Download The structure of PTSD symptoms

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Posttraumatic stress disorder wikipedia , lookup

Treatments for combat-related PTSD wikipedia , lookup

Transcript
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
299
British Journal of Clinical Psychology (2007), 46, 299–313
q 2007 The British Psychological Society
The
British
Psychological
Society
www.bpsjournals.co.uk
The structure of PTSD symptoms: A test of
alternative models using confirmatory factor
analysis
Ask Elklit1 and Mark Shevlin2*
1
2
University of Aarhus, Denmark
University of Ulster (Magee), Northern Ireland
Objectives. This study aimed to examine the structure of self-reported posttraumatic stress disorder (PTSD) symptoms.
Design. Based on previous factor analytic findings and the DSM-IV formulation, six
confirmatory factor models were specified and estimated that reflected different
symptom clusters.
Methods. The analyses were based on responses from 1116 participants who had
suffered whiplash injuries and screened for full or subclinical PTSD using the Harvard
Trauma Questionnaire.
Results. A correlated four-factor model with re-experiencing, avoidance, dysphoria
and arousal factors fitted the data very well. Correlations with criteria measures
showed that these factors were associated with other trauma related variables in a
theoretically predictable way and showed evidence of unique predictive utility.
Conclusions. These results concur with previous research findings using different
trauma populations but do not reflect the current DSM-IV symptom groupings.
Since the inclusion of post-traumatic stress disorder (PTSD) into the Diagnostic and
Statistical Manual of Mental Disorders-III (DSM-III: American Psychiatric Association,
1980) the symptom groupings within the diagnostic criteria have undergone a number
of changes. The current grouping from the DSM-IV (American Psychiatric Association,
1994) specifies three clusters of symptoms: re-experiencing, avoidance and arousal.
However this structure of symptoms has consistently failed to be found using factor
analytic methods. Alternative multifactor solutions have been reported comprising two
to four factors.
The evidence for a two-factor model of posttrauma symptoms initially came from
Horowitz’s (1979) two-factor information processing model of PTSD. Horowitz, Wilner,
and Alvarez (1979) developed the Impact of Event Scale (IES) to measure the two
* Correspondence should be addressed to Mark Shevlin, School of Psychology, University of Ulster at Magee, L’Derry, BT48 7JL,
Northern Ireland (e-mail: [email protected]).
DOI:10.1348/014466506X171540
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
300 Ask Elklit and Mark Shevlin
symptom groups of intrusion and avoidance and since then exploratory and
confirmatory factor analytic studies have provided support for this two factor structure
(e.g. Hodgkinson & Joseph, 1995; Joseph, Williams, Yule, & Hodgkinson, 1993; Shevlin,
Hunt, & Robbins, 2000; Zilberg, Weiss, & Horowitz, 1982). Even with the incorporation
of the arousal symptoms in the DSM-III-R two-factor solutions have been reported.
Taylor, Kuch, Koch, Crockett, and Passey (1998) carried out an exploratory factor
analysis using the PTSD Symptomatology Scale (PSS: Foa, Riggs, Dancu, & Rothbaum,
1993) and a structured clinical interview. They used data from two samples; 103 motor
accident victims and 419 United Nation peacekeepers, who had experienced wartime
action in Bosnia. A similar two-factor solution was reported for both samples with the
intrusion and avoidance symptoms loading together on the first factor and the numbing
and arousal symptoms loading on the second factor. Buckley, Blanchard, and Hickling
(1998) successfully replicated this two-factor structure using confirmatory factor
analysis based on 217 motor accident victims.
Evidence for a three-factor structure of symptoms as proposed by the DSM-IV has
also been found. Cordova, Studts, Hann, Jacobsen, and Andrykowski (2000) tested a
three-factor model with a second order PTSD factor using confirmatory factor analysis in
a sample of 142 breast cancer patients. Their support for the three-factor model was
equivocal: the model ‘ : : : reflected a moderate fit of the symptom structure implied by
the DSM-IV. These findings provide some tentative support for the DSM-IV clustering of
PTSD symptoms’ (p. 301). A post hoc modification of a correlated error between the two
effortful avoidance items was found to significantly improve the fit of the model. This
suggests the presence of a fourth factor, separating the original avoidance factor into a
separate effortful avoidance (of trauma-specific stimuli) factor (C1, C2) and an emotional
numbing factor (C3, C4, C5, C6 and C7).
King, Leskin, King, and Weathers (1998) used an ‘alternative models’ approach to
model testing by specifying four competing models using data from a sample of 524
male military veterans who were interviewed using the Clinician Administered PTSD
Scale (CAPS: Blake et al., 1990). The first model was a correlated four-factor model with
the factors of re-experiencing, effortful avoidance, emotional numbing and arousal.
The second used the same four factors but included two correlated second-order
factors, one factor accounting for the re-experiencing and effortful avoidance symptoms
and the other accounting for the emotional numbing and arousal symptoms. The third
model specified the four first-order factors and a single second-order factor of PTSD. The
fourth model specified a single-factor with all the 17 symptoms loading on this PTSD
factor. The first model that specified a correlated four-factor model was found to be the
best representation of the sample data, although the other models that included the four
first-order factors were acceptable. This suggested that irrespective of the nature and
number of possible high-order factors, the four first-order factors are useful in describing
PTSD symptoms. Interestingly, the lowest correlation between the factors was between
the effortful avoidance and emotional numbing factors, which supports the distinction
between these symptoms.
