare (cf.
Asmundson, Stapleton, & Taylor, 2004). In contrast, the majority of
factor analytic studies have provided support for four-factor mod-
els. The two models receiving the most attention and support to
date include that proposed by King et al. (1998), and that proposed
by Simms et al. (2002). The major differences between the two
models lie in their comparative placement of items D1 (sleeping
difficulty), D2 (irritability or anger), and D3 (concentration difficul-
by allocating five items to an Intrusion factor (B1–B5), two items to
an Avoidance factor (C1–C2), four items to an Emotional Numbing
factor (C3–C7), and five items to an Arousal factor (D1–D5). This
model differs from the three-factor DSM-IV model in that it splits
the avoidance/emotional numbing factor into two separate fac-
tors based on evidence that emotional numbing and avoidance are
separate pathology-related constructs (Asmundson et al., 2004).
In contrast, the Simms et al. (2002) model groups the 17 PTSD
items by allocating five items to an Intrusion factor (B1–B5), two
items to an Avoidance factor (C1–C2), eight items to a Dyspho-
0887-6185/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
ria factor (C3–C7 and D1–D3), and two items to an Arousal factor
phoria factor items to other mood and anxiety disorders (Watson,
2005). Reflecting this similarity in content across disorders, recent
studies have found that the Dysphoria factor is less specific to
PTSD than the other three PTSD factors (Armour, McBride, Shevlin,
& Adamson, in press; Armour & Shevlin, 2010; Elklit, Armour, &
all PTSD items, Armour et al. (in press) and Elklit et al. (2010) both
est degree of attenuation occurring in the Dysphoria factor.
The current literature is ambiguous with respect to which of
the two models yields a more accurate representation of the latent
the King et al. (1998) model (most recently in Elhai, Engdahl, et al.,
2009; Lancaster, Melka, & Rodriguez, 2009; Naifeh, Elhai, Kashdan,
& Grubaugh, 2008; Palmieri, Marshall, & Schell, 2007; Schinka,
Brown, Borenstein, & Mortimer, 2007), as well as the Simms et al.
(2002) model (most recently in Armour & Shevlin, 2010; Boelen,
van den Hout, & van den Bout, 2008; Elhai, Ford, Ruggiero, & Frueh,
2009; Elklit & Shevlin, 2007; Palmieri, Weathers, Difede, & King,
2007). Accordingly, a general consensus as to which of these two
models provides the best representation of the underlying dimen-
sions of PTSD has yet to be reached.
As part of continuing efforts to clarify the strengths and poten-
tial drawbacks of the two models, recent studies have highlighted
a number of moderating variables that may be linked to the latent
structure of PTSD. Identifying potential moderators is of particular
value because it helps to clarify the circumstances under which a
given model fits best, and by extension, to identify specific popula-
tions, settings, or applications for which the model may be better
suited. For example, studies have found that PTSD model fit varies
(Palmieri, Weathers, et al., 2007), and whether respondents are
instructed to rate PTSD symptoms based on their worst trauma vs.
their global trauma history (Elhai, Engdahl, et al., 2009).
In addition to these efforts to evaluate the comparative fit of the
two PTSD factor models, a related line of studies have focused on
evaluating the stability (as gauged by the variance vs. invariance of
selected model parameters) of key model features as a function of
specific sample or setting characteristics. For example, the facto-
rial invariance of models of PTSD has been evaluated across groups
that differ in their native language (Marshall, 2004; Norris, Perilla,
Hourani, & Babeu, 2010; Simms et al., 2002), and era of military
service combined with treatment-seeking status (McDonald et al.,
2008). Of particular note, few studies have conducted compre-
hensive measurement invariance testing, but instead have tested
for model invariance in one or two focal model parameters (e.g.
McDonald et al., 2008).
To date, gender remains an understudied potential modera-
tor of the factor structure of PTSD. The relevance of gender to
the ongoing evaluation of competing models of PTSD is under-
scored by findings that although men tend to experience higher
rates of exposure to potentially traumatic events, women have
a two-fold higher risk of experiencing PTSD following exposure
to traumatic events (Breslau, Davis, Andreski, & Peterson, 1991;
Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995). This two-fold
risk for PTSD in trauma-exposed women was supported by a meta-
analysis of 52 studies (Tolin & Foa, 2006). Women are also more
likely to develop chronic forms of PTSD compared to men (Breslau
& Davis, 1992; Breslau et al., 1998).
Few compelling explanations for these gender differences in
PTSD prevalence rates have been offered. One explanation centers
and circumstances differ in their etiological potency for developing
PTSD (cf. Layne et al., 2010); (2) men and women have differen-
tial rates of exposure to different types of trauma; and therefore,
(3) the specific types of trauma to which women have compar-
atively higher rates of exposure are more likely to lead to PTSD.
However, multiple studies have documented higher rates of PTSD
in women even after controlling for trauma categories that occur
more frequently in women, including sexual assault and domestic
violence (Breslau, Chilcoat, Kessler, & Davis, 1999; Fullerton et al.,
2001; Stein, Walker, & Forde, 2000).
A second explanation for differential prevalence rates in PTSD
between the genders is based on findings that women are twice
more likely than men to report depression and anxiety symptoms
(Tolin & Breslau, 2007). Accordingly, higher prevalence rates of
PTSD in women may reflect a higher general prevalence of psycho-
logical distress or of psychiatric disorder. However, Breslau, Davis,
Andreski, Peterson, & Schultz (1997) reported that gender differ-
psychiatric disorders such as depression and anxiety.
lence rates centers on PTSD Criterion A, which, serves a key
“gatekeeper” function to a PTSD diagnosis. Specifically, Criterion
A is comprised of both objective (A1: traumatic experience) and
dence that women are more likely than men to disclose initial
adverse emotional reactions to trauma exposure by endorsing Cri-
experienced a Criterion A event.
