Clinical perspective

The Latent Structure of Acute Stress Disorder: A Posttraumatic Stress Disorder Approach

Cherie Armour University of South Denmark and University of Ulster

at Magee Campus

Ask Elklit University of South Denmark

Mark Shevlin University of Ulster at Magee Campus

Acute stress disorder (ASD) was first included in the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM–IV; American Psychiatric Association, 1994) to account for the psychological symptoms present during the one-month period between trauma exposure and a posttraumatic stress disorder (PTSD) diagnosis. The diagnostic criteria sets of both ASD and PTSD are similar; however, ASD includes additional dissociative items. Factor analytic research into ASD is rare, whereas there is a plethora of research on the factor structure of PTSD symptoms. This study tested whether the latent structure of ASD is similar to the latent structure of PTSD. Five models were tested by using data from Danish rape victims (N � 380); a unidimensional model, the DSM–IV 4-factor ASD model, a King, Leskin, King, and Weathers (1998) replication model, a Simms, Watson, and Doebbeling (2002) replication model, and a 3-factor model. Model fit was assessed by using a number of fit indices, including the root-mean-square error of approximation, comparative fit index, Tucker-Lewis index, and standardized root-mean-square residual. However, based on the fit indices, 3 models were deemed indistinguishable. Chi-square difference tests concluded that a 3-factor model and two 4-factor models did not differ in fit. Overall, the current 4-factor ASD latent structure proposed by the DSM–IV was not supported. A 3-factor structure was deemed preferential on the basis of parsimony. Furthermore, of all models, the unidimensional model provided the poorest fit to the data. These findings are pertinent given that the DSM-5 ASD task force is considering implementing either a 4-factor conceptualization or a unidimensional approach to the ASD diagnosis.

Keywords: acute stress disorder, posttraumatic stress disorder, CFA, rape

An ongoing debate relates to whether or not acute stress disorder (ASD) and posttraumatic stress disorder (PTSD) are similar or different constructs (Brooks et al., 2008). Indeed, the ASD diag- nostic criteria were closely modeled on the PTSD diagnostic criteria (cf. Bryant & Harvey, 2000). The main differences are the addition of the ASD dissociative items and the diagnostic time- frame. Therefore, an interesting line of enquiry relates to whether the latent structure of ASD is similar to the latent structure of PTSD. A plethora of research has proven that four-factor structures better account for the underlying dimensionality of PTSD com- pared to the Diagnostic and Statistical Manual of Mental Disor-

ders (4th ed.; DSM–IV; American Psychiatric Association [APA], 1994) three-factor structure. Notably, a less explored area relates to the latent structure of ASD.

ASD was introduced as a formal diagnosis into the DSM no- menclature in 1994 (DSM–IV; APA, 1994). A diagnosis of ASD requires that an individual must experience a traumatic event, which results in an intense emotional reaction (Criterion A). The individual must endorse a number of peritrauma symptoms from a dissociative (Criterion B) symptom factor. The individual must also endorse a number of posttrauma symptoms from the intrusion (Criterion C) symptom factor. In addition, individuals must show marked avoidance (Criterion D) and marked arousal. Other diag- nostic requirements relate to a display of significant occupational or social impairment or distress (Criterion F), which lasts between 2 days and 4 weeks (Criterion G). ASD was introduced as an attempt to identify those individuals who had experienced a trauma and would subsequently develop PTSD. The ASD diagnosis was therefore introduced to fill a diagnostic gap, as PTSD can only be diagnosed when symptoms have persisted for at least one month (Bryant, 2007).

Posttraumatic stress disorder (PTSD) was first included in the DSM–III (APA, 1980). The current DSM–IV (APA, 1994) diag- nostic criteria states that for an individual to receive a diagnosis of PTSD they must have experienced a traumatic event (Criterion A1), which has resulted in an intense feeling of fear, helplessness,

This article was published Online First August 15, 2011. Cherie Armour, National Centre for Psychotraumatology, University of

Southern Denmark, Odense Campus, Funnen, Denmark, and School of Psychology, Faculty of Life and Health Sciences, University of Ulster at Magee Campus, L’Derry , Northern Ireland; Ask Elklit, National Centre for Psychotraumatology, University of Southern Denmark, Odense Cam- pus, Funnen, Denmark; Mark Shevlin, School of Psychology, Faculty of Life and Health Sciences, University of Ulster at Magee Campus, L’Derry, Northern Ireland.

Correspondence concerning this article should be addressed to Cherie Armour, School of Psychology, Faculty of Life and Health Sciences, University of Ulster at Magee Campus, L’Derry BT48 7JL, Northern Ireland. E-mail: [email protected]

Psychological Trauma: Theory, Research, Practice, and Policy © 2011 American Psychological Association 2013, Vol. 5, No. 1, 18 –25 1942-9681/13/$12.00 DOI: 10.1037/a0024848

18

or horror (Criterion A2). In addition, it is a requirement that individuals also endorse a number of symptoms found across three symptom factors termed intrusion, avoidance/emotional numbing, and arousal (Criterion B, C, and D, respectively). The reported symptoms must persist in excess of one month (Criterion E) and cause clinically significant distress or impairment in the daily functioning of the individual (Criterion F). The DSM–IV diagnos- tic criterion for PTSD therefore neglects the psychological reac- tions that occur in the one-month period between initial trauma exposure and diagnosis.