Asmundson et al. (2000) tested five alternative models using data form 349 primary
care patients who had experienced a range of traumatic accidents. PTSD symptom
scores were assessed using the PTSD Checklist (PCL-C: Weathers, Litz, Huska, & Keane,
1994). The models included a four-factor model with and without a second-order PTSD
factor, a DSM-IV based three-factor model with and without a second-order PTSD factor,
and the two-factor model proposed by Taylor et al. (1998). Based on the results of
confirmatory factor analyses the four-factor model with a second-order PTSD factor was
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
301
found to be the best fitting model. Similarly, DuHamel et al. (2004) tested seven models
using data from 236 cancer survivors. PTSD symptoms were assessed using the PTSD
Checklist (PCL: Weathers & Ford, 1996) and the models were estimated using
confirmatory factor analysis. A correlated four-factor model (re-experiencing, effortful
avoidance, emotional numbing and arousal) was found to be the best fitting model.
Using an alternative models approach this four-factor structure has also be found to
represent the best fitting model using samples of sexually harassed women (Palmieri &
Fitzgerald, 2005) and adult survivors of community violence (Marshall, 2004) based on
the PTSD Checklist.
The two most recent factor analytic studies of PTSD symptoms have proposed
models that more radically differ from the DSM-IV in that they go beyond modelling
avoidance as two separate factors. Simms, Watson, and Doebbeling (2002) suggested a
four-factor model with re-experiencing, avoidance, dysphoria and arousal factors. The
dysphoria factor comprises the emotional numbing symptoms and the irritability/anger,
difficulty sleeping and difficulty concentrating symptoms. The hypervigilance and
exaggerated startle response symptoms comprised the arousal factor. The confirmatory
factor analyses were based on data from a large sample (N ¼ 3; 695) of Gulf War
veterans and non-deployed controls using the PTSD Checklist – Military Version (PCL-M:
Weathers, Huska, & Keane, 1991). Simms et al. found this model provided the best fit
compared with five other models, and this finding was replicated using different
samples. Subsequently, Baschnagel, O’Connor, Colder, and Hawk (2005) replicated this
finding based on the Posttraumatic Diagnostic Scale (Foa, Cashman, Jaycox, & Perry,
1997) using a sample of undergraduate students following the September 11th World
Trade Centre attacks. However, the authors note that the fit of this confirmatory factor
model provided only a ‘slight advantage’ (p. 682) over the four-factor model proposed
by King et al. (1998). McWilliams, Cox, and Asmundson (2005) compared nine models
(all based on specifications described above) using confirmatory factor analysis based on
data from the PTSD module of the National Comorbidity study (for details see Kessler,
Sonnega, Bromet, Highes, & Nelson, 1995) but claimed that none met the criteria for
acceptable model fit based on the fit indices. The models proposed by King et al. and
Simms et al. were rejected on the basis of one fit index although the others suggested a
reasonable degree of fit. McWilliams et al. proceeded to carry out an exploratory
principal components analysis and extracted four factors which were labelled
dysphoria, cued re-experiencing, uncued re-experiencing and rumination. The solution
contained two sizable cross-factor loadings.
Until the publication of the analyses by Simms et al. (2002) there was a degree of
consensus regarding the structure of PTSD symptoms with a four-factor model being
widely supported in the research literature: this model separated the traditional
avoidance factor into a conscious avoidance and emotional numbing factor (Andrews,
Shevlin, Troop, & Joseph, 2004; Asmundson et al., 2000; King et al., 1998). However,
findings reported by Simms et al., McWilliams et al. (2005) and Baschnagel et al. (2005)
all suggested a dysphoria factor composed of symptoms from both the avoidance (more
specifically emotional numbing) and arousal clusters.
This paper aimed to include these two models in a comprehensive analysis of
alternative models of PTSD symptoms. The structure of the six models tested is
presented in Table 1.
The emotional reactivity to trauma cues (B4) and physiological reactivity to trauma
cues (B5) symptoms were assessed using one question (‘Sudden emotional or physical
reaction when reminded of the traumatic event’) so there are four, rather than five,
RE/AV
RE/AV
RE/AV
RE/AV
RE/AV
RE/AV
A/N
A/N
A/N
A/N
A/N
A/N
A/N
A/N
A/N
A/N
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
Two-factor model
(Taylor, et al. 1998)
One-factor
model
AV
AV
A
A
A
A
A
AV
AV
AV
AV
AV
RE
RE
RE
RE
Three –factor
model (DSM-IV)
Model 3
N
N
A
A
A
A
A
AV
AV
N
N
N
RE
RE
RE
RE
Four-factor model
(King et al., 1998)
Model 4a
D
D
D
D
D
A
A
AV
AV
D
D
D
RE
RE
RE
RE
Four-factor model
(Simms et al. 2002)
Model 4b
D
D
R
D
D (cl)
C-RE
U-RE
C-RE
C-RE
R
D
U-RE
R
U-RE (cl)
U-RE
C-RE
Four-factor model
(McWilliams et al., 2005)
Model 4c
Note. P, PTSD; RE, Re-experiencing; AV, Avoidance; N, Numbing; A, Arousal; D, Dysphoria; U-RE, Uncued re-experiencing; C-RE, Cued re-experiencing; R,
Rumination; cl, cross-factor loading.
