Behaviour Research and Therapy 38 (2000) 203±214

www.elsevier.com/locate/brat

Dimensionality of posttraumatic stress symptoms: a con®rmatory factor analysis of DSM-IV symptom clusters and other symptom models Gordon J.G. Asmundson a, b,*, Inger Frombach a, b, John McQuaid c, Paolo Pedrelli c, Rebecca Lenox c, Murray B. Stein c a

b

Regina Health District, Regina, Saskatchewan, Canada Department of Psychology, University of Regina, Regina, Saskatchewan, Canada c San Diego VA Health Care System, San Diego, California, USA Received 13 October 1998; received in revised form 29 January 1999

Abstract Recent exploratory [Taylor, S., Kuch, K., Koch, W. J., Crockett, D. J., & Passey, G. (1998). The structure of posttraumatic stress symptons. Journal of Abnormal Psychology, 107, 154±160.] and con®rmatory [Buckley, T. C., Blanchard, E. B., & Hickling, E. J. (1998). A con®rmatory factor analysis of posttraumatic stress symptons. Behaviour Research and Therapy, 36, 1091±1099; King, D. W., Leskin, G. A., King, L. A., & Weathers, F. W. (1998). Con®rmatory factor analysis of the clinicianadministered PTSD scale: evidence for the dimensionality of posttraumatic stress disorder. Psychological Assessment, 10, 90±96.] factor analytic investigations suggest that the three symptom clusters of posttraumatic stress disorder (PTSD) as de®ned in the Diagnostic and Statistical Manual [4th ed.; DSMIV; American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.] may not provide the best conceptualization of symptom dimensionality. However, the alternative models have not been in agreement, nor have they been compared against each other or models based on the DSM-IV. The purpose of the present investigation was to test a series of dimensional models suggested by these recent factor analytic investigations and the DSM-IV. Using data collected with the PTSD ChecklistÐCivilian Version [Weathers, F. W., Litz, B. T., Huska, J. A., & Keane, T. M. (1994). PCL-C for DSM-IV. Boston: National Center for PTSDÐBehavioral Science Division.] from 349 referrals to a primary care medical clinic, we used con®rmatory factor analysis to evaluate a: (1) hierarchical four-factor model, (2) four-factor intercorrelated model, (3) hierarchical

* Corresponding author. Clinical Research and Development Program, Regina Health District, 2180 23rd Avenue, Regina, Saskatchewan, S4S 0A5, Canada. Tel.: +1-306-766-5384; fax: +1-306-766-5530. E-mail address: [email protected] (G.J.G. Asmundson) 0005-7967/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 0 6 1 - 3

204

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

three-factor model, (4) three-factor intercorrelated model, and (5) hierarchical two-factor model. The hierarchical four-factor model (comprising four ®rst-order factors corresponding to reexperiencing, avoidance, numbing, and hyperarousal all subsumed by a higher-order general factor) provided the best overall ®t to the data; although, all models met some standards speci®ed for good model ®t. More research is needed to establish the dimensional nature of PTSD symptoms and to assess whether identi®ed dimensions di€er as a function of the trauma experience. Implications for assessment, diagnosis, and treatment are also discussed. # 2000 Elsevier Science Ltd. All rights reserved.