In light of these gender-linked differences in variables linked to
the diagnosis or prevalence of PTSD, the current study examined
links between gender and the factorial invariance of an empiri-
cally supported model of PTSD. Of particular interest, both the King
et al. (1998) and Simms et al. (2002) models were originally devel-
oped using samples consisting of mostly or entirely adult men –
specifically military veterans. Notwithstanding this methodolog-
ical artifact, factor analytic support has been found in all-female
2005), as well as the Simms et al. (2002) model (Krause, Kaltman,
Goodman, & Dutton, 2007).
male military veterans to adult female civilian samples is a notable
advance, PTSD factor analytic research to date has nevertheless
been largely embedded within the adult literature. Few studies,
strong evidence that these younger age groups experience PTSD
symptoms in response to varying trauma types (Goenjian et al.,
1995; McLeer, Deblinger, Henry, & Orvaschel, 1992; Sack, Clarke,
& Seeley, 1995). Valuable exceptions to this general trend include
Saul, Grant, and Carter’s (2008) CFA study, using a sub-sample of
1581 adolescents from the National Survey of Adolescents (NSA),
and Elhai, Ford, et al.’s (2009) study of 4023 adolescents, the full
NSA sample, both of which found support for the King et al. model.
Saul et al. (2008) found that magnitudes of factor loadings of the
King et al. model significantly differed between boys and girls. This
evidence of factorial variance across gender groups points to the
ables may play as moderators of the latent structure of PTSD across
diverse groups. Of particular interest, measurement invariance and
comparative fit of both the King et al. and Simms et al. models of
PTSD have not been comprehensively tested across gender groups
within the same study sample in either the adult or adolescent
Given lack of clear consensus in the factor analytic literature as
PTSD, we did not formulate specific hypotheses concerning which
of the two models would best fit the data. However, we planned
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
a priori to use the model that provided the best model fit in all
subsequent invariance-testing analyses. Specifically, based on evi-
dence of a two-fold increase in PTSD for women (Breslau et al.,
chronic forms of PTSD (Breslau & Davis, 1992; Breslau et al., 1998),
we formed the general hypothesis that girls would report higher
mean levels of symptomatic distress compared to boys (as gauged
parameters (including factor loadings, covariances, variances, and
means); however, given a lack of prior evidence or well-supported
theory relating to gender-linked differences in PTSD model param-
eters in adolescents, we did not formulate any specific hypotheses
regarding the values of these parameters.
secondary schools who completed a screening survey during the
1997–1998 school year. All schools were located in regions in Cen-
tral Bosnia that were heavily war-exposed during the 1992–1995
Bosnian conflict. The purpose of the survey was to identify appro-
priate candidates for Trauma and Grief Component Therapy, a
UNICEF-sponsored school-based intervention (Layne et al., 2001,
2008; see also Cox, Moata, Clara, & Asmundson, 2008; Kutlac et al.,
2002). Demographic characteristics of the study’s effective sample
are described below.
In the Fall of 2000, in accordance with a study protocol designed
to reduce potential stigmatization of individual “high risk” stu-
dents, 16 trained school counselors used purposive sampling to
of severely war-exposed students. The counselors were psychol-
ogists and pedagogues (the duties of pedagogues most closely
resemble those of guidance counselors and learning tutors in U.S.
high schools). The counselors used all professional tools available
to them to select classrooms, including school records (which indi-
cated whether a father or mother had been killed in the war, and
whether the family was internally displaced due to the war); dis-
cussions with other teachers, parents, and school administrators;
and information gathered in their roles as teachers and counselors
at the school (cf. Layne et al., 2008, for further details concerning
classroom preparation, obtaining informed consent of caregivers
and informed consent of students, and survey administration pro-
cedures). Approximately 100 students per school completed the
risk screening survey. Students who reported significant current
distress were then selected for further screening and considera-
tion for specialized mental health services (cf. Layne et al., 2001,
2008; for further details). Screening survey measures and proce-
dures were approved both by an ad hoc institutional review board
(IRB) formed of Bosnian mental health professionals and a local
UNICEF officer, and by the IRB of Brigham Young University.
The screening survey included measures of pre-war, war-time,
and post-war trauma and loss exposure, post-war adversities, fre-
quency of exposure to current trauma and loss reminders, and
appropriateness and were forward- and back-translated (between
English and the local Serbian/Bosniak Muslim/Croatian native lan-
guages) by local doctoral-level psychology students and health
professionals as described elsewhere (Layne et al., 2008). Instru-
ments relevant to the present paper are described below.
2.3.1. War Trauma Screening Inventory (WTSI)
The WTSI (Layne, Stuvland, Saltzman, Djapo, & Pynoos, 1999) is
a measure of exposure to a broad range of war-related traumatic
events. The WTSI version used contained 35 “yes”/“no” self-report
civil war in the Former Yugoslavia that were generated based on
field research in post-war Bosnia and on a review of war expo-
sure instruments (e.g., Macksoud & Aber, 1996). Dimensions of war
exposure assessed by the WTSI included (1) direct physical injury,
(2) witnessing violence, (3) life threat, (4) traumatic bereavement,
(5) harm to loved ones, (6) threat to loved ones, (7) displace-
ment, and (8) separations from loved ones caused by the war (see
Layne et al., 2010). All 35 items constituted Criterion A1-qualifying
events. After completing the WTSI, students were asked to record a
brief handwritten description of their “most upsetting war-related
fied,” “horrified,” or “helpless” during that event (in three separate
“yes”/“no” items, thereby assessing PTSD Criterion A2). Respon-
dents were subsequently instructed to rate their current PTSD
symptoms relating to this most upsetting traumatic experience
(a common PTSD assessment strategy, reviewed in Elhai, Engdahl,
et al., 2009) using the UCLA PTSD Reaction Index-Revised.