Since the inclusion of ASD in the DSM–IV, multiple studies have assessed the predictive utility of ASD with regard to identi- fying individuals who will subsequently meet the criteria for PTSD. Most studies confirm that ASD has good predictive utility for PTSD (recently in Elklit & Brink, 2004; Kangas, Henry, & Bryant, 2005; Kassam-Adams, Fleisher, & Winston, 2009). Re- sults have reported that between 30% (Creamer, O’Donnell, & Pattison, 2004) and 83% (Brewin, Andrews, Rose, & Kirk, 1999; Bryant & Harvey, 1998) of individuals diagnosed with ASD are subsequently diagnosed with PTSD, and as such the ASD diagno- sis enables early intervention that facilitates the reduction of long- term psychiatric disorder (Bryant, Molds, Guthrie, & Nixon, 2005). However, there is also evidence that many individuals who are diagnosed with PTSD did not initially meet the criteria for ASD; estimates range from 10% to 72% (Schnyder, Moergeli, Klaghofer, & Buddeberg, 2001; Creamer, O’Donnell, & Pattison, 2004; respectively). Nonetheless, it is important to consider that study results may vary as a function of the varying methodologies and populations. However, Bryant and Harvey (2000) stated that although studies indicated that approximately three quarters of individuals who are given an ASD diagnosis go on to develop PTSD, only a small minority of people with PTSD had an initial diagnosis of ASD. Thus, it has been suggested that the claim, that ASD diagnosis alone is a useful tool for identifying trauma victims who will eventually develop PTSD, may have been premature (Bryant & Harvey, 2000). Indeed, Bryant (2003) stated that there may “. . . be greater utility in focusing on the interaction between [ASD] symptoms, biological responses, and cognitive factors in predicting who will develop PTSD” (p. 793). Bryant further pro- posed that the interaction of these factors “. . . may enhance the predictive power provided by the ASD diagnosis” (p. 793). How- ever, it is still reasonable to suggest that ASD and PTSD are phenomenologically similar. Indeed, the ASD diagnostic criteria were closely modeled on the PTSD diagnostic criteria (cf. Bryant & Harvey, 1998; Bryant & Harvey, 2000). The main differences are the addition of the dissociative items, found in the ASD diagnostic criteria, and the diagnostic timeframe.

Given the extensive factor analytic literature of PTSD symp- toms (cf. Armour, McBride, Shevlin, & Adamson; Armour & Shevlin, 2010; Elhai, Ford, Ruggiero, & Frueh, 2009; Elklit, Armour, & Shevlin, 2010; Elklit & Shevlin, 2007; King, Leskin, King, & Weathers, 1998; Simms, Watson, & Doebbeling, 2002) the lack of factor analytic ASD research is apparent. To date, four studies have assessed this issue. Cardeña, Koopman, Classen, Waelde, and Spiegel (2000) conducted an exploratory factor anal- ysis (EFA) on data from a sample of survivors of a firestorm (N � 187). Data from the Stanford Acute Stress Reaction Questionnaire (SASRQ) yielded a three-factor structure. The three factors were termed dissociation, reexperiencing, and anxiety/hyperarousal.

This structure accounted for 56% of the variance; however, not all items were placed in the expected factors, for example, “the item related to the individuals mind going blank” loaded on the anxiety/ hyperarousal factor. Bryant et al. (2000) conducted an EFA on data from a sample of accident and assault victims (N � 99) and a sample of bushfire survivors (N � 107). Data from accident and assault victims samples yielded a three-factor solution, whereas, the bushfire survivors sample yielded a four-factor solution.

Brooks et al. (2008) conducted the first confirmatory factor analysis (CFA) of ASD symptoms. They used participants (N � 587) who had been admitted to a level-one trauma center in Australia for various injuries, including work-related accidents, transport accidents, and assaults. Forty-four (7.5%) participants were diagnosed with ASD based on the Acute Stress Disorder Interview (ASDI; Bryant, Harvey, Dang, & Sackville, 1998). However, all participants were included in the analysis. A four- factor correlated model was specified (cf. Table 2). Results con- cluded that the model provided good fit to the data, �2(89, N � 584) � 154.90, p � .001, comparative fit index (CFI) � .97, Tucker-Lewis index (TLI) � .99, root-mean-square error of ap- proximation (RMSEA) � .036. However, as the model produced strong interfactor correlations the authors also tested a model with a second-order factor. The fit statistics for this model were com- parable to the first model, �2(75, N � 584) � 159.68, p � .001, CFI � .97, TLI � .98, RMSEA � .032, and therefore the authors reported that on the basis of parsimony the preferred model was the four-factor single-order correlated model. However, we sug- gest that the second-order model was more parsimonious as it used two less parameters (e.g., four higher order loadings compared to six correlations).

Most recently, Wang, Li, Shi, Zhang, and Shen (2010) extended the factor analytic research by using a CFA alternative models approach. They used data from a sample of Chinese earthquake victims (N � 353). Fifty-four (15.3%) individuals were identified as probable ASD cases based on the Acute Stress Disorder Scale (ASDS; Bryant, Moulds, & Guthrie, 2000). Again, all participants were included in the analysis. They specified five ASD factor models: the DSM–IV four-factor structure, the three-factor struc- ture, and four-factor structure proposed by EFA (cf. Bryant et al., 2000); a three-factor structure combining intrusion and arousal symptoms; and the DSM–IV model with a second-order factor. Results concluded that the DSM–IV model provided the best fit to the data.

To date, both CFA studies of ASD (Brooks et al., 2008; Wang et al., 2010) provided support for the four-factor conceptualization as outlined in the DSM–IV. Brooks et al. (2008) welcomed the empirical support provided by their study. They stated that “It is worth considering the current finding in the context of the robust finding that chronic PTSD is best explained by 4-factor models . . .” (p. 354). The authors also discussed the conceptual similarity between the dissociative items in ASD and the passive-avoidance items (e.g., numbing, withdrawal) in PTSD. In addition, Brooks et al. (2008) highlighted that further re- search into the factor structure of ASD is required for a number of reasons, including the fact that “there is considerable debate concerning the extent to which ASD and PTSD are similar or different constructs . . .” (p. 353) and “. . . it is useful to determine if [the] proposed factor structure of ASD is compa- rable to the reported factor structures in PTSD” (p. 353).