1. Introduction Posttraumatic stress disorder (PTSD) is classi®ed in the Diagnostic and Statistical Manual (4th ed.; DSM-IV; American Psychiatric Association, 1994) as an anxiety disorder characterized by three distinct clusters of symptoms that develop following exposure to an event perceived to be traumatic. The symptom clusters include reexperiencing of the traumatic event (criterion B: e.g., nightmares and other intrusive recollections), avoidance of cues associated with the event along with general emotional numbing; (criterion C: e.g., avoidance of people and/or places related to the traumatic event, restricted a€ect), and hyperarousal (criterion D: e.g., sleep diculties, exaggerated startle). In order to receive a DSM-IV de®ned diagnosis of PTSD, an individual must be exposed to a traumatic event with actual or perceived threat, the response must involve intense fear and/or helplessness, and the individual must experience at least one re-experiencing symptom, at least three avoidance and numbing symptoms, and at least two hyperarousal symptoms. Existing PTSD symptom clusters (i.e., criteria B, C and D) were determined by expert consensus. However, on the basis of evidence attained from recent factor analytic investigations of symptom ratings (e.g., Buckley, Blanchard & Hickling, 1998; King, Leskin, King & Weathers, 1998; Taylor, Kuch, Koch, Crockett & Passey, 1998), there is some question as to whether the existing PTSD symptom clusters are representative of the basic underlying dimensions (or factors) of PTSD. This suggests that empirically derived dimensions of PTSD may be more appropriate and would serve e€orts of clinicians and clinical researchers to better understand, diagnose and treat this complex disorder. Taylor et al. (1998) have conducted the most methodologically sound exploratory factor analysis of PTSD symptoms to date. Three previous exploratory factor analytic investigations of PTSD symptoms as presented in the DSM-III-R and DSM-IV have been conducted (i.e., Foa, Riggs & Gershuny, 1995; Keane, 1993; King, King, Leskin & Foy, 1995); but, these investigations were criticized by Taylor et al. (1998) on the basis of sample size limitations and/ or use of data analysis techniques that prevented identi®cation of stable factor structures. Taylor et al. (1998) used principal axis factor analysis with oblique rotation and employed parallel analysis (i.e., a statistical method for determining the break in the scree; Zwick & Velicer, 1986) to examine PTSD symptom data collected from 103 motor vehicle accident (MVA) victims and 419 United Nations peace keepers exposed to combat-related events in Bosnia. Symptom information was collected by structured clinical interview in the MVA victims and using the PTSD Symptom Scale (Foa, Riggs, Dancu & Rothbaum, 1993) in the peace keepers. Results indicated support for a two-factor solution comprising intrusions and

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

205

avoidance (i.e., all ®ve symptoms from criterion B and the two avoidance symptoms from criterion C) and hyperarousal and numbing (i.e., most of the symptoms from criterion D and the ®ve numbing symptoms of criterion C). These two factors possessed good simple structure, loaded on a higher-order factor of PTSD, and were reasonably stable across the MVA and combat groups. On the basis of these results, Taylor et al. (1998) concluded that PTSD may be a function of a general factor that gives rise to most PTSD symptoms and two speci®c factors that contribute to intrusions/avoidance and hyperarousal/numbing. Buckley et al. (1998) have recently con®rmed the hierarchical two-factor model proposed by Taylor et al. (1998). Symptom data collected from 217 MVA victims using the Clinician Administered PTSD Scale (CAPS; Blake et al., 1997) was submitted to con®rmatory factor analysis (CFA). Results indicated that the Taylor et al. (1998) model provided a reasonable to good ®t to the data (e.g., goodness of ®t index [GFI]=0.87, root mean square error of approximation [RMSEA]=0.078, comparative ®t index [CFI]=0.91). These results re¯ect favorably upon the hierarchical two-factor model and further suggest that PTSD may be best conceptualized as comprising two speci®c factorsÐintrusions/avoidance and hyperarousal/ numbingÐeach subsumed by a higher-order general factor. However, these con®rmatory results are limited in that they were based on one of the same trauma groups (i.e., MVA victims) assessed by Taylor et al. (1998). Moreover, alternative models, such as indicated in the DSM-IV, were not tested against the hierarchical two-factor model. Relative to these limitations, King et al. (1998) have used CFA to compare a series of nested models re¯ecting alternative conceptualizations of PTSD symptom dimensionality. They used CAPS derived symptom data collected from 524 treatment seeking military veterans to test: (1) a four-factor intercorrelated model (i.e., reexperiencing, avoidance, numbing, and hyperarousal); (2) a two-factor hierarchical model (i.e., second-order reexperiencing/avoidance and numbing/hyperarousal not subsumed under a general PTSD factor); (3) a hierarchical four-factor model (i.e., reexperiencing, avoidance, numbing, and hyperarousal subsumed under a general PTSD factor), and (4) a single factor or unidimensional model (i.e., all 17 symptoms loading on a general PTSD factor). They did not test the Taylor et al. (1998) model and, given the temporal proximity of the investigations, may have been unaware of it. King et al. (1998) interpreted their results as indicating that the four-factor intercorrelated model provided the best ®t to the data (e.g., GFI=0.95; RMSEA=0.044; CFI=0.95) and that this model was signi®cantly superior to the other models tested. Thus, contrary to the ®ndings of Taylor et al. (1998) and Buckley et al. (1998), King et al. (1998) suggest four distinct yet intercorrelated ®rst-order factors corresponding to re-experiencing, avoidance, emotional numbing, and hyperarousal that are not manifest as part of a higher-order general PTSD factor. It should be noted, however, that interpretation of this solution as superior to the other models was based primarily on the chi-square di€erence. As the obtained RMSEA values were only marginally di€erent from one another, it is tenable that both the two-factor hierarchical and hierarchical four-factor solutions provide a similarly adequate account of the data. The results of the two CFAs reviewed above provide a preliminary indication that the symptom clusters presented in the DSM-IV (i.e., criteria B, C and D) may not provide the best conceptualization of PTSD symptoms. To this end, both Buckley et al. (1998) and King et al. (1998) provide recommendations, albeit ones that di€er as a function of the observed factor solution, for how assessment and treatment may be better focused. However, it might be