2.3.2. UCLA PTSD Reaction Index-Revised
The Reaction Index (Steinberg, Brymer, Decker, & Pynoos, 2004)
is a self-report scale of PTSD symptom frequency experienced dur-
ing the previous month. The Reaction Index version used in this
study contained 17 items corresponding to the 17 DSM-IV PTSD
diagnostic criteria. Items were measured using a five-point fre-
quency scale ranging from 0 (Never) to 4 (Almost Always). The
˛=.87), two-week test–retest reliability (r=.75), and criterion-
referenced validity in relation to a range of distress measures (r’s
ranging from .30 to .70) in post-war Bosnian adolescents sampled
in 2001 (Layne et al., 2009).
3.1. Preparatory analyses
Of 1572 students screened, 43 students endorsed no trauma
did not identify their gender (leaving 1480 students as the effec-
tive sample size). All but four students in the effective sample were
missing fewer than three Reaction Index items. Missing items were
found to be missing completely at random (MCAR) as measured
by Little’s MCAR ?2(458, N=1480)=479.92, p>.05. We conducted
multiple imputation with an iterative Markov chain Monte Carlo
method using the Gibbs Sampler procedure (using SPSS’s Version
17 Missing Value Analysis software) to estimate missing item-level
PTSD data as generated across five imputed datasets. In addition
to the individual PTSD items, gender was also used as a predictor
variable in the imputation to account for possible between-group
missing data variation. CFA parameter estimates were averaged
across the five datasets (see Schafer & Graham, 2002) using Mplus
5.2 software (Muthén & Muthén, 2007).
(2002) four-factor PTSD models using the Reaction Index items.
(The last loading on each of the King model factors were fixed to
1.0, and the same items in the Simms factors were also fixed to
1.0, for scale identification purposes.) All items were normally dis-
tributed, with the highest univariate skewness level being 1.55 and
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
the highest kurtosis level being 1.86. The CFA’s thus used standard
maximum likelihood estimation. Goodness of fit indices used to
evaluate model fit included the comparative fit index (CFI; Bentler,
1990), Tucker–Lewis Index (TLI; Tucker & Lewis, 1973), root mean
square error of approximation (RMSEA; Steiger, 1990), and stan-
dardized root mean square residual (SRMR; Jöreskog & Sörbom,
1993). Good-fitting models are generally considered to fall within
.10 for adequate-fitting models) (Hu & Bentler, 1999). In addition,
within-group comparisons of the King et al. (1998) vs. Simms et al.
(2002) models that involved non-nested model comparisons (i.e.,
comparing a model that is not a subset of another model) used
the Bayesian Information Criterion (BIC). A 10-point BIC difference
between models denotes a 150:1 likelihood that the model with
the lower BIC value has better fit (p<.05) (Raftery, 1995).
to zero, to test measurement invariance using parameter equality
constraints across the boys (n=471) and girls (n=1009). We tested
for invariance/non-invariance across groups on factor loadings,
observed variable intercepts, observed variable residual variances,
factor variances and covariances, and factor means, following
established procedures (e.g., Gregorich, 2006; Meredith, 1993;
Meredith & Teresi, 2006).
Model A allowed groups (boys and girls) to vary on all param-
eters (i.e., testing configural invariance). Subsequent models tested
progressively more conservative restrictions, constraining particu-
lar parameter estimates to be equal across groups, as tested against
the prior step’s model (except as otherwise noted). Model B con-
strained factor loadings as equal across groups (thereby testing
metric or pattern invariance). Metric invariance is established if
the corresponding factor loadings are equivalent across groups
(Gregorich, 2006). Model C additionally constrained observed vari-
able intercepts to be equal (thereby testing strong or scalar factorial
invariance). Model D additionally constrained residual variances
to be equal (thereby testing strict factorial invariance). Additional
analyses further tested structural invariance but did not constrain
residual variances, including Model E, which additionally con-
strained factor variances and covariances to be equivalent across
gender groups (tested against Model C); and Model F, which addi-
tionally constrained factor means to be equivalent across gender
groups (tested against Model E).
Tests of statistical significance between Models A through F
were assessed with chi-square difference tests, comparing a given
CFA model assuming equal parameter estimates across groups
(e.g., factor loadings) against a model allowing those estimates to
vary across gender groups. Comparing model fit statistics between
invariance testing models is not recommended given that it leads
to inaccurate results (Fan & Sivo, 2009).
(n=1009) girls and 31.8% (n=471) boys. Age ranged from 14 to 21
years (M=16.18, SD=1.07). Most participants were in the second
year of high school (n=549, 37.4%), followed by third year (n=425,
29.0%), first year (n=332, 22.6%), and fourth year (n=160, 10.9%)
(n=14 not reported). The mean summed Reaction Index score of
25.19 for girls (SD=13.81; n=1009) was significantly higher than
that of boys (M=16.07; SD=11.50; n=471), F(1, 1478)=154.97,
p<.001 (Cohen’s d=.66).
Analysis of BIC indices indicated that the King et al. (1998)
model provided equally good fit in both the girls and boys sub-
groups; the same was true of the Simms et al. (2002) model.
Notably, BIC differences can only be used to compare models
from within a single sample (e.g., girls) and so cannot be used
to compare models across the two gender groups. For girls, the
King et al. model fit well, ?2(113, N=1009)=580.18, p<.001;
CFI=.93, TLI=.92, RMSEA=.06, SRMR=.04, BIC=49,704.059, as
Unstandardized observed item parameter estimates for the boy and girl groups.