19FACTOR MODELS OF ASD

Our study tested whether the factor structure of ASD would be better conceptualized in light of empirically supported, preexist- ing, PTSD factor structures. Indeed, many studies have concluded that the underlying dimensionality of PTSD is best represented by a four-factor structure. The large majority of support currently lies with two four-factor models (King et al., 1998; Simms et al., 2002). The King et al. (1998) model is comprised of four factors: intrusion, avoidance, emotional numbing, and arousal. Factor an- alytic support for this model is common across various trauma samples and measurement instruments (for recent support cf. Cox, Moata, Clara, & Asmundson, 2008; DuHamel et al., 2004; McDonald et al., 2008; Palmieri, Marshall, & Schell, 2007). The Simms et al. (2002) model also specified four factors: intrusion, avoidance, dysphoria, and arousal. Support for this model concep- tualization is also common and again covers multiple trauma samples, which use varying measurement instruments (for recent support cf. Boelen, van den Hout, & van den Bout, 2008; Elhai et al., 2009; Elklit et al., 2010; Olff, Sijbrandij, Opmeer, Carlier, & Gersons, 2009). The two models differ with regard to the place- ment of three symptoms: sleeping difficulties, irritability or anger, and concentration difficulties. These are specified to load on the emotional numbing factor in the King model and the dysphoria factor in the Simms model.

ASD and PTSD are conceptually related, diagnostically similar (APA, 1994), and the latent structure of both appears to be best represented by four-factor models (Brooks et al., 2008; King et al., 1998; Simms et al., 2002; Wang et al., 2010). The conceptual similarity between the dissociative items of ASD and the passive- avoidance items of PTSD (Brooks et al., 2008) suggest some hypotheses that concern the factor structure of ASD. The aim of this study was to test the latent structure of ASD symptoms by using data from Danish rape victims who completed the ASDS (Bryant et al., 2000). Studies that address the latent structure of ASD and indeed PTSD are pertinent given that the individual

symptom factors may have differing mechanisms (Foa, Zinbarg, & Rothbaum, 1992). Likewise, individual symptom factors may ev- idence varying degrees of associations with alternative psychiatric disorders (Elklit & Shevlin, 2007). In addition, the development of symptoms within different factors may be differentially related to trauma experiences (Armour & Shevlin, 2010). Thus, knowledge of the latent structure of ASD will prove useful when ascertaining how particular symptoms develop and the relationship between individual symptoms or clusters of symptoms to alternative psy- chiatric disorders or co-occurring difficulties (Scher, McCreary, Asmundson, & Resick, 2008). Using an alternative models ap- proach, this study tested four models. These were a unidimensional ASD model, the DSM–IV model of ASD (as empirically supported by Brooks et al., 2008, and Wang et al., 2010), a four-factor model based on King et al. (1998), and a four-factor model based on Simms et al. (2002). Note that the King and Simms replication models mapped the ASD items onto the preexisting PTSD factor structures while assigning dissociative symptoms to passive- avoidance factors (emotional numbing and dysphoria, respec- tively). See Table 1 for the symptom distributions of the proposed factor models.

Method

Participants

Three hundred and eighty rape victims were contacted through the Centre for Rape Victims (CRV) at the University hospital in Aarhus, Denmark. All participants completed a questionnaire within 4 weeks of experiencing the assault. The questionnaire assessed a variety of demographic variables, a number of peritrau- matic factors, and information related to the type of assault expe- rienced. Sixty-nine percent (n � 263) of the sample met the DSM–IV diagnostic criteria for the four symptom clusters of dis-

Table 1 Model Specifications for the Alternative Models of Acute Stress Disorder (ASD)

ASD scale items One-factor model:

Model 1

DSM-IV four-factor model: Model 2

(Brooks et al., 2008)

King replication four-factor model:

Model 3

Simms replication four-factor model:

Model 4

Three-factor model, combining intrusion and

arousal items of the DSM-IV model: Model 5

Intrusive memories (ASD 6) ASD Intrusion Intrusion Intrusion Intrusion/Arousal Nightmares (ASD 7) ASD Intrusion Intrusion Intrusion Intrusion/Arousal Flashbacks (ASD 8) ASD Intrusion Intrusion Intrusion Intrusion/Arousal Distressed on reminders (ASD 9) ASD Intrusion Intrusion Intrusion Intrusion/Arousal Numbness (ASD 1) ASD Dissociation Numbing Dysphoria Dissociation Dazed (ASD 2) ASD Dissociation Numbing Dysphoria Dissociation Derealization (ASD 3) ASD Dissociation Numbing Dysphoria Dissociation Depersonalization (ASD 4) ASD Dissociation Numbing Dysphoria Dissociation Dissociative amnesia (ASD 5) ASD Dissociation Numbing Dysphoria Dissociation Avoidance of thoughts (ASD 10) ASD Avoidance Avoidance Avoidance Avoidance Avoidance of conversations (ASD 11) ASD Avoidance Avoidance Avoidance Avoidance Avoidance of reminders (ASD 12) ASD Avoidance Avoidance Avoidance Avoidance Avoidance of emotions (ASD 13) ASD Avoidance Avoidance Avoidance Avoidance Insomnia (ASD 14) ASD Arousal Arousal Dysphoria Intrusion/Arousal Irritability (ASD 15) ASD Arousal Arousal Dysphoria Intrusion/Arousal Concentration difficulties (ASD 16) ASD Arousal Arousal Dysphoria Intrusion/Arousal Hypervigilance (ASD 17) ASD Arousal Arousal Arousal Intrusion/Arousal Startle response (ASD 18) ASD Arousal Arousal Arousal Intrusion/Arousal Physiological reactivity (ASD 19) ASD Arousal Intrusion Intrusion Intrusion/Arousal

20 ARMOUR, ELKLIT, AND SHEVLIN

sociation, intrusion, avoidance, and arousal. They had experienced a range of sexual assaults including, completed rape (154, 60.4%), attempted rape (37, 14.5%), and sexual touching (24, 9.4%). Ages ranged from 13 to 57 years (M � 22.91, SD � 8.69). Almost all participants were female (255, 99.6%) with only one male case. The most common residential circumstances were “living with parents” (95, 37.1%) and “living alone” (52, 20.3%). Over half classified their occupation as school–student (128, 50.4%), the second most common occupation was “working” (50, 19.7%), followed by “unemployed” (31, 12.2%).