B1. Intrusive recollection
B2. Recurrent dreams
B3. Event recurring
B4/5. Psychological and
physiological distress
C1. Efforts to avoid thoughts
C2. Efforts to avoid activities
C3. Memory impairment
C4. Diminished interest in activities
C5. Feeling of
detachment from others
C6. Restricted range of affect
C7. Sense of foreshortened future
D1. Sleeping difficulty
D2. Irritability or anger
D3. Difficulty concentrating
D4. Hypervigilance
D5. Exaggerated startle response
DSM-IV PTSD symptom
Model 2
Model 1
Table 1. Model specifications for the alternative models of the PTSD symptoms
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
302 Ask Elklit and Mark Shevlin
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
303
symptoms specifying the re-experiencing cluster. In order to test the broadest range of
models we also tested higher-order variants of the four-factor models. These models
included a single higher-order factor used to explain the variation and covariation
among the four first-order factors. The concurrent validity of the factors from the best
fitting model was assessed by examining their associations with criteria variables. It was
predicted that factors would be significantly related to trauma related cognitions (selfblame for accident, ruminating on reason for accident), emotional responses
(hopelessness), social support (having nobody to rely on), dissociation and general
distress (depression, anxiety). Based on the findings of Simms et al. (2002) it was also
predicted that if found to be consistent with the data, factors representing non-specific
components of PTSD would have the highest associations with variables representing
depressive symptomatology.
Method
Participants
A total of 1709 participants were recruited. The mean age was 43.10 years (SD ¼ 10:30)
and 79% (N ¼ 1349) were female. Males were significantly older than females,
tð1700Þ ¼ 3:58, p , :005, although the mean difference was small (2.19 years).
The participants had been exposed to trauma resulting in whiplash on average 62
months prior to participating in the study and were recruited through the Danish
Society for Polio, Traffic and Accident Victims. The society generally receives referrals
from the Danish National Health Service and other health related sources. All society
members with whiplash (N ¼ 2320) were contacted with a response rate of 74%.
The female/male ratio of the final sample closely matches the complete patient group
(approximately 4:1) and is similar to the findings from epidemiological studies, which
have shown that females have a higher risk of developing and maintaining whiplash
related symptoms (Harder, Veilleux, & Suissa, 1998). All participants were contacted by
post and invited to participate in the study by completing the enclosed questionnaires.
Almost all participants (N ¼ 1527; 89%) sustained whiplash through a motor vehicle
accident and half (N ¼ 831; 49%) sustained other physical injuries. A quarter of the
participants (N ¼ 426; 25%) were hospitalized after the accident, and for those the
mean length of stay was 6.9 days. Almost half the participants contacted a doctor within
24 hours (N ¼ 829; 49%), with smaller numbers waiting up to 48 hours (N ¼ 407; 24%).
Almost all the participants (N ¼ 1601; 94%) had sought medical help within 4 weeks of
the accident. The diagnosis of whiplash was given on the same day for almost half the
participants (N ¼ 779; 46%). All analyses in this study were based only on participants
who met the criteria for clinical or subclinical PTSD (see Measures section for details).
Measures
The Harvard Trauma Questionnaire Part IV (HTQ: Mollica et al., 1992)
The HTQ assess both DSM-IV symptoms and culture-specific symptoms associated with
PTSD. The scale yields both a PTSD diagnosis according to DSM-IV criteria and a
measure of PTSD symptom severity. The HTQ asks the respondents how much each
symptom has bothered them in the last week. The 30 items are answered on a 4-point
Likert-type scale (‘not at all’ (1), ‘a little’ (2), ‘quite a bit’ (3) and ‘all the time’ (4)). The
summed scores provide a score for symptom severity. The first sixteen items were
derived directly from the seventeen DSM-IV criteria for PTSD. The HTQ uses one item to
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
304 Ask Elklit and Mark Shevlin
assess both psychological and physiological reactions to events that symbolize or
resemble aspects of the traumatic event (in accordance with DSM-IV this item is part of
the re-experiencing cluster). The items are divided into three subscales that correspond
to the three main symptom groups of PTSD: re-experiencing, avoidance and arousal.
Similar to Mollica et al. (1992), Bödvarsdóttir and Elklit (2004) report high estimates of
reliability for each of the subscales and the scale as a whole (a ¼ 0:94). Following
the DSM-IV, the diagnosis of PTSD was made if participants reported at least one reexperiencing symptom, three avoidance symptoms and two arousal symptoms.
Subclinical PTSD was characterized by participants endorsing at least one symptom in
each of the three PTSD symptom clusters (Stein, Walker, Hazen, & Forde, 1997).
A symptom was rated as present if the item corresponding to the symptom was scored 3
(‘quite a bit’) or greater (this is a more conservative approach than that taken by others,
such as Foa et al. (1997) who include scores corresponding to our ‘point 2’ or greater for
symptom endorsement). Complete HTQ scores for the PTSD classification were
provided by 1491 of the participants.