Items BoysGirlsBoys GirlsBoys Girls
Emotional cue reactivity
Physiological cue reactivity
Avoidance of thoughts
Avoidance of reminders
Loss of interest
did the Simms et al. model, ?2(113, N=1009)=570.42, p<.001;
CFI=.94, TLI=.92, RMSEA=.06, SRMR=.04, BIC=49,694.26; the BIC
difference was less than 10 units. The King et al. model fit ade-
quately for the boys, ?2(113, N=471)=334.12, p<.001; CFI=.92,
TLI=.90, RMSEA=.06, SRMR=.05, BIC=22,126.41, and was nearly
adequate for the Simms et al. model, ?2(113, N=471)=343.18,
(the BIC’s again differed by fewer than 10 units). Thus, neither the
tested separately in both the girls subgroup and the boys subgroup.
Given the lack of superiority in fit for either model, we conducted
our subsequent analyses using the King et al. (1998) model for two
reasons. First, the King et al. model is an older and more estab-
lished model, and second, the model more closely resembles the
latent PTSD structure proposed for the DSM-V (APA, 2010).
Measurement invariance analyses formally tested differences
between girls and boys on the structural parameters of the King
et al. (1998) four-factor model (see Tables 1 and 2). Model A, which
allowed for variation in parameter estimates across groups, fit
the data adequately, ?2(226, N=1480)=913.54, p<.001; CFI=.93,
TLI=.92, RMSEA=.06, SRMR=.04. Based on difference tests (see
Table 3), analyses revealed that all parameter estimates differed
significantly (all p’s<.05) between subgroups. Specifically, the girls
subgroup had higher observed intercepts, indicating greater PTSD
item severity than the boys. The girls also had larger residual
error variances, demonstrating greater residual error that is not
Unstandardized factor parameter estimates for the A2 and non-A2 groups.
R with A
R with N
R with H
A with N
A with H
N with H
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
Comparisons of boys and girls using measurement invariance procedures.
Chi-square difference test
Difference testsModels tested
A vs. B
B vs. C
C vs. D
C vs. E
E vs. F
Note: Table presents the difference between chi-square values for the model
comparisons; degrees of freedom are in parentheses. Model A=no parameters con-
strained equal across groups; Model B=factor loadings constrained to be equal;
Model C=observed variable intercepts and factor loadings constrained to be equal;
Model D=residual variances, factor loadings and observed variable intercepts con-
strained to be equal; Model E=factor variances and covariances, factor loadings
and observed variable intercepts constrained to be equal; Model F=factor means,
factor variances and covariances, factor loadings and observed variable intercepts
constrained to be equal.
accounted for by PTSD’s factors. The girls also evidenced larger fac-
higher co-variances indicating larger inter-correlations for PTSD’s
Although the great majority of factor analytic studies suggests
that the latent structure of PTSD is comprised of four factors,
no clear consensus has yet emerged regarding which four-factor
model (King et al., 1998; Simms et al., 2002) best represents the
latent structure of PTSD. The current study fit both the King et al.
(1998) and Simms et al. (2002) four-factor models to data from a
by gender. Results indicated that both the King et al. and Simms
et al. models provided equivalently good fit for girls compared
to boys. Invariance testing revealed that the girl subgroup sig-
nificantly differed from the boy subgroup in all factor structure
The initial CFA produced nearly identical model fit for the girls
vs. boys subgroups in both the King et al. (1998) and Simms et al.
(2002) models. Given this finding, and the rationale that the King
limited adolescent factor analytic literature (Elhai, Gray, Docherty,
Kashdan, & Kose, 2007; Saul et al., 2008), and is more similar to
proposed PTSD criteria for the DSM-V (APA, 2010), we tested the
groups. Given prior evidence that women have a two-fold greater
risk for PTSD following trauma exposure (Breslau et al., 1991;
compared to men (Breslau & Davis, 1992; Breslau et al., 1998), we
of distress compared to boys as gauged by the item-level indicator
intercepts. This hypothesis was confirmed, in that all 17 indicator
respective counterparts in the boys subgroup. More generally, our
exploratory tests of factorial invariance found systematic differ-
ences between boys and girls subgroups on all model parameters.
group manifested higher residual error variances, factor variances,
loadings across groups, however, yielded mixed results.
We also found higher inter-factor covariances in the girls sub-
group, pointing to a greater degree of collinearity between the
factors in girls. This finding suggests that the latent structure of
PTSD symptom factors may be more oblique (and thus potentially
more redundant across factors) in adolescent girls compared to
4.1. Potential contributors to gender-linked differences in PTSD
The finding that the girls’ observed indicator residual error vari-
ances were significantly higher than those in the boys subgroup
suggests that the girls’ PTSD symptom factors did not capture as
finding suggests a greater degree of measurement error in girls’
observed responses compared to boys – a statistical artifact that
may be traceable back to an early methodological artifact: early
research into the latent structure of PTSD was largely conducted
using adult male samples (King et al., 1998; Simms et al., 2002).
This finding raises the question of whether gender discrepancies
(at least in adolescence) exist in the current diagnostic formula-
tion of PTSD, such that its factor structure maps more closely onto
its clinical manifestations as they appear in males compared to
Although this study found evidence that gender functions as
a moderator of structural parameters of latent variable models of
PTSD in adolescents, the existing child and adolescent war trauma
literature provides mixed (and still ambiguous) evidence that gen-
der performs a related function – moderating the predictive (or
causal) links between war exposure and post-war distress (Fayyad,
Karam, Karam, Tabet, & Ghosn, 2004; Powell & Durakovi´ c-Belko,
2002; Shaw, 2003). For example, in a rare follow-up study con-
(2003) found that demographic variables including gender and age
traumatic stress reactions (p. 21). In contrast, in a cross-sectional
sample of 393 war-exposed Bosnian secondary school students,
Durakovicˇ ı-Belko, Kulenovicˇ ı, and Dapicˇ ı (2003) found that female
gender was a risk factor for both PTSD and depressive symptoms
ticular interest, the authors found evidence of differential relations
as a function of predicted outcome, in that war-related loss was
the most potent predictor of PTSD symptoms, whereas female gen-
der was the most potent predictor of depressive symptoms (cf.