Measures

Acute Stress Disorder Scale. The Acute Stress Disorder Scale (ASDS; Bryant et al., 2000) is a self-report measure with 19 items that measure dissociative, intrusion, avoidance, and arousal symptoms as specified by the DSM–IV. Items are answered on a 5-point Likert scale, ranging from 1 (not at all) to 5 (very much). The ASDS has been reported as producing reliable scores with reliability coefficients of .85, .90, and .84 (Armour, Shevlin, Elklit, & Mroczek, in press; Elklit & Christiansen, in press; Elklit, Due, & Christiansen, 2009, respectively). Cronbach’s alpha coefficients in this study were .67 for the dissociation factor, .64 for the intrusion factor, .67 for the avoidance factor, .62 for the arousal factor, and .80 for the total ASDS. Bryant et al. (2000) reported high internal consistency, sensitivity, and specificity. For the pur- poses of this analysis, scores on the ASDS items greater than or equal to three were coded as positively endorsed items and thus were used to ascertain those individuals who met the ASD diag- nostic criteria. This procedure has been used previously (cf. Elklit & Christiansen, in press).

Analytic plan. All statistical analyses were conducted with Mplus 5.2 software (Muthén & Muthén, 2007) and were based on those who met the DSM–IV diagnostic criteria for the four symp- tom clusters of dissociation, intrusion, avoidance, and arousal (263, 69.0%). Five competing models were specified (see Table 1). The model parameters were estimated by using robust maximum likelihood (MLR) estimator based on a covariance matrix. For all models, the factors were allowed to correlate. Goodness-of-fit indices were used to assess model fit, including the chi-square, CFI (Bentler, 1990), TLI (Tucker & Lewis, 1973), RMSEA (Steiger,1990), standardized root-mean-square residual (SRMR; Jöreskog & Sörbom, 1993) and the Bayesian information criterion (BIC; Schwarz, 1978). Hoyle and Panter (1995) proposed that acceptable model fit requires a nonsignificant chi-square value. Hu and Bentler (1999) have stated that good model fit is demonstrated by RMSEAs that are less than .06, CFIs and TLIs that are greater than .95, and SRMRs that are less than .08. Adequate model fit is demonstrated by RMSEAs that are less than .08, CFIs and TLIs that are between .90 and .94, and SRMRs that are less than .10. A 10-point BIC difference represents a 150:1 likelihood that the model with the lower BIC value fits best, p � .05 (Raftery, 1995).

Results

Table 2 displays the goodness-of-fit statistics for all models. The chi-square for all models was statistically significant. Schumaker and Lomax (1996) reported that the chi-square value tends to become nonsignificant when assessing model fit with samples thatT

ab le

2 F

it In

d ic

es fo

r th

e F

o u r

A lt

er n a ti

ve M

o d el

s o f

A cu

te S tr

es s

D is

o rd

er (A

S D

)

It em

O ne

-f ac

to r

m od

el :

M od

el 1

D S M

-I V

fo ur

-f ac

to r

m od

el :

M od

el 2

K in

g re

pl ic

at io

n fo

ur -f

ac to

r m

od el

: M

od el

3 S

im m

s re

pl ic

at io

n fo

ur -f

ac to

r m

od el

: M

od el

4

T hr

ee -f

ac to

r m

od el

, co

m bi

ni ng

th e

in tr

us io

n an

d ar

ou sa

l it

em s

of th

e D

S M

-I V

m od

el (B

ro ok

s et

al .,

20 08

)

(M ap

pe d

to th

e fo

ur -f

ac to

r K

in g

et al

., 19

98 ,

P T

S D

m od

el )

(M ap

pe d

to th

e fo

ur -f

ac to

r S

im m

s et

al .,

20 02

, P

T S

D m

od el

)

� 2

37 9.

34 27

6. 12

28 6.

78 27

4. 08

27 8.

74 d f

(p )

15 2

(. 00

) 14

6 (.

00 )

14 7

(. 00

) 14

6 (.

00 )

14 9

(. 00

) R

M S

E A

(9 5%

C I)

.0 75

(. 06

6– .0

85 )

.0 58

(. 04

8– .0

69 )

.0 60

(. 05

0– .0

70 )

.0 58

(. 04

7– .0

68 )

.0 58

(. 04

7– .0

68 )

C F

I .6

7 .8

0 .7

9 .8

1 .8

0 T

L I

.6 1

.7 7

.7 5

.7 7

.7 6

S R

M R

.0 8

.0 7

.0 8

.0 7

.0 7

B IC

15 84

0. 96

7 15

76 4.

28 5

15 76

9. 63

6 15

76 1.

80 8

15 75

1. 14

3

N o te

. D

S M

-I V

� D

ia g n o st

ic a n d

S ta

ti st

ic a l

M a n u a l

o f

M en

ta l

D is

o rd

er s

(4 th

ed .)

; P

T S

D �

po st

tr au

m at

ic st

re ss

di so

rd er

; R

M S

E A

� ro

ot -m

ea n-

sq ua

re er

ro r

of ap

pr ox

im at

io n;

C I

� co

nf id

en ce

in te

rv al

; IF

I �

in cr

em en

ta l

fi t

in de

x; C

F I

� co

m pa

ra ti

ve fi

t in

de x;

S R

M R

� st

an da

rd iz

ed ro

ot -m

ea n

re si

du al

; B

IC �

B ay

es ia

n in

fo rm

at io

n. C

ri te

ri a.