206

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

Fig. 1. Simpli®ed factor models tested using con®rmatory factor analysis; (1) hierarchical four-factor model; (2) four-factor intercorrelated model; (3) hierarchical three-factor model; (4) three-factor intercorrelated model; and (5) hierarchical two-factor model. PTSD=posttraumatic stress disorder.

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

207

argued that these recommendations are premature given that neither Buckley et al. (1998) nor King et al. (1998) actually tested their preferred model against models based on the three symptom clusters proposed in the DSM-IV. Moreover, without direct comparison of the hierarchical two-factor model (Buckley et al., 1998; Taylor et al., 1998) and the four-factor intercorrelated model preferred by King et al. (1998), it is dicult to determine which model, if either, provides the best conceptualization of PTSD symptom dimensionality. The purpose of the present investigation was to empirically evaluate the aforementioned models against each other and relative to models suggested by the DSM-IV symptom clusters. Cattell (1978) has argued that each factor identi®ed through factor analysis represents a distinct causal mechanism. To the extent that this is true, the con®rmation of the factor structure of PTSD symptoms is not merely a psychometric exercise but, rather, o€ers to inform e€orts to better understand the nature of the disorder itself. The following models, presented graphically in Fig. 1, were speci®ed and tested: (1) the King et al. (1998) hierarchical fourfactor model; (2) the King et al. (1998) four-factor intercorrelated ®rst-order model; (3) a hierarchical three-factor model stemming from the DSM-IV; (4) a three-factor intercorrelated ®rst-order model derived from the DSM-IV; and (5) the Taylor et al. (1998) hierarchical twofactor model. Because other investigators have focused on speci®c types of trauma (e.g., MVA, combat-related), we chose to evaluate symptoms in a primary care medical setting. It was our intention, by so doing, to determine which factor structure provides the best representation of symptoms reported across patients experiencing symptoms related to a wide range of traumatic experiences (see recommendations of Taylor et al., 1998).

2. Method 2.1. Participants Participants were 349 consecutively approached patients with routine medical problems presenting at the University of California, San Diego Primary Care Clinic at Mira Mesa, who gave informed and written consent to participate (response rate was approximately 60%). All participants were asked to complete the PTSD ChecklistÐCivilian Version (PCL-C; Weathers, Litz, Huska & Keane, 1994), while in the waiting room prior to their appointment, as part of a study of trauma in the general population. The average age of participants was 38.9 (SD=15.3). Approximately 58% of the participants were women. The majority of respondents were Caucasian (60.9%), 12% were Filipino, 8% were Latin American, 8% were African American, 4% were Asian American and 1.7% were native American. Approximately 13% of the participants met the diagnostic criteria for full (n = 13) or partial (i.e., meeting criteria A and B but falling one symptom short on criteria C and/or D; n = 33) PTSD. 2.2. Measures PTSD symptoms were assessed using the PCL-C (Weathers et al., 1994). The 17 items on this self-report measure correspond to the symptoms associated with the DSM-IV diagnostic criteria for PTSD. Respondents indicate, on a scale anchored from 1 (not at all ) to 5

208

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

(extremely ), the degree to which they have been bothered by particular symptoms stemming from potentially stressful life experiences (unspeci®ed) occurring over the past month. Test± retest reliability for the PCL-C (retest interval of 2±3 days; F. W. Weathers, personal communication, January 27, 1999) has been reported at 0.96 (Weathers, Litz, Herman, Huska & Keane, 1993) and the overall diagnostic eciency has been found to be acceptably high at 0.90 (Blanchard, Jones-Alexander, Buckley & Forneris, 1996).