Layne et al., 2010). Notably, much of the extant gender-related
predictor or grouping variable in relation to a distress-related out-
come, rather than examined in a formal test of interaction as tests
of “true” moderation require (cf. Tolin & Foa, 2006).
Considered more generally, reasons why females tend to report
higher levels of distress compared to males following exposure to
traumatic events are not well understood. Although several expla-
nations have been put forward, most explanations pointing to a
single source or cause (e.g., exposure to more traumatic events)
have been rejected after further study (e.g., epidemiological sur-
veys found that males report higher levels of trauma exposure;
Breslau et al., 1991; Kessler et al., 1995; Tolin & Foa, 2006). Instead,
it is more likely that gender-linked differences in PTSD are linked
to a complex array of factors, including differences in exposure
to and appraisals of different types of stress, biological develop-
ment, secondary adversities set in motion by prior trauma and loss,
and the differential use of adaptive vs. maladaptive coping strate-
gies (Compas, Connor-Smith, Saltzman, Thomsen, & Wadsworth,
2001). One possible explanation, for example, is that the higher
levels of distress (at the item level) reported by the girls in the
current study may be related to higher levels of emotionality and
the use of different coping strategies compared to their male coun-
terparts (cf. Compas, Orosan, & Grant, 1993). For example, studies
suggest that women favor the use of ‘emotion’ focused coping as
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
opposed to men who favor the use of ‘problem’ focused coping
(Gavranidou & Rosner, 2003). Clohessy and Ehlers (1999) reported
that, compared to men, women are more likely to rely on wish-
ful thinking, memory suppression, and disengagement (avoidant
emotion coping strategies), all of which have been shown to corre-
late positively with severe PTSD symptoms. In addition, Coyne and
Racioppo (2000) reported that emotion-focused coping (especially
avoidant emotional coping), when compared to active emotional
coping, is related to poorer mental health outcomes. Tolin and
Foa (2006) also suggest that social expectancy (e.g., women are
expected to be more vulnerable to trauma and its consequences,
whereas men are expected to be more resilient and “tough”) may
also partly account for gender differences. Further study of spe-
other trauma and loss-related events is warranted, given that both
problem-focused and avoidant emotion-focused coping are posi-
tively associated with PTSD symptom severity (Schnider, Elhai, &
4.2. Beyond gender: other potential moderators of PTSD
Efforts to evaluate the implications of these findings for the
factor analytic literature and for future research should consider
other potential moderators of the latent structure of PTSD. For
example, Palmieri, Weathers, et al. (2007) found that the type
or appearance of the measurement instrument employed (e.g.,
administration format) is linked to the factor structure of PTSD.
Specifically, the authors found that the Simms et al. (2002) model
list (PCL; Weathers, Litz, Huska, & Keane, 1994), whereas the King
et al. (1998) model provided better fit for data collected via the
Clinician-Administered PTSD Scale (CAPS; Blake et al., 1990). How-
ever, between-model differences in fit statistics were modest in
size. With respect to gender-linked differences in PTSD, Tolin and
Foa (2006) reported that between-gender differences appear to
be more pronounced when studies employ self-report question-
clinical interviews. Of particular interest, Tolin and Foa observed
that some interview-based studies have failed to find significant
ods have generally found higher female PTSD rates. This potential
interaction between instrument format (self-report vs. interview)
and gender raises the possibility that the specific type of mea-
surement instrument employed may be linked to both systematic
differences in prevalence rates of PTSD, as well as in the latent
Essentially, when using self-report measures, girls may be over-
reporting whereas boys may be underreporting. If this is the case,
higher rates of false positives in girls, and false negatives in boys.
The clinical implications are that a disproportionally higher rate of
girls may be selected for specialized treatments.
Other format-related features of assessment instruments may
also moderate the latent structure of PTSD symptoms. Elhai,
Palmieri, Biehn, Frueh, and Magruder (2010) found systematic dif-
ferences in the latent structure of PTSD symptoms linked to the
specific self-report CAPS (Blake et al., 1990) and format used (e.g.,
rating the frequency vs. the intensity of PTSD symptoms). Further,
the limited research to date relating to the factorial invariance
of PTSD models between ethnic groups (Hoyt & Yeater, 2010;
Marshall, 2004; Norris et al., 2001) points to the need to consider
the role of culture given the ethnically diverse region (post-war
Bosnia) in which the data were collected.
Study strengths and limitations. This study is, to our knowledge,
the first to systematically test gender as a moderator of all relevant
structural parameters of models of PTSD symptoms. We employed
a robust latent variable modeling approach and drew from a large
sample of war-exposed adolescents that is unique in the PTSD
report measure rather than clinical interview. However, although
clinical interviews are generally regarded as the “gold standard”
in the clinical assessment of PTSD, diagnoses based on self-report
measures vs. clinical interviews often reach an acceptable level
of agreement (Coffey, Dansky, Falsetti, Saladin, & Brady, 1998;
Harrington & Newman, 2007). Further, the latent structure of the
King et al. (1998) model of PTSD has been supported as assessed
via both clinical interviews and self-report measures. Indeed, fac-
tor analytic research has often been conducted on data collected
via self-report measures (cf. Armour & Shevlin, 2010).