As part of the validation process responses to four other items from the HTQ were
used as criteria measures. These items reflected cognitions (self-blame for accident,
ruminating on reason for accident), emotional responses (hopelessness) and social
support (having nobody to rely on) that are associated with the traumatic response.
Based on diagnosis using the HTQ 42.2% of the participants (N ¼ 639) were
classified as having PTSD, and 32.0% (N ¼ 477) as having subclinical PTSD making the
total sample size 1116. PTSD was absent in 25.2% (N ¼ 375) of the sample. All analyses
in this study were based only on participants who met the criteria for clinical or
subclinical PTSD.
The Trauma Symptom Checklist (TSC-33: Briere & Runtz, 1989)
The TSC-33 was originally developed to assess the long-term impact of rape and child
sexual abuse. As TSC-33 is also responsive to physical abuse, Briere and Runtz (1989)
suggested that TSC-33 could be responsive to a variety of traumatic experiences.
According to Elklit (1990), this is to be expected, as the items cover symptoms that are
associated with various psychological states, and highly overlapping with the Symptom
checklist (SCL-90) (Derogatis & Coons, 1993) and the Hopkins Symptom Checklist
(HSCL) (Derogatis, Lipman, Rickels, Uhlenhuth, & Covi, 1974). In this study three
subscales were used (with Cronbach’s alpha in parentheses): depression (0.70), anxiety
(0.66) and dissociation (0.68). The items are answered on a 4-point Likert scale from
‘never’ (1) to ‘very often’ (4). The depression subscale has ten items, anxiety eight items
and dissociation seven items. The TSC-33 has previously been found to be internally
consistent, both subscales and total scores, and has good discriminant validity (Briere &
Runtz, 1989; Elklit, 1990).
Analysis
Six confirmatory factor models were specified and estimated by using LISREL 8.72
(Jöreskog & Sörbom, 2005a). A covariance matrix and an asymptotic weight matrix
were computed using PRELIS 2.72 (Jöreskog & Sörbom, 2005b) based on the first
sixteen items of the HTQ, and the model parameters estimated using maximum
likelihood. The use of an asymptotic weight matrix allows for weaker assumptions
regarding the distribution of the observed variables and results in improved fit and test
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
305
statistics (Curran, West, & Finch, 1996; Satorra, 1992). For all variables the percentage
of missing values was very small (.18–4.85%). The missing values were imputed using
the EM algorithm. Bunting, Adamson, and Mulhall (2002) demonstrated the benefits of
using the EM algorithm to treat missing data over traditional methods such as listwise
deletion.
The structure of the six models is shown in Table 1. All factors were allowed to
correlate and no correlated errors were included in any of the models. Following the
guidelines suggested by Hoyle and Panter (1995) the goodness of fit for each model was
assessed using a range of fit indices including the Satorra–Bentler scaled chi-squared
(S–Bx2), the Incremental Fit Index (IFI: Bollen, 1989), and the Comparative Fit Index
(CFI: Bentler, 1990). A non-significant chi-squared, and values greater than .95 for the IFI
and CFI are considered to reflect acceptable model fit. In addition, the Root Mean Square
Error of Approximation (RMSEA: Steiger, 1990) with 90% confidence intervals (90%CI)
were reported, where a value less than .05 indicates close fit and values up to .08
indicating reasonable errors of approximation in the population (Jöreskog & Sörbom,
1993). The standardized root-mean-square residual (SRMR: Jöreskog & Sörbom, 1981)
has been shown to be sensitive to model mis-specification and its use recommended by
Hu and Bentler (1999). Values less than .08 are considered to be indicative of acceptable
model fit (Hu & Bentler, 1998). The comparative fit of the models was assessed using the
Expected Cross Validation Index (ECVI; Browne & Cudeck, 1989), an index used for the
purposes of model comparison, with the smallest value being indicative of the best
fitting model.
Results
Confirmatory factor analysis and reliability analysis
The fit indices are reported in Table 2. On the basis of meeting the criteria associated
with the RMSEA and the SRMR fit indices only Model 4 and Model 5 were judged to
exhibit reasonable model fit (for Model 4 the values of the IFI and CFI were just below
the stringent cut-off value of .95). Although the chi-squared for Models 2 and 4 were
large relative to the degrees of freedom, and statistically significant, this should not lead
to the rejection of the models as the large sample size increases the power of the test
(Tanaka, 1987). The increased power of the chi-squared test can result in models with
no serious misspecification being rejected as minor discrepancies between the sample
and implied covariance matrix are detected. Models 4 and 5 also had lower ECVI values
when compared with the other models. Models 4 and 5 are not tested. Hence, a
statistical test of difference of fit, such as likelihood ratio difference test, was not
possible. However, Model 5 has a lower chi-squared/df ratio (4.40) compared with
Model 4 (6.06), has a lower RMSEA (the upper 90% confidence interval for Model 5
(.060) is less than the lower 90% confidence interval for Model 4 (.062)), a lower ECVI
value, higher values for the IFI and CFI and a lower SRMR value. On the basis of this it
is proposed that Model 5 represents an adequate description of the data, and is judged
the best of the alternative models1. However, the adequacy of Model 5 must also
be considered in terms of the conceptual clarity of the parameter estimates.