3. Results Descriptive statistics for the PCL-C items and total score are presented in Table 1. Distribution of individual item scores did not vary signi®cantly from normal. CFA procedures were conducted using AMOS (Analysis of Moment Structures; Arbuckle, 1997a,b). Covariance matrices of the PCL-C items were analyzed using maximum likelihood estimation procedures. As recommended by Bollen and Long (1993), multiple ®t indices were used to judge how well the proposed models ®t the data. An e€ort was made to choose indices that would allow comparison with existing literature in the area (i.e., Buckley et al., 1998; King et al., 1998) or that are commonly used in CFA. We examined the goodness of ®t index (GFI) and the adjusted goodness of ®t index (AGFI), requiring values of >0.85 and >0.80, respectively, to denote good ®t (Marsh, Balla & McDonald, 1988). The GFI and AGFI have been criticized as providing insucient evidence of model ®t (MacCallum & Hong, 1997; Floyd & Widaman, Table 1 Descriptive statistics for the PTSD ChecklistÐCivilian Versiona Item

Content

Mean

SD

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Intrusive recollections Distressing dreams Flashbacks Distress at reminders Reaction to reminders Avoid thoughts/feelings Avoid reminders Psychogenic amnesia Anhedonia Detached from others Psychic numbing Foreshortened future Disturbed sleep Irritability/anger Concentration diculty Hypervigilance Exaggerated startle Total score

1.94 1.55 1.48 1.85 1.46 1.78 1.65 1.50 1.85 1.96 1.71 1.71 2.18 2.14 2.06 1.69 1.64 30.14

1.09 0.91 0.89 1.04 0.81 1.08 1.03 0.94 1.15 1.17 1.06 1.07 1.30 1.12 1.10 1.01 0.97 12.42

a

Possible (and observed) item scores range from 1=not at all to 5=extremely.

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

209

1995), but have been commonly reported. The root mean square error of approximation (RMSEA) and standardized root mean square residual (RMR), an index that provides an estimate of the magnitude of the average absolute discrepancy between predicted and observed correlations for each model, were also included as measures of model ®t. Here, values less than 0.05 represent good ®t, values of 0.05±0.08 represent moderate ®t, and values of 0.08 to 0.10 represent adequate ®t (Brown & Cudeck, 1993). Finally, two relative ®t indicesÐthe TuckerLewis Index (TLI) and comparative ®t index (CFI)Ðwere used with values of >0.90 considered to be consistent with good model ®t (Bentler, 1990; Stevens, 1996). Chi-square values were used for purposes of model comparison. The data indicated that the hierarchical four-factor model (labeled in Fig. 1 as `Model 1') was the only solution that met all of the standards of good ®t or, in the case of RMSEA and RMR, moderately good ®t (see Table 2). Further, the chi-square di€erence test revealed that this model was signi®cantly better than any of the competing models. Speci®cally, the hierarchical four-factor model o€ered a better ®t to the data than the Taylor et al. (1998) hierarchical two-factor model (w 2 di€erence=130.69, df = 2, P < 0.001), the three-factor intercorrelated model (w 2 di€erence=142.55, df = 5, P < 0.001), the hierarchical three-factor model (w 2 di€erence=95.38, df = 4, p < .001), and the King et al. (1998) four-factor intercorrelated model (w 2 di€erence=70.52, df = 5, P < 0.001). It is noteworthy, however, that the relative di€erences between various models on some of the ®t indices were marginal and that all of the models tested met some of the standards outlined for good model ®t. Based on chi-square data, scores on the various ®t indices, and consistency with theoretical and clinical presentations (see Foa et al., 1995), we identi®ed the hierarchical four-factor model as providing the best account of the data. The factor loadings of the hierarchical four-factor model are presented in Table 3. With the exception of PCL-C item 13, corresponding to trouble falling or staying asleep, all of the individual PCL-C items had standardized parameter estimates greater than 0.60 on their associated factor. These parameter estimates further arm the appropriateness and strength of this model (Chin, 1998). Table 2 Goodness-of-®t indices for the PTSD models (n = 349)a Fit indices Model

w2

df

GFI

AGFI

RMSEA

RMR

TLI

CFI

Hierarchical two-factor Three-factor Hierarchical three-factor Four-factor Hierarchical four-factor