In conclusion, neither the King nor the Simms model was statis-
tically superior to the other in both subgroups indicating that they
provided equivalently good fit. Invariance testing revealed that the
tor structure parameters of the King et al. (1998) model. A greater
degree of measurement error in girls’ observed responses com-
pared to boys indicates that the factor structure maps more closely
onto its clinical manifestations as they appear in males compared
to females possibly attributable to the fact that, the construct was
originally ‘coined’ during the Vietnam War Era with young men,
and the fact that early research on the latent structure of PTSD
2002). The higher mean levels of distress reported by females may
be attributable to higher levels of emotionality and the use of dif-
ferent coping strategies compared to their male counterparts (cf.
Compas, Orosan, & Grant, 1993). In addition, the potential inter-
action between instrument format (self-report vs. interview) and
gender raises the possibility that the specific type of measurement
in prevalence rates of PTSD, as well as in the latent structure of
implications being that females may be disproportionally selected
for specialized treatments.
American Psychiatric Association. (1980). Diagnostic and statistical manual of mental
disorders (3rd ed.). Washington, DC: American Psychiatric Association.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental
disorders (4th ed.). Washington, DC: American Psychiatric Association.
American Psychiatric Association. DSM-5 Development (2010). Proposed draft
revisions to DSM disorders and criteria. Retrieved March 16, 2010, from
Armour, C., McBride, O., Shevlin, M., & Adamson, G. Testing the robustness of the
Theory, Research, Practice and Policy, in press.
Armour, C., & Shevlin, M. (2010). Testing the dimensionality of PTSD and the speci-
ficity of the dysphoria factor. Journal of Loss and Trauma, 15, 11–27.
Asmundson, G., Stapleton, J. A., & Taylor, S. (2004). Are avoidance and numbing
distinct PTSD symptom clusters? Journal of Traumatic Stress, 7, 467–475.
Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological
Bulletin, 107, 238–246.
Blake, D., Weathers, F., Nagy, L., Kaloupek, D., Klauminzer, G., Charney, D., et al.
1. Behavior Therapist, 13, 187–188.
Boelen, P. A., van den Hout, M. A., & van den Bout, J. (2008). The factor structure of
posttraumatic stress disorder symptoms among bereaved individuals: a confir-
matory factor analysis. Journal of Anxiety Disorders, 22, 1377–1383.
Breslau, N., Kessler, R. C., Chilcoat, H. D., Schultz, L. R., Davis, G. C., & Andreski, P.
(1998). Posttraumatic stress disorder in the community: the 1996 Detroit area
survey of trauma. Archives of General Psychiatry, 55, 626–632.
Breslau, N., & Davis, G. (1992). Posttraumatic stress disorder in an urban population
of young adults: risk factors for chronicity. American Journal of Psychiatry, 149,
Breslau, N., Davis, G. C., Andreski, P., & Peterson, E. (1991). Traumatic events and
PTSD in an urban population of young adults. Archives of General Psychiatry, 48,
Breslau, N., Chilcoat, H. D., Kessler, R. C., & Davis, G. C. (1999). Previous exposure to
trauma and PTSD effects of subsequent trauma: results from the Detroit area
survey of trauma. American Journal of Psychiatry, 156, 902–907.
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611
Breslau, N., Davis, G. C., Andreski, P., Peterson, E. L., & Schultz, L. R. (1997). Sex
differences in posttraumatic stress disorder. Archives of General Psychiatry, 54,
Brewin, C. R., Andrews, B., & Rose, S. (2000). Fear, helplessness and horror in post-
crime. Journal of Traumatic Stress, 13, 400–509.
Clohessy, S., & Ehlers, A. (1999). PTSD symptoms, responses to intrusive memories
and coping in ambulance service workers. British Journal of Clinical Psychology,
for PTSD in a substance abuse sample: psychometric properties of a modified
version of the PTSD symptom scale self-report. Journal of Traumatic Stress, 11,
Compas, B. E., Connor-Smith, J. K., Saltzman, H., Thomsen, A. H., & Wadsworth,
M. E. (2001). Coping with stress during childhood and adolescence: prob-
lems, progress, and potential in theory and research. Psychological Bulletin, 127,
Compas, B. E., Orosan, P. G., & Grant, K. E. (1993). Adolescents stress and coping:
Implications for psychopathology during adolescents. Journal of Adolescence,
Cox, B. J., Moata, N., Clara, I., & Asmundson, G. J. G. (2008). The symptom structure of
posttraumatic stress disorder in the National Comorbidity Replication Survey.
Journal of Anxiety Disorders, 22, 1523–1528.
Coyne, J. C., & Racioppo, M. W. (2000). Never the twain shall meet? Closing the gap
between coping research and clinical intervention research. American Psycholo-
gist, 55, 655–664.
Durakovicˇ ı-Belko, E., Kulenovicˇ ı, A., & Ðapicˇ ı, R. (2003). Determinants of Posttrau-
matic Adjustment in Adolescents from Sarajevo Who Experienced War. Journal
of Clinical Psychology, 59, 27–40.
Elhai, J. D., Engdahl, R. M., Palmieri, P. A., Naifeh, J. A., Schweinle, A., & Jacobs, G.
A. (2009). Assessing posttraumatic stress disorder with or without reference
to a single, worst traumatic event: examining differences in factor structure.
Psychological Assessment, 21, 629–634.
Elhai, J. D., Ford, J. D., Ruggiero, K. J., & Frueh, B. C. (2009). Diagnostic alterations for
posttraumatic stress disorder: examining data from the national comorbidity
survey replication and national survey of adolescents. Psychological Medicine,
Elhai, J. D., Gray, M. J., Docherty, A. R., Kashdan, T. B., & Kose, S. (2007).