1
All models were analysed using data for all participants, not only those who met the criteria for clinical or subclinical PTSD,
and the estimates and model selection were very similar.
2011.22 104 (.00)
.129 (.123–.133)
2.00 (1.7–2.3)
.69
.69
.098
Model 1
892.31 103 (.00)
.083 (.078–.083)
.90 (.77–.95)
.87
.87
.078
Model 2
1124.77 101 (.00)
.095 (.090–.100)
1.13 (.98–1.17)
.84
.84
.089
Model 3
594.36 98 (.00)
.067 (.062–.072)
.63 (.53–.67)
.92
.92
.063
Model 4a
704.40 100 (.00)
.074 (.069–.079)
.74 (.62–.77)
.90
.90
.076
Model 4a-h
431.84 98 (.00)
.055 (.050–.060)
.48 (.40–.52)
.95
.95
.050
Model 4b
477.70 100 (.00)
.058 (.053–.063)
.52 (.44–.56)
.94
.94
.057
Model 4b-h
858.23 96 (.00)
.084 (.079–.089)
.89 (.76–.93)
.88
.88
.082
Model 4c
894.00 98 (.00)
.085 (.080–.090)
.92 (.79–.96)
.87
.87
.081
Model 4c-h
Note. Models with ‘-h’ extension denote higher-order model. S–Bx2, Satorra–Bentler scaled chi-square; RMSEA, Root Mean Square Error of Approximation;
CI, confidence intervals; ECVI, Expected Cross Validation Index; IFI, Incremental Fit Index; CFI, Comparative Fit Index; SRMR, Standardized Root Mean Residual.
S–Bx2 df (p)
RMSEA 90% CI
ECVI 90% CI
IFI
CFI
SRMR
Item
Table 2. Fit indices for the alternative factor models of the Harvard trauma questionnaire
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
306 Ask Elklit and Mark Shevlin
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
307
The standardized factor loadings are presented in Table 3. All loadings were statistically
significant (p , 0:05).
The factor correlations and estimates of reliability (Cronbach’s alpha) are reported in
Table 4.
Table 3. Factor loadings for four-factor model (Model 4b) of PTSD symptoms
Re-experience
B1. Intrusive recollection
B2. Recurrent dreams
B3. Event recurring
B4/5. Psychological & physiological distress
C1. Efforts to avoid thoughts
C2. Efforts to avoid activities
C3. Memory impairment
C4. Diminished interest in activities
C5. Feeling of detachment from others
C6. Restricted range of affect
C.7 Sense of foreshortened future
D1. Sleeping difficulty
D2. Irritability or anger
D3. Difficulty concentrating
D4. Hypervigilance
D5. Exaggerated startle response
Avoidance
Dysphoria
Arousal
.66
.53
.74
.61
.60
.66
.21
.59
.64
.52
.64
.27
.49
.45
.73
.60
Note. All factor loadings statistically significant (p , :05).
Table 4. Factor correlations and reliability estimates for four-factor model (Model 4b) of PTSD
symptoms
Re-experiencing
Avoidance
Dysphoria
Arousal
Re-experience
Avoidance
Dysphoria
Arousal
(.71)
.75
.20
.56
(.57)
.41
.66
(.68)
.49
(.61)
Note. All factor correlations statistically significant (p , :05).
All factor correlations were statistically significant (p , :05) and ranged from r ¼ :20
to r ¼ :75. The estimates of reliability ranged from a ¼ :57 (avoidance) to a ¼ :71
(re-experiencing).
Validity analyses
Scale scores based on the factors from Model 5 were calculated by summing the relevant
items from the HTQ. These scale scores were correlated with the depression, anxiety
and dissociation scale scores from the TSC and the four items from the HTQ that
reflected trauma related cognitions, emotional responses and social support.
Correlations were corrected for attenuation due to measurement error (Gulliksen,
1987). The correlations are reported in Table 5.
.36 (.25)
.38 (.24)
.88 (.61)
.58 (.38)
.37 (.25)
.39 (.24)
.76 (.51)
.70 (.44)
Anxiety
.45
.40
.84
.44
(.31)
(.25)
(.57)
(.28)
Dissociation
.15
.24
.10
.06
(.13)
(.18)
(.08)
(.05)
Self-blame
.18 (.15)
.28 (.21)
.68 (.56)
.34 (.26)
Hopelessness
.37 (.31)
.40 (.30)
.29 (.24)
.26 (.20)
Rumination
.21 (.18)
.31 (.23)
.51 (.42)
.39 (.30)
Social support
Note. The social support item is worded negatively, and asks about how often the respondent felt that they ‘had nobody to rely on’; the largest correlation with each
criterion measure is presented in boldface; all correlations significant (p , :05); N ¼ 1116 for all correlations.
Re-experiencing
Avoidance
Dysphoria
Arousal
Depression
Table 5. Pearson correlations corrected for measurement error (uncorrected) between PTSD scales and criteria measures
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
308 Ask Elklit and Mark Shevlin
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
309
All the correlations are statistically significant. Tests of differences between nonindependent correlation (see Steiger, 1980) showed that the correlations between the
dysphoria scale and the TSC subscales were significantly higher (p , :05) than for the
other PTSD scales. The corrected correlations were all high and positive ranging from
r ¼ :76 to r ¼ :88. The dysphoria scale also had the highest corrected correlation with
the items measuring hopelessness (r ¼ :68) and social support (r ¼ :51). These were
significantly higher than the correlations for the other PTSD scales (p , :05).