522.90 534.76 487.60 462.73 392.21

116 119 115 119 114

0.85 0.83 0.84 0.86 0.88

0.80 0.78 0.80 0.82 0.84

0.10 0.10 0.10 0.09 0.08

0.07 0.34 0.07 0.46 0.07

0.85 0.85 0.86 0.88 0.90

0.87 0.87 0.88 0.89 0.91

a

GFI, goodness-of-®t index; AGFI, adjusted goodness-of-®t index; RMSEA, root mean square error of approximation; RMR, standardized root mean residual; TLI, Tucker Lewis index; CFI, comparative ®t index. The degrees of freedom presented for the hierarchical models are minus the error terms for the second order latent factors. Covariance matrices used in the analyses are available upon request. Fit indices identi®ed with an asterisk () approximate the percentage of variance accounted for by the model (Hu & Bentler, 1995).

210

Maximum likelihood estimates Item/DSM-IV symptom

Factor 1, intrusion/reexperiencing

B1. Intrusive recollections B2. Distressing dreams B3. Flashbacks B4. Distress at reminders B5. Reaction to reminders C1. Avoid thoughts/feelings C2. Avoid reminders C3. Psychogenic amnesia C4. Anhedonia C5. Detached from others C6. Psychic numbing C7. Foreshortened future D1. Disturbed sleep D2. Irritability/anger D3. Concentration diculty D4. Hypervigilance D5. Exaggerated startle

0.76 0.81 0.64 0.77 0.71

a

All parameter estimates are standardized.

Factor 2, avoidance

0.79 0.85

Factor 3, numbing

0.67 0.80 0.83 0.74 0.67

Factor 4, hyperarousal

0.55 0.76 0.76 0.71 0.69

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

Table 3 Factor loadings for the hierarchical four-factor modela

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

211

4. Discussion Recent exploratory (Taylor et al., 1998) and con®rmatory (Buckley et al., 1998; King et al., 1998) factor analytic investigations have been conducted, in an e€ort to better understand the dimensionality of PTSD symptoms. These investigations have led to suggestions of alternatives to the symptom dimensions (or clusters) of PTSD presented in the DSM-IV. The suggested models, di€erent not only from the DSM-IV but from each other, are particularly salient given that existing PTSD symptom clusters were determined by expert consensus rather than empirical criteria. Notwithstanding, these models had not previously been compared against each other nor to the DSM-IV symptoms clusters. The present investigation was designed with the intent of addressing this limitation in the existing literature and to thereby make progress toward identifying the most suitable conceptualization of PTSD symptoms from among the existing theoretical and empirical models. Our ®ndings from referrals to a primary care medical clinic indicated that, of the ®ve models tested, the hierarchical four-factor model ®t the data best. Speci®cally, this model comprised four ®rst-order factors corresponding to reexperiencing (criterion B items), avoidance (two criterion C items), numbing (®ve criterion C items), and hyperarousal (criterion D items) all of which were subsumed as dimensions of a second-order general factor. These results extend existing evidence by indicating that the hierarchical four-factor model is empirically superior (i.e., provides a better ®t to the data) to the symptom clusters de®ned in the DSM-IV. It is, however, noteworthy that all of the models tested met some criteria speci®ed for good model ®t (as illustrated in Table 2). Determining which model provides the best overall ®t is a complex process involving not only consideration of multiple ®t indices and parameter estimates but also issues of parsimony and theoretical and clinical appropriateness (Hu & Bentler, 1995; Stevens, 1996). Our endorsement of the hierarchical four-factor model was based on its superiority as established by the chi-square di€erence and assessed ®t indices, its consistency with theoretical conceptualizations of PTSD (Foa, Zinbard & Rothbaum, 1992), and its potential clinical impact (as described below). While future investigation may lead to consensus on an empirically derived model of PTSD symptom clusters, existing theoretical and empirical models possess some degree of merit (at least as determined by CFA). How does our hierarchical four-factor model compare to the hierarchical two-factor model proposed by Taylor et al. (1998) and the four-factor intercorrelated model preferred by King et al. (1998)? There are some obvious di€erences and some conceptual similarities. Relative to the Taylor et al. (1998) model, the di€erence pertains to the number of ®rst-order factors, with the two speci®c factors of intrusions/avoidance and hyperarousal/numbing breaking into four distinct factors of intrusions, avoidance, numbing, and hyperarousal. These same ®rst-order factors were present in the King et al. (1998) model; however, their preferred model did not include a second-order general factor for PTSD. (It should be noted that both their two-factor hierarchical model and hierarchical four-factor model, although not the models of choice, did have good ®t indices.) Consistent with King et al. (1998), our results suggest that PTSD symptoms are best conceptualized along four speci®c dimensions. Taylor et al. (1998) hinted at the possibility of more than two ®rst-order factors but noted, in the case of exploratory factor analysis, that more items assessing each of the symptom dimensions would be required to allow a greater number of factors to emerge. It is possible, had they tested a four-factor model,