Structural validity of the posttraumatic stress disorder checklist among col-
lege students with a trauma history. Journal of Interpersonal Violence, 22,
Elhai, J. D., Palmieri, P. A., Biehn, T. L., Frueh, B. C., & Magruder, K. M. (2010). Post-
traumatic stress disorder’s frequency and intensity ratings are associated with
factor structure differences in military veterans. Psychological Assessment, 22,
Elklit, A., Armour, C., & Shevlin, M. (2010). Testing alternative factor models of
PTSD and the robustness of the dysphoria factor. Journal of Anxiety Disorders,
Elklit, A., & Shevlin, M. (2007). The structure of PTSD symptoms: a test of alternative
models using confirmatory factor analysis. British Journal of Clinical Psychology,
Fan, X., & Sivo, S. A. (2009). Using ? goodness-of-fit indexes in assessing mean
structure invariance. Structural Equation Modeling, 16, 54–67.
and adolescents following war. In: R. R. Silva (Ed.), Posttraumatic stress disorders
in children and adolescents. New York: W.W. Norton.
Fullerton, C. S., Ursano, R. J., Epstein, R. S., Crowley, B., Vance, K., Kao, T., et al.
(2001). Gender differences in posttraumatic stress disorder after motor vehicle
accidents. American Journal of Psychiatry, 158, 1486–1491.
Gavranidou, M., & Rosner, R. (2003). The weaker sex? Gender and posttraumatic
stress disorder. Depression and Anxiety, 17, 130–139.
Gregorich, S. E. (2006). Do self-report instruments allow meaningful compar-
isons across diverse population groups?: testing measurement invariance using
the confirmatory factor analysis framework. Medical Care, 44(11, Suppl. 3),
I., et al. (1995). Psychiatric co morbidity in children after the 1988 earthquake
in Armenia. Journal of the American Academy of Child and Adolescent Psychiatry,
Harrington, T., & Newman, E. (2007). The psychometric utility of two self-report
measures of PTSD among women substance users. Addictive Behaviours, 32,
structure models in Hispanic and white college students. Psychological Trauma;
Theory, Research, Practice, and Policy, 2, 19–30.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance struc-
ture analysis: conventional criteria versus new alternatives. Structural Equation
Modeling, 6, 1–55.
Jöreskog, K., & Sörbom, D. (1993). LISREL 8 user’s reference guide. Chicago: Scientific
Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. B. (1995).
Posttraumatic-stress-disorder in the national comorbidity survey. Archives of
General Psychiatry, 52, 1048–1060.
King, D. W., Leskin, G. A., King, L. A., & Weathers, F. W. (1998). Confirmatory factor
analysis of the clinician-administered PTSD scale: evidence for the dimension-
ality of posttraumatic stress disorder. Psychological Assessment, 10, 90–96.
Krause, E. D., Kaltman, S., Goodman, L. A., & Dutton, M. A. (2007). Longitudinal factor
Psychological Assessment, 19, 165–175.
Kuterovac-Jagodicˇ ı, G. (2003). Posttraumatic Stress Symptoms in Croatian Chil-
dren Exposed to War: A Prospective Study. Journal of Clinical Psychology, 59,
Kutlac, M., Layne, C. M., Wood, J., Saltzman, W. S., Stuvland, R., & Pynoos, R. S.
(2002). Psychological adjustment in war-exposed secondary school students
two years after the war: results of a large-scale risk screening survey. In: S. Pow-
ell, & E. Durakovic-Belko (Eds.), Sarajevo 2000: the psychosocial consequences of
war. Results of empirical research from the territory of former Yugoslavia. Sarajevo,
Lancaster, S. L., Melka, S. E., & Rodriguez, B. F. (2009). A factor analytic com-
parison of five models of PTSD symptoms. Journal of Anxiety Disorders, 23,
Layne, C. M., Olsen, J. A., Baker, A., Legerski, J.-P., Isakson, B., Paˇ sali´ c, A., et al. (2010).
Unpacking trauma exposure risk factors and differential pathways of influence:
Predicting postwar mental distress in Bosnian adolescents. Child Development,
Layne, C. M., Pynoos, R. S., Saltzman, W. R., Arslanagic, B., Black, M., Savjak, N., et al.
(2001). Trauma/grief-focused group psychotherapy: school-based post-war
intervention with traumatized Bosnian adolescents. Group Dynamics: Theory,
Research, and Practice, 5, 277–290.
Layne, C. M., Saltzman, W. R., Poppleton, L., Burlingame, G. M., Pasalic, A., Durakovic,
E., et al. (2008). Effectiveness of a school-based group psychotherapy program
for war-exposed adolescents: a randomized controlled trial. Journal of the Amer-
ican Academy of Child and Adolescent Psychiatry, 47, 1048–1062.
Layne, C. M., Stuvland, R., Saltzman, W. R., Djapo, N., & Pynoos, R. S. (1999). War
Layne, C. M., Warren, J. S., Hilton, S., Lin, D., Pasalic, A., Fulton, J., et al. (2009). Mea-
suring adolescent perceived support amidst war and disaster: the multi-sector
social support inventory. In: B. K. Baker (Ed.), Adolescents and war: How youth
deal with political violence (pp. 145–176). New York: Oxford.
Macksoud, M. S., & Aber, J. L. (1996). The war experiences and psychosocial devel-
opment of children in Lebanon. Child Development, 67, 70–88.
ance of posttraumatic stress disorder symptoms among U.S. military personnel.
Journal of Traumatic Stress, 23, 91–99.