21FACTOR MODELS OF ASD

exceed 200 participants, as in our sample. Therefore, a significant chi-square value should not lead to the rejection of the models. On the basis of the guidelines proposed by Hu and Bentler (1999) all models provided acceptable fit. Indeed, based on the RMSEA and the SRMR alone all models provided adequate to excellent fit. Models 2 and 4 produced the lowest RSMEA values and the lowest SRMR values. All models failed to meet the guidelines with regard to the CFI and TLI indices, indeed the values on these indices failed to come close to conventional guidelines for ade- quate model fit. However, the highest CFI was produced by Model 4, and the highest TLI was produced by Models 2 and 4. Both Models 2 and 4 were evidenced by the lowest BIC values; how- ever, there were minimal differences (15764.285 and 15761.808, respectively). Therefore, Models 2 and 4 were indistinguishable based on a number of fit indices. A closer inspection of the parameter estimates revealed a very high correlation between the intrusion and arousal factors for both Model 2 (r � .91) and Model 4 (r � .81). This was also the case in the DSM–IV ASD models reported in the Brooks et al. (2008) study (r � .91) and the Wang et al. (2010) study (r � .94). Therefore, a supplementary analysis was conducted in which a three-factor model that combines the intrusion and arousal factors of the DSM–IV model was specified and estimated (Model 5). Again the fit indices (RSMEA � .05, CFI � .80, TLI � .76, SRMR � .07) were indistinguishable from those reported by Model 2 and Model 4. However, this was not the case for the BIC value. Indeed, the BIC value for Model 5 (15751.143) represents a 10-point decrease compared to the BIC values of Model 2 (15764.285) and Model 4 (15761.808). On the basis of the guidelines that state that a 10-point difference in BIC values represents a 150:1 likelihood that the model with the lower BIC value fits best ( p � .05) this suggests that the three-factor

model is the best-fitting model. However, as it is recommended that model fit should be judged on a “two index presentation strategy” (Hu & Bentler, 1999) tests of statistical significance between Models 2 and 5 and Models 2 and 4 were assessed with chi-square difference tests (Satorra & Bentler, 2001). Results were nonsignificant for both Model 2 versus Model 5 (��2 � 2.98, �df � 3, p � .39) and Model 4 versus Model 5 (��2 � 4.81, �df � 3, p � .19).

The differences in fit between Models 2, 4, and 5 were not significant. Model 5 is composed of three-factors compared with Model 2 and Model 4, both of which are composed of four factors. Therefore, Model 5 is the preferred model as it evidenced the lowest BIC value by 10 points (Raftery, 1995) and is the more parsimonious model. The adequacy of model fit must also be judged in terms of the parameter estimates. The standardized factor loadings and factor correlations of Model 5 were all positive and statistically significant with one exception (cf. Table 3).

Discussion

This study assessed the fit of four alternative ASD models (cf. Table 1). On the basis of a number of fit indices all models provided acceptable fit to the data. However, poorer fit was pro- vided by the unidimensional model and the King replication model. The remaining models were indistinguishable with regard to model fit. On the basis of a high intercorrelation between intrusion and arousal in both models and a high correlation be- tween these factors in previous studies (Brooks et al., 2008; Wang et al., 2010) questions are raised about the redundancy of these factors. Therefore, supplementary analysis specified and estimated a three-factor model (Model 5) combining the intrusion and

Table 3 Standardized Factor Loadings (SEs) and Factor Correlations for Model 5

ASD scale items Intrusion/Arousal Dissociation Avoidance

Intrusive memories(ASD 6) 0.38 (.06) Nightmares(ASD 7) 0.47 (.07) Flashbacks (ASD 8) 0.53 (.06) Distressed on reminders (ASD 9) 0.31 (.07) Numbness (ASD 1) 0.48 (.09) Dazed (ASD 2) 0.61 (.07) Derealization (ASD 3) 0.45 (.08) Depersonalization (ASD 4) 0.49 (.08) Dissociative amnesia (ASD 5) 0.11 (.08)*

Avoidance of thoughts (ASD 10) 0.63 (.08) Avoidance of conversations (ASD 11) 0.57 (.07) Avoidance of reminders (ASD 12) 0.29 (.09) Avoidance of emotions (ASD 13) 0.50 (.08) Insomnia (ASD 14) .50 (.06) Irritability (ASD 15) .47 (.06) Concentration difficulties (ASD 16) .56 (.06) Hypervigilance (ASD 17) .50 (.06) Startle response (ASD 18) .59 (.06) Physiological reactivity (ASD 19) .59 (.05) Factor correlations Avoidance 0.25 (.10) .30 (.12) NA Dissociation 0.54 (.08) NA NA Intrusion/Arousal NA NA NA

Note. ASD � acute stress disorder. All effects are statistically significant ( p � .05) except those loadings highlighted by an asterisk

22 ARMOUR, ELKLIT, AND SHEVLIN

arousal factors of the DSM–IV four-factor model. Again the model fit was indistinguishable (with the exception of a lower BIC value) from the DSM–IV model and the Simms replication model. Chi- square difference testing concluded that there were no significant differences between the fit provided by the two four-factor models (DSM–IV & Simms replication) compared to the three-factor model. Therefore, the three-factor model was deemed the preferred model on the basis of a lower BIC value and parsimony.

Neither the King or Simms model provided superior fit to the data, despite strong empirical support for these models (cf. Armour & Shevlin, 2010; Elhai et al., 2009; Elklit et al., 2010; Elklit & Shevlin, 2007; Palmieri et al., 2007a & b) combined with the conceptual similarities between ASD and PTSD (Brooks et al., 2008). Brooks et al. (2008) have offered an explanation. They highlighted that although the dissociative symptoms of ASD are conceptually similar to the passive-avoidance symptoms in PTSD there is one notable difference between the two. The dissociative items are reflective of reactions that occurred during the trauma and therefore are considered to be peritraumatic factors, whereas, the passive avoidance items are reflective of reactions that occur after the trauma and therefore are considered to be posttraumatic factors. As such the dissociative items of ASD and the passive avoidance items of PTSD may be qualitatively distinct and there- fore play a different role in the etiology of the two disorders and thus are inappropriately classified within one factor.