The avoidance scale had the highest correlation with rumination scores (r ¼ :40) and
self-blame scores (r ¼ :24), although the correlations were not significantly different
from those for re-experiencing (p . :05). There was evidence of unique predictive
utility of the PTSD scales as the pattern of correlations between the PTSD scales and
each of the criteria measures is dissimilar.
Discussion
The aim of this study was to test alternative models of the structure of PTSD symptoms.
Six models based on previous research findings were specified and estimated by using
confirmatory factor analysis. On the basis of fit indices reported in Table 2 a four-factor
model (Model 5) was found to provide an adequate fit to the data and was considered to
be better than the alternative models. This model had been proposed by Simms et al.
(2002) including re-experiencing, avoidance, dysphoria and arousal factors. The fourfactor model proposed by King et al. (1998) was also found to be an acceptable
description of the data, although not as good as that proposed by Simms et al.. Fit indices
showed that the models proposed by Simms et al. and King et al. were much superior to
the structure proposed by the DSM-IV. The relative fit of these three models directly
replicates those reported by Simms et al. and McWilliams (2005). This study supported
the Simms model based on data from a sample representing a non-American population
thereby replicating the four-factor structure in a different cultural context. In addition,
this study was based on a previously unused measure of PTSD.
All the factor loadings for this model were positive and statistically significant. The
loadings reported in Table 3 are generally slightly lower than those reported by Simms
et al. (2002), in particular the memory impairment (.21) and sleep disturbance (.27)
symptoms. The estimates of reliability are low to moderate and lower than those
reported by Simms et al.. The biggest difference is for the dysphoria scale (a ¼ :86 in
Simms et al.) which could account for the two weak factor loadings associated with that
factor.
The pattern of factor correlations was reported in Table 4 and it showed that the
re-experiencing and avoidance scales were highly correlated (r ¼ :75) as were the
avoidance and arousal scales (r ¼ :66). These correlations were higher than those
reported by Simms et al. (2002) who reported a correlation of r ¼ :58 between the
re-experiencing and avoidance scales and r ¼ :75 between the avoidance and arousal
scales. The correlation of r ¼ :20 between the re-experiencing and dysphoria scales was
much lower than the correlation of r ¼ :61 reported by Simms et al.. Interestingly, the
three correlations that are associated with the dysphoria scale are the smallest. This
suggests that the dysphoria scale shares less conceptual overlap with the other scales
than they share among themselves, that is, it does not fit within the cluster of the other
symptoms. This fits with the hierarchical view of anxiety and depression (Brown,
Chorpita, & Barlow, 1998; Clark & Watson, 1991) that such disorders have specific and
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
310 Ask Elklit and Mark Shevlin
non-specific components often referred to as ‘general distress’ or negative affectivity’.
As Simms et al. suggested the dysphoria factor could be a non-specific component of
PTSD. A model that comprised four specific and one non-specific component of PTSD
was found to be a well-fitting model by Andrews, Joseph, Shevlin, and Troop (2006).
They tested a series of confirmatory factor models based on data from emergency
personnel using the PSS. The best fitting model specified a general factor with all PTSD
symptom measures loading on it, along with four correlated factors of re-experiencing,
avoidance, numbing and arousal. The general factor was labelled ‘psychological distress’
and the factor loadings for the four specific symptom groups were lower when the
model included this general factor. This supports the hypothesis of specific and nonspecific components of PTSD, although the small percentage of participants (8.04%)
with PTSD makes it difficult to generalize these findings to populations with PTSD.
The correlations reported in Table 5 show that the PTSD scales correlate positively
with other trauma related cognitions, emotional responses, social support and
significant dimensions of psychopathology. It adds further weight to the hypothesis that
the dysphoria factor may be a non-specific component of PTSD as the dysphoria scale
was positively correlated with all other criteria measures but the highest correlation was
with depression scores (r ¼ :88). The correlations of the other PTSD scales and the
criteria measures imply substantial unique predictive utility. The pattern of corrected
correlations with the criteria measures was different for each of the three PTSD scales;
the re-experiencing scale correlated highest with dissociation (r ¼ :45), the highest for
avoidance was rumination (r ¼ :40), and the highest for arousal was anxiety (r ¼ :70).
Any future studies in this area could incorporate design elements to overcome the
limitations of this study. First, clinical interview-based diagnosis of PTSD would be
superior to those based on self-report. Second, a longitudinal study would allow the
temporal changes in the structure of PTSD to be examined. Third, the inclusion of other
groups who have experienced different traumas would allow tests of invariance of PTSD
structure. Fourth, the low reliability of the avoidance and arousal factors could be
improved by using additional indicators. Finally, future research should use a range of
external criterion variables in an effort to further understand the nature of the specific
and the non-specific components of PTSD.