212

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

that Buckley et al. (1998) would have found it to have better ®t indices relative to the Taylor et al. (1998) model. Does the hierarchical four-factor model correspond with current theoretical models of fear structures and PTSD? Emerging models suggest that fear and anxiety are hierarchical in nature (Taylor, 1998; Zinbarg & Barlow, 1998). Although not speci®c to PTSD, these models provide a relevant framework within which to conceptualize a hierarchically structured response to traumata. Likewise, evolving conceptualizations of PTSD posit a general posttraumatic stress response which may comprise a variety of distinct symptom presentations that are dependent on the individual and the trauma experience. So, in a general theoretical sense, our hierarchical four-factor model may be preferable to the ®rst-order intercorrelated four-factor model (King et al., 1998) and the more parsimonious but less speci®c two-factor models (Buckley et al., 1998; King et al., 1998; Taylor et al., 1998). The hierarchical four-factor model also corresponds to the proposed theoretical mechanisms that underlie the development of PTSD. For example, Foa et al. (1992) have summarized evidence of the importance of reexperiencing (intrusions) and hyperarousal to posttraumatic stress responses. They propose that avoidance and numbing of general responsiveness involve two separate mechanisms, the former regulated by strategic (or e€ortful) mechanisms and the latter by more automatic mechanisms. Accordingly, avoidance serves as a strategic means of escaping exposure to trauma-related stimuli that may intrude on awareness while numbing serves as an automatic response to hyperarousal. Consistent with Foa et al. (1992), our results suggest speci®c factors (or mechanisms) for reexperiencing, avoidance, numbing, and hyperarousal, and that these are integral parts of the general syndrome known as PTSD. More speci®cally, and echoing Foa et al. (1995), it may be prudent to disjoin criterion C into separate symptom dimensions for arousal and numbing. This may, in turn, necessitate an increased item pool for assessing trauma-related avoidance (Taylor et al., 1998). The divergent models that have been identi®ed by factor analytic investigations may di€er from one another, in part, as a function of the unique aspects of the di€erent trauma groups studied (e.g., MVA victims, military veterans, sexual assault victims) and, possibly, as a result of di€erences in the various tools used to evaluate symptoms. These issues need to be addressed in future investigations. To the extent that identi®ed factors represent underlying causal mechanisms (Cattell, 1978), the model deemed to be of choice will ultimately hold important implications with regard to assessment, diagnosis, and treatment. Both our hierarchical four-factor model and the preferred model of King et al. (1998) indicate that each of four speci®c dimensions of PTSD should be emphasized in assessment. Separation of e€ortful/active avoidance from numbing implies that diagnostic criteria should be altered to allow for di€erent patterns of symptom presentation (King et al., 1998) or, in a more restrictive manner, so that symptoms of avoidance and symptoms of numbing are both required (Foa et al., 1995). In either case, this would acknowledge the distinctiveness of e€ortful/active avoidance and numbing and, possibly, would permit identi®cation of PTSD subtypes. Related to this point, the mechanisms underlying both avoidance and numbing may require di€erent approaches to treatment (i.e., treatment should be tailored to alleviate speci®c symptom presentations). In order to facilitate research and clinical advancements in the area of PTSD, it is essential that an empirically sound model of PTSD symptom dimensionality be identi®ed. This model

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

213

also needs to be clinically relevant and appropriate to di€erent and diverse trauma populations. Ultimately, further empirical investigation of the issues raised above will be required before any ®rm conclusions regarding PTSD symptom dimensionality and the appropriateness of existing DSM-IV symptom clusters and diagnostic criteria can be drawn.