Contrary to previous factor analytic studies (Brooks et al., 2008; Wang et al., 2010) the results demonstrated that the DSM–IV model was not the best conceptualisation of the latent structure of ASD. Instead, a model which combined the Intrusion and Arousal factors while retaining the original DSM–IV placement of items across Dissociation and Avoidance was deemed the preferred model. Interestingly, a four-factor model developed through EFA, reported by Bryant et al. (2000) also included an Intrusion/Arousal factor. They stated that “. . . the observed clustering of reexperi- encing [intrusion] and arousal symptoms is consistent with the proposal that acute arousal is strongly related to distress associated with intrusive and distressing memories” (p. 66). In addition, the observed clustering of Intrusion and Arousal may to some extent be explained by fear conditioning models. Such models have shown that extreme arousal during the traumatic experience or in the initial posttrauma period leads to trauma memory consolida- tion. Research by Charney, Deutch, Krystal, Southwick, and Davis (1993) suggested that this may be attributable to increased states of arousal that result in the release of stress neurochemicals, which facilitate traumatic memory consolidation. Pitman, Shalev, and Orr (2000) stated that the overconsolidation of traumatic memories can lead to intrusive recollections, thus illustrating the interdependency between the factors of intrusion and arousal (Nixon & Bryant, 2005). However, it is also important to note this model conceptu- alization was previously specified and estimated by Wang et al. (2010). Wang et al. reported that the three-factor model provided poorer fit compared to the four-factor DSM–IV model. Notably, however the differences in fit indices between these two models in the Wang et al. (2010) study were minimal (e.g., RMSEA � .39 vs. .38; CFI � 99 & TLI � 99). Model superiority was deemed on the basis of a lower Akaike information criteria (AIC; Akaike, 1987).

It is interesting to note that with regard to the parameter esti- mates of the model, the standardized factor loadings and factor

correlations were all statistically significant with the exception of the factor loading for dissociative amnesia. Despite the possibility that the weak loading of this item may be related to the frequency of which this item is endorsed, supplementary analyses revealed that 65% of the ASD diagnosed sample positively endorsed this item. Indeed, this item was more frequently endorsed than items related to nightmares (57.8%), flashbacks (50.6%), and the avoid- ance of conversations (58.6%), all of which loaded significantly on their corresponding factors. The weak loading of this dissociative item was also highlighted by Brooks et al. (2008). They suggested that the operational definition of this item is in need of clarification and that the role of this item within the ASD diagnostic criteria set needs further investigation. This is an issue that has been high- lighted with regard to the weak loading of the memory impairment item within the diagnostic criteria for PTSD: Baschnagel, O’Conner, Colder, & Hawk, 2005 (.35); Elklit & Shevlin, 2007 (.21); Scher, McCreary, Asmundson, & Resnick, 2008 (.18 –.37); Olff et al., 2009 (.39 & .28). Notably, of all the aforementioned studies, the factor loading of the PTSD memory impairment item is consistently the lowest factor loading across all items.

Wang et al. (2010) stated that previous ASD factor analytic studies (e.g., Brooks et al., 2008; Bryant et al., 2000) have reported that the dissociative cluster correlates relatively poorly with the remaining three clusters in the DSM–IV model. However, a closer look reveals that this is an assumption based on the fact that an EFA in the Bryant et al. (2000) study resulted in all items from the intrusion, avoidance, and arousal factors loading on a single factor, whereas the dissociative items loaded on two separate factors. Bryant et al. (2000) therefore reported that this suggested that the intrusion, avoidance, and arousal factor were strongly interrelated. However, they did not report the interfactor correlation values for any given model. They did report on the factor correlations be- tween the ASD and the ASDI; however, the reported correlations between the dissociative factor and other factors ranged from r � .54 to r � .64 and therefore are regarded as large. Likewise, the correlations between the dissociative factor and the three remain- ing factors of the DSM–IV model in the Brooks et al. (2008) study ranged from r � .63 to r � .77. Furthermore, in their own study, Wang et al. (2010) correctly noted strong interfactor correlations: dissociation with intrusion (r � .88), dissociation with avoidance (r � .74), dissociation with arousal (r � .85). Therefore, the results of previous factor analytic studies suggest that the dissociative factor is well placed within the ASD factor analytic structure. Indeed, the correlation coefficients in this study are also all sig- nificant: dissociation with intrusion/arousal, r � .54, and dissoci- ation with avoidance, r � .30. However, the interfactor correla- tions reported herein are lower than those reported previously. Indeed the correlation reported for dissociation and avoidance (r � .30) and the correlation reported for intrusion/arousal with avoid- ance (r � .25) are moderate and small, respectively. Therefore, the results of this study suggest that there is a good degree of orthog- onality between the ASD factors, suggesting that they are statis- tically independent but still subsumed under the ASD construct.

This study also tested a unidimensional model of ASD. The poor fit provided by this model compared to the four alternative models is pertinent in light of the proposed revision to the ASD symptom criteria in the DSM-5 (cf. APA, 2010). The most notable revision relates to the fact that individuals will be required to endorse at least eight ASD symptoms across the subheadings of intrusion,

23FACTOR MODELS OF ASD

dissociation, avoidance, and arousal. The number of items required from each factor is not stated. Indeed, there is a note that states that the DSM-5 task force is considering the removal of topic headings, thus the removal of symptom factors. This unidimensional ap- proach has not received empirical support in this article.