Marshall, G. N. (2004). posttraumatic stress disorder symptom checklist: factor
structure and English–Spanish measurement invariance. Journal of Traumatic
Stress, 17, 223–230.
McDonald, S. D., Beckham, J. C., Morey, R., Marx, C., Tupler, L. A., & Calhoun, P. S.
(2008). Factorial invariance of posttraumatic stress disorder symptoms across
three veteran samples. Journal of Traumatic Stress, 21, 309–317.
McLeer, S. V., Deblinger, E., Henry, D., & Orvaschel, H. (1992). Sexually abused chil-
dren at high risk for post-traumatic stress disorder. Journal of the American
Academy of Child and Adolescent Psychiatry, 31, 875–879.
Meredith, W. (1993). Measurement invariance, factor analysis and factorial invari-
ance. Psychometrika, 58, 525–543.
Meredith, W., & Teresi, J. A. (2006). An essay on measurement and factorial invari-
ance. Medical Care, 44(11, Suppl. 3), S69–S77.
Muthén, B. O., & Muthén, L. K. (2007). Mplus (Version 5). Los Angeles, California.
Naifeh, J. A., Elhai, J. D., Kashdan, T. B., & Grubaugh, A. L. (2008). The PTSD symptom
scale’s latent structure: an examination of trauma-exposed medical patients.
Journal of Anxiety Disorders, 22, 1355–1368.
Norris, F., Perilla, J., & Murphy, A. (2001). Postdisaster stress in the United
States and Mexico: a cross-cultural test of the multicriterion conceptual
model of posttraumatic stress disorder. Journal of Abnormal Psychology, 110,
stress symptoms in sexually harassed women. Journal of Traumatic Stress, 18,
Palmieri, P. A., Marshall, G. N., & Schell, T. L. (2007). Confirmatory factor analysis
of posttraumatic stress symptoms in Cambodian refugees. Journal of Traumatic
Stress, 20, 207–216.
Palmieri, P. A., Weathers, F. W., Difede, J., & King, D. W. (2007). Confirmatory factor
ter workers exposed to the world trade centre ground zero. Journal of Abnormal
Psychology, 116, 329–341.
Powell, S., & Durakovi´ c-Belko, E. (Eds.) (2002). Sarajevo 2000: the psychosocial con-
sequences of war. Results of empirical research from the territory of former
Yugoslavia. Internet: http://www.psih.org.
ology, 25, 111–163.
Saul, A. L., Grant, K. E., & Carter, J. S. (2008). Post-traumatic reactions in adolescents:
how well do the DSM-IV PTSD criteria fit the real life experience of trauma
exposed youth? Journal of Abnormal Child Psychology, 36, 915–925.
Sack, W. H., Clarke, G. N., & Seeley, J. (1995). Posttraumatic stress disorder across
two generations of Cambodian refuges. Journal of the American Academy of Child
and Adolescent Psychiatry, 24, 1160–1166.
Schafer, J. L., & Graham, J. W. (2002). Missing data: our view of the state of the art.
Psychological Methods, 7, 147–177.
Schinka, J. A., Brown, L. M., Borenstein, A. R., & Mortimer, J. A. (2007). Confirmatory
factor analysis of the PTSD checklist in the elderly. Journal of Traumatic Stress,
C. Armour et al. / Journal of Anxiety Disorders 25 (2011) 604–611 Download full-text
Schnider, K. R., Elhai, J. D., & Gray, M. J. (2007). Coping style use predicts
posttraumatic stress and complicated grief symptom severity among col-
lege students reporting a traumatic loss. Journal of Counseling Psychology, 54,
Shaw, J. A. (2003). Children exposed to war/terrorism. Clinical Child and Family Psy-
chology Review, 6, 237–246.
Shevlin, M., McBride, O., Armour, C., & Adamson, G. (2009). Reconciling the differ-
ences between the King et al. (1998) and Simms et al. (2002) factor models of
PTSD. Journal of Anxiety Disorders, 23, 995–1001.
Simms, L. J., Watson, D., & Doebbeling, B. N. (2002). Confirmatory fac-
tor analyses of posttraumatic stress symptoms in deployed and non-
deployed veterans of the Gulf War. Journal of Abnormal Psychology, 111,
Spitzer, R. L., First, M. B., & Wakefield, J. C. (2007). Saving PTSD from itself in DSM-V.
Journal of Anxiety Disorders, 21, 233–241.
Steiger, J. H. (1990). Structural model evaluation and modification: an interval esti-
mation approach. Multivariate Behavioral Research, 25, 173–180.
Stein, M. B., Walker, J. R., & Forde, D. R. (2000). Gender differences in suscep-
tibility to posttraumatic stress disorder. Behaviour Research and Therapy, 38,
Steinberg, A. M., Brymer, M., Decker, K., & Pynoos, R. S. (2004). The UCLA PTSD
reaction index. Current Psychiatry Reports, 6, 96–100.
Tolin, D. F., & Breslau, N. (2007). Sex differences in risk of PTSD. PTSD Research
Quarterly, 18. Spring
Tolin, D. F., & Foa, E. B. (2006). Sex differences in trauma and posttraumatic stress
disorder: a quantitative review of 25 years of research. Psychological Bulletin,
Tucker, C., & Lewis, C. (1973). A reliability coefficient for maximum likelihood factor
analysis. Psychometrika, 38, 1–10.
Watson, D. (2005). Rethinking the mood and anxiety disorders: a quantita-
tive hierarchical model for DSM-V. Journal of Abnormal Psychology, 114,
Weathers, F. W., Litz, B. T., Huska, J. A., & Keane, T. M. (1994). The PTSD checklist-
civilian version (PCL-C). Boston, MA: National Center for PTSD.