References American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Arbuckle J. L. (1997a). AMOS-Version 3.6. Chicago: Smallwaters Corporation. Arbuckle J. L. (1997b). Amos Users' Guide Version 3.6. Chicago: Smallwaters Corporation. Bentler, P. M. (1990). Comparative ®t indices in structural models. Psychological Bulletin, 107, 238±246. Blake D., Weathers F., Nagy L., Kaloupek D., Klauminzer G., Charney D., Keane T. 1997 Clinician-Administered PTSD Scale (CAPS). Boston: National Center for Post-Traumatic Stress Disorder, Behavioural Science Division, Boston, VA. Blanchard, E. B., Jones-Alexander, J., Buckley, T. C., & Forneris, C. A. (1996). Psychometric properties of the PTSD Checklist (PCL). Behaviour Research and Therapy, 34, 669±673. Bollen, K. A., & Long, J. S. (1993). Testing structural equation models. Newbury Park, CA: Sage. Brown, M. W., & Cudeck, R. (1993). Alternative ways of assessing model ®t. In K. A. Bollen, & J. S. Long, Testing structural equation models (pp. 136±162). Newbury Park, CA: Sage. Buckley, T. C., Blanchard, E. B., & Hickling, E. J. (1998). A con®rmatory factor analysis of posttraumatic stress symptoms. Behaviour Research and Therapy, 36, 1091±1099. Cattell, R. B. (1978). The scienti®c use of factor analysis in the behavioral and life sciences. New York: Graywind. Chin, W. W. (1998). Issues in structural equation modeling. MIS Quarterly, 22, vii±xvi. Floyd, F., & Widaman, K. F. (1995). Factor analysis in the development and re®nement of clinical assessment instruments. Psychological Assessment, 7, 286±299. Foa, E. C., Zinbarg, R., & Rothbaum, B. O. (1992). Uncontrollability and unpredictability in post-traumatic stress disorder: An animal model. Psychological Bulletin, 112, 218±238. Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993). Reliability and validity of a brief instrument for assessing post-traumatic stress disorder. Journal of Traumatic Stress, 6, 459±473. Foa, E. B., Riggs, D. S., & Gershuny, B. S. (1995). Arousal, numbing, and intrusion: symptom structure of PTSD following assault. American Journal of Psychiatry, 152, 116±120. Hu, L., & Bentler, P. M. (1995). Evaluating model ®t. In R. H. Hoyle, Structural equation modeling (pp. 76±99). Thousand Oaks, CA: Sage. Keane, T. M. (1993). Symptomatology of Vietnam veterans with post-traumatic stress disorder. In J. R. T. Davidson, & E. B. Foa, Posttraumatic stress disorder: DSM-IV and beyond (pp. 99±111). Washington, DC: American Psychiatric Press. King, L. A., King, D. W., Leskin, G., & Foy, D. W. (1995). The Los Angeles Symptom Checklist: a self-report measure of posttraumatic stress disorder. Assessment, 2, 1±17. King, D. W., Leskin, G. A., King, L. A., & Weathers, F. W. (1998). Con®rmatory factor analysis of the clinicianadministered PTSD Scale: evidence for the dimensionality of posttraumatic stress disorder. Psychological Assessment, 10, 90±96. MacCallum, R. C., & Hong, S. (1997). Power analysis in covariance structure modeling using GFI and AGFI. Multivariate Behavioral Research, 32, 193±210. Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness-of-®t indexes in con®rmatory factor analysis: the e€ect of sample size. Psychological Bulletin, 103, 391±410. Stevens, J. (1996). Applied multivariate statistics for the social sciences. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

214

G.J.G. Asmundson et al. / Behaviour Research and Therapy 38 (2000) 203±214

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. Taylor, S. (1998). The hierarchic structure of fears. Behaviour Research and Therapy, 36, 205±214. Weathers F. W., Litz B. T., Herman D. S., Huska J. A., & Keane T. M. (1993). The PTSD Checklist: reliability, validity, & diagnostic utility. Paper presented at the Annual Meeting of the International Society for Traumatic Stress Studies, San Antonio, TX, October. Weathers F. W., Litz B. T., Huska J. A., & Keane T. M. (1994). PCL-