The findings reported in this study must be interpreted in light of several limitations. First, the study used the ASDS (Bryant et al., 2000), a self-report measure, rather than the ASDI (Bryant et al., 1998), which is based on a clinical interview. The PTSD literature has shown that the factor structure of PTSD may vary as a function of the measurement instrument used (Palmieri, Weath- ers, Difede, & King, 2007), thus, as Brooks et al. (2008) used the use of the ASDI this raises questions with regard to the true comparability between results. In addition, this study was con- ducted on Danish trauma victims and therefore the reliability with which the results will generalize is not known. However, given the large number of PTSD factor analytic studies from the United States the fact that this study is based on Danish participants could also be regarded as a strength. Further factor analytic work across varying trauma samples and cultures must be conducted before making any firm conclusions. In addition, and as highlighted by Brooks et al. (2008), although all models provided acceptable fit to the data including the four-factor DSM–IV conceptualization, this must not be taken as support for the use of the ASD diagnosis. Despite these limitations this study used a robust latent variable modeling approach. In addition, the analysis was conducted only on individuals who met the diagnostic requirements of ASD symp- tom factors. This is in contrast to previous studies in which analysis was conducted on the full sample despite only 7.5% of participants meeting ASD diagnostic criteria (Brooks et al., 2008) and 15.3% of participants being classified as probable ASD (Wang et al., 2010). Furthermore, these results are pertinent in light of the proposed revisions to the ASD diagnostic criteria in the DSM-5 (APA, 2010) and as such add to the debate about how ASD will be conceptualized in the near future.

References

Akaike, H. (1987). Factor-analysis and AIC. Psychometrika, in52, 317– 332.

American Psychiatric Association. (1980). Diagnostic and statistical man- ual of mental disorders (3rd ed.). Washington, DC: Author.

American Psychiatric Association. (1994). Diagnostic and statistical man- ual of mental disorders (4th ed.). Washington, DC: Author.

American Psychiatric Association. (2010). DSM-5 development. Retrieved March 11th, 2010, from http://www.dsm5.org/ProposedRevisions/ Pages/proposedrevision.aspx?rid�165#

Armour, C., McBride, O., Shevlin, M., & Adamson, G. (2011). Testing the robustness of the dysphoria factor of the Simms et al. (2002) model of PTSD. Psychological Trauma: Theory, Research, Practice and Policy, 3, 139 –147.

Armour, C., & Shevlin, M. (2010). Testing the dimensionality of PTSD and the specificity of the dysphoria factor. Journal of Loss and Trauma, 15, 11–27.

Armour, C., Shevlin, M., Elklit, A., & Mroczek, D. (in press). A latent growth mixture modelling approach to PTSD symptoms in rape victims. Traumatology.

Baschnagel, J. S., O’Conner, R. M., Colder, C. R., & Hawk, L. W. (2005). Factor structure of posttraumatic stress among Western New York undergraduates following the September 11th terrorist attack on the World Trade Centre. Journal of Traumatic Stress, 18, 677– 684.

Bentler, P. M. (1990). Comparative fit indices in structural models. Psy- chological Bulletin, 107, 238 –246.

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 confirmatory factor analysis. Journal of Anxiety Disor- ders, 22, 1377–1383.

Brewin, C. R., Andrews, A., Rose, S., & Kirk, M. (1999). Acute stress disorder and posttraumatic stress disorder in victims of violent crime. American Journal of Psychiatry, 156, 360 –366.

Brooks, R., Silove, D., Bryant, R., O’Donnell, M., Creamer, M., & McFarlane, A. (2008). A confirmatory factor analysis of acute stress disorder interview. Journal of Traumatic Stress, 21, 352–355.

Bryant, R. A. (2007). Early intervention for post-traumatic stress disorder. Early Intervention in Psychiatry, 1, 19 –26.

Bryant, R. A., & Harvey, A. G. (1998). The relationship between Acute Stress Disorder and Posttraumatic Stress Disorder following mild trau- matic brain injury. American Journal of Psychiatry, 155, 625– 629.

Bryant, R. A., & Harvey, A. G. (2000). Acute stress disorder: A handbook of theory, assessment, and treatment. Washington, DC: American Psy- chological Association.

Bryant, R. A., Harvey, A. G., Dang, S. T., & Sackville, T. (1998). Assessing acute stress disorder: Psychometric properties of a structured clinical interview. Psychological Assessment, 10, 215–220.

Bryant, R. A., Moulds, M. L., & Guthrie, R. M. (2000). Acute Stress Disorder scale: A self-report measure of acute stress disorder. Psycho- logical Assessment, 12, 61– 68.

Bryant, R. A., Moulds, M. L., Guthrie, R. M., & Nixon, R. D. (2005). The Additive Benefit of Hypnosis and Cognitive-Behavioral Therapy in Treating Acute Stress Disorder. Journal of Consulting and Clinical Psychology, 73, 334 –340.

Cardeña, E., Koopman, C., Classen, C., Waelde, L. C., & Spiegel, D. (2000). Psychometric properties of the Stanford Acute Stress Reaction Questionnaire (SASRQ): A valid and reliable measure of acute stress. Journal of Traumatic Stress, 13, 719 –734.

Charney, D. S., Deutch, A. Y., Krystal, J. H., Southwick, S. M., & Davis, M. (1993). Psychobiologic mechanisms of posttraumatic stress disorder. Archives of General Psychiatry, 50, 294 –305.

Cox, B. J., Moata, N., Clara, I., & Asmundson, G. J. G. (2008). The symptom structure of posttraumatic stress disorder in the National Co- morbidity Replication Survey. Journal of Anxiety Disorders, 22, 1523– 1528.

Creamer, M. C., O’Donnell, M. L., & Pattison, P. (2004). The relationship between acute stress disorder and posttraumatic stress disorder in se- verely injured trauma survivors. Behaviour Research and Therapy, 42, 315–328.