Clinical theory and practice might be affected in several ways if the findings from this
study were to be consistently replicated in future studies. First, the findings reported in
this, and many other papers, have consistently failed to support the existing threesymptom cluster model proposed by the currewnt DSM-IV. Second, it would be
expected that future editions of the DSM would accommodate such findings, and
change the diagnostic criteria accordingly. Obviously debate would focus on whether
only specific components, or all components, of PTSD would be included in the
formulation of any diagnosis. Third, treatment might also be more focused, and possibly
effective, if the dysphoria component of PTSD was to be conceived as secondary to the
specific components and less important to target in early phases of intervention.
In conclusion, confirmatory factor analyses showed that a correlated four-factor
model of PTSD symptoms with re-experiencing, avoidance, dysphoria and arousal
factors was the best fitting model based on a large sample of participants exposed to
trauma that resulted in whiplash. These results concur with previous research findings
using different trauma populations but do not reflect the current DSM-IV symptom
groupings. In all studies testing competing models of PTSD symptoms using
confirmatory factor analysis (Andrews et al., 2006; Andrews et al., 2004; Asmundson
et al., 2000; DuHamel et al., 2004; McWilliams et al., 2005; Simms et al., 2002) the
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
311
DSM-IV model of PTSD symptoms has never been found to be an acceptable model.
The dysphoria factor reported in this study may be a non-specific component of PTSD,
with the re-experiencing, avoidance and arousal factors being the specific components.
References
American Psychiatric Association (1980). Diagnostic and statistical manual of mental disorders
(3rd ed.). Washington, DC: Author.
American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders
(4th ed.). Washington, DC: Author.
Andrews, L., Joseph, S., Shevlin, M., & Troop, N. (2006). Confirmatory factor analysis of
posttraumatic stress symptoms in emergency personnel: An examination of seven alternative
models. Personality and Individual Differences, 41, 213–224.
Andrews, L., Shevlin, M., Troop, N., & Joseph, S. (2004). Multidimensionality of intrusion and
avoidance: Alternative factor models of the Impact of Event Scale. Personality and Individual
Differences, 36, 431–446.
Asmundson, G. J. G., Frombach, I., McQuaid, J., Pedrelli, P., Lenox, R., & Stein, M. B. (2000).
Dimensionality of posttraumatic stress symptoms: A confirmatory factor analysis of DSM-IV
symptom clusters and other symptom models. Behaviour Research and Therapy, 38,
203–214.
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. (1990). A clinician
rating scale for assessing current and lifetime PTSD: The CAPS-1. Behavior Therapist, 13,
187–188.
Bödvarsdóttir, I., & Elklit, A. (2004). Psychological reactions in Icelandic earthquake survivors.
Scandanavian Journal of Psychology, 45, 3–13.
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Briere, J., & Runtz, M. (1989). The Trauma Symptom Checklist (TSC-33): Early data on a new scale.
Journal of Interpersonal Violence, 4, 151–163.
Browne, M. W., & Cudeck, R. (1989). Single sample cross-validation indices for covariation
structures. Multivariate Behavioral Research, 24, 445–455.
Brown, T. A., Chorpita, B. F., & Barlow, D. H. (1998). Structural relationships among dimensions of
the DSM-IV anxiety and mood disorders and dimensions of negative affect, positive affect, and
autonomic arousal. Journal of Abnormal Psychology, 107, 179–192.
Buckley, T. C., Blanchard, E. B., & Hickling, E. J. (1998). A confirmatory factor analysis of
posttraumatic stress symptoms. Behaviour Research and Therapy, 36, 1091–1099.
Bunting, B. P., Adamson, G., & Mulhall, P. (2002). A Monte Carlo examination of MTMM model with
planned incomplete data structures. Structural Equation Modeling, 9, 369–389.
Clark, L. A., & Watson, D. (1991). Tripartite model of anxiety and depression: Psychometric
evidence and taxonomic implications. Journal of Abnormal Psychology, 100, 316–336.
Cordova, M., Studts, J., Hann, D., Jacobsen, P., & Andrykowski, M. (2000). Symptom structure of
PTSD following breast cancer. Journal of Traumatic Stress, 13, 301–319.
Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and
specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29.
Derogatis, L. R., & Coons, M. L. (1993). Self-report measures of stress. In L. Goldberger &
S. Breznitz (Eds.), Handbook of stress (2nd ed., pp. 200–234). New York: The Free Press.
Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., & Covi, L. (1974). The Hopkins
Symptom Checklist (HSCL): A self-report symptom inventory. Behavioural Science, 19, 1–15.
DuHamel, K. N., Ostroff, J., Ashman, T., Winkel, G., Mundy, E. A., Keane, T. M., et al. (2004).
Construct validity of the posttraumatic stress disorder checklist in cancer survivors: Analyses
based on two samples. Psychological Assessment, 16, 255–266.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
312 Ask Elklit and Mark Shevlin
Elklit, A. (1990). Måling af belastninger efter voldeligt overfald med TSC-33 (traume symptom
checkliste). [Measurement of stress after physical assault using the TSC-33 (trauma symptom
checklist)]. Nordisk Psykologi, 42, 281–289.
Foa, E. B., Cashman, L., Jaycox, L., & Perry, K. (1997). The validation of a self-report measure of
posttraumatic stress disorder: The posttraumatic diagnostic scale. Psychological Assessment,
9, 445–451.
Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993). Reliability and validity of a brief
instrument for assessing posttraumatic stress disorder. Journal of Traumatic Stress, 6,
459–473.
Gulliksen, H. (1987). Theory of mental tests. Hillsdale, NJ: Lawrence Erlbaum.
Harder, S., Veilleux, M., & Suissa, S. (1998). The effect of socio-demographic and crash-related
factors on the prognosis of whiplash. Journal of Clinical Epidemiology, 51, 377–384.
Hodgkinson, P., & Joseph, S. (1995). Factor analysis of the impact of event scale with female bank
staff following a raid. Personality and Individual Differences, 19, 773–775.
Horowitz, M. (1979). Psychological response to serious life events. In V. Hamilton & D. M. Warburton
(Eds.), Human stress and cognition: An information processing approach. New York: Wiley.
Horowitz, M., Wilner, N., & Alvarez, W. (1979). Impact of event scale: A measure of subjective
stress. Psychosomatic Medicine, 41, 209–218.
Hoyle, R. H., & Panter, A. T. (1995). Writing about structural equation models. In R. H. Hoyle (Ed.),
Structural equation modeling: Concepts, issues and applications (pp. 158–198). London:
Sage.
Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to
underparameterized model misspecification. Psychological Methods, 4, 424–453.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Jöreskog, K., & Sörbom, D. (1981). LISREL V: Analysis of linear structural relationships by the
method of maximum likelihood. Chicago: National Educational Resources.
Jöreskog, K., & Sörbom, D. (2005a). LISREL 8.72. Chicago: Scientific Software.
Jöreskog, K., & Sörbom, D. (2005b). PRELIS 2.72. Chicago: Scientific Software.
Jöreskog, K. G., & Sörbom, D. (1993). Structural equation modeling with the SIMPLIS command
language. Chicago: Scientific Software.
Joseph, S., Williams, R., Yule, W., & Hodgkinson, P. (1993). The Herald of free enterprise disaster:
Measuring post-traumatic symptoms 30 months on. British Journal of Clinical Psychology,
32, 327–331.
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., Leskin, G., King, L., & Weathers, F. (1998). Confirmatory factor analysis of the clinicianadministered PTSD Scale: Evidence for the dimensionality of posttraumatic stress disorder.
Psychological Assessment, 10, 90–96.
McWilliams, L. A., Cox, B. J., & Asmundson, G. J. G. (2005). Symptom structure of posttraumatic
stress disorder in a nationally representative sample. Journal of Anxiety Disorders, 19,
626–641.
Mollica, R. F., Caspi-Yavin, Y., Bollini, P., Truong, T., Tor, S., & Lavelle, J. (1992). The harvard trauma
questionnaire: Validating a cross-cultural instrument for measuring torture, trauma and
posttraumatic stress disorder in Indochinese refugees. Journal of Nervous and Mental
Disease, 180, 111–116.
Satorra, A. (1992). Asymptotic robust inferences in the analysis of mean and covariance structures.
Sociological Methodology, 22, 249–278.
Satorra, A., & Bentler, P. (2001). A scaled difference chi-square test statistic for moment structure
analysis. Psychometrika, 66, 507–514.
Shevlin, M., Hunt, N., & Robbins, I. (2000). A confirmatory factor analysis of the impact of event
scale using a sample of world war two and Korean war veterans. Psychological Assessment,
12, 414–417.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Factor structure of PTSD symptoms
313
Simms, L. J., Watson, D., & Doebbeling, B. N. (2002). Confirmatory factor analyses of posttraumatic
stress symptoms in deployed and nondeployed veterans of the Gulf war. Journal of Abnormal
Psychology, 111, 637–647.
Steiger, J. H. (1980). Tests for comparing elements of a correlation matrix. Psychological Bulletin,
87, 245–251.
Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation
approach. Multivariate Behavioral Research, 25, 173–180.
Stein, M. B., Walker, J. R., Hazen, A. L., & Forde, D. R. (1997). Full and partial posttraumatic stress
disorder: Findings from a community survey. American Journal of Psychiatry, 154,
1114–1149.
Taylor, S., Kuch, K., Koch, W. J., Crockett, D. J., & Passey, G. (1998). The structure of posttraumatic
stress symptoms. Journal of Abnormal Psychology, 107, 154–160.
Weathers, F., & Ford, J. (1996). Psychometric properties of the PTSD checklist (PCL-C, PCL-S,
PCL-M, PCL-PR). In B. H. Stamm (Ed.), Measurement of stress, trauma, and adaptation.
Lutherville, MD: Sidran Press.
Weathers, F., Huska, J., & Keane, T. (1991). The PTSD checklist military version (PCL-M). Boston,
MA: National Center for PTSD.
Weathers, F. W., Litz, B. T., Huska, J. A., & Keane, T. M. (1994). The PTSD checklist (PCL). Boston,
MA: National Center for PTSD.
Zilberg, N. J., Weiss, D. S., & Horowitz, M. J. (1982). Impact of event scale: A cross-validation study
and some empirical evidence supporting a conceptual model of stress response syndromes.
Journal of Consulting and Clinical Psychology, 50, 407–414.
Received 2 July 2006; revised version received 15 December 2006