DuHamel, K. N., Ostroff, J., Ashman, T., Winkel, G., Mundy, E. A., Keane, T. M., . . . Burkhalter, J. (2004). Construct validity of the Posttraumatic Stress Disorder Checklist in cancer survivors: Analyses based on two samples. Psychological Assessment, 16, 255–266.

Elhai, J. D., Ford, J. D., Ruggiero, K. J., & Frueh, C. B. (2009). Diagnostic alterations for post-traumatic stress disorder: Examining data from the National Comorbidity Survey Replication and National Survey of Ad- olescents. Psychological Medicine, 39, 1957–1966.

Elklit, A., Armour, C., & Shevlin, M. (2010). Testing alternative factor models of PTSD and the robustness of the dysphoria factor. Journal of Anxiety Disorder, 24, 147–154.

Elklit, A., & Brink, O. (2004). Acute stress disorder as a predictor of post-traumatic stress disorder in physical assault victims. Journal of Interpersonal Violence, 19, 709 –716.

Elklit, A., & Christiansen, D. M. (in press). ASD and PTSD in rape victims. Journal of Interpersonal Violence.

Elklit, A., Due, L., & Christiansen, D. M. (2009). Predictors of acute stress symptoms in rape victims. Traumatology, 15, 3– 45.

Elklit, A., & Shevlin, M. (2007). The structure of PTSD symptoms: A test

24 ARMOUR, ELKLIT, AND SHEVLIN

of Alternative models using confirmatory factor analysis. British Journal of Clinical Psychology, 46, 299 –313.

Foa, E. B., Zinbarg, R., & Rothbaum, B. O. (1992). Uncontrollability and unpredictability in post-traumatic stress disorder: An animal model. Psychological Bulletin, 112, 218 –238.

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, England: Sage.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struc- tural Equation Modeling, 6, 1–55.

Jöreskog, K., & Sörbom, D. (1993). LISREL 8 User’s reference guide. Chicago, IL: Scientific Software International.

Kangas, M., Henry, J. L., & Bryant, R. A. (2005). The relationship between acute stress disorder and posttraumatic stress disorder following cancer. Journal of Consulting and Clinical Psychology, 73, 360 –364.

Kassam-Adams, N., Fleisher, C. L., & Winston, F. K. (2009). Acute stress disorder and posttraumatic stress disorder in parents of injured children. Journal of Traumatic Stress, 22, 294 –302.

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 dimensionality of posttraumatic stress disorder. Psy- chological Assessment, 10, 90 –96.

McDonald, S. D., Beckham, J. C., Morey, R., Marx, C., Tupler, L. A., & Calhoun, P. S. (2008). Factorial invariance of posttraumatic stress dis- order symptoms across three veteran samples. Journal of Traumatic Stress, 21, 309 –317.

Muthén, B. O., & Muthén, L. K. (2007). Mplus (Version 5). Los Angeles, CA: Muthén & Muthén.

Nixon, R. D. V., & Bryant, R. A. (2005). Induced arousal and reexperi- encing in acute stress disorder. Journal of Anxiety Disorders, 19, 587– 594.

Olff, M., Sijbrandij, M., Opmeer, B. C., Carlier, I. V. E., & Gersons, B. P. R. (2009). The structure of acute posttraumatic stress symptoms: ‘reexperiencing’, ‘active avoidance’, ‘dysphoria’, and ‘hyperarousal.’ Journal of Anxiety Disorders, 23, 656 – 659.

Palmieri, P. A., Marshall, G. N., & Schell, T. L. (2007a). 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. (2007b). Confirmatory factor analysis of the PTSD Checklist and the Clinician Administered PTSD Scale in Disaster workers exposed to the World

Trade Centre Ground Zero. Journal of Abnormal Psychology, 116, 329 –341.

Pitman, R. K., Shalev, A. Y., & Orr, S. P. (2000). Posttraumatic stress disorder: Emotion, conditioning and memory. In M. D. Corbetta & M. Gazzaniga (Eds.), The new cognitive neurosciences (2nd ed.). New York, NY: Plenum Press.

Raftery, A. E. (1995). Bayesian model selection in social research. Soci- ological Methodology, 25, 111–163.

Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66, 507–514.

Scher, C. D., McCreary, D. R., Asmundson, G. J. G., & Resick, P. A. (2008). The structure of post-traumatic stress disorder symptoms in three female trauma samples: A comparison of interview and self-report measures. Journal of Anxiety Disorders, 22, 1137–1145.

Scher, C. D., McCreary, D. R., Asmundson, G. J. G., & Resnick, H. S. (2008). Survey replication and National Survey of Adolescents. Psycho- logical Medicine, 39, 1957–1966.

Schnyder, U., Moergeli, H., Klaghofer, R., & Buddeberg, C. (2001). Incidence and prediction of posttraumatic stress disorder symptoms in severely injured accident victims. American Journal of Psychiatry, 158, 594 –599.

Schumaker, R. E., & Lomax, R. G. (1996). A beginner’s guide to structural equation modeling. Mahwah, NJ: Erlbaum.

Schwarz, G. E. (1978). Estimating the dimension of a model. Annals of Statistics, 6, 461– 464.

Simms, L. J., Watson, D., & Doebbeling, B. N. (2002). Confirmatory factor analyses of posttraumatic stress symptoms in deployed and non- deployed veterans of the Gulf War. Journal of Abnormal Psychology, 111, 637– 647.

Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation approach. Multivariate Behavioral Research, 25, 173–180.

Tucker, C., & Lewis, C. (1973). A reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38, 1–10.

Wang, L., Li, Z., Shi, Z., Zhang, S., & Shen, J. (2010). Factor structure of acute stress disorder symptoms in Chinese earthquake victims: A con- firmatory factor analysis of the acute stress disorder scale. Personality and Individual Differences, 48, 798 – 802.

Received March 19, 2010 Revision received October 7, 2010

Accepted November 27, 2010 �

25FACTOR MODELS OF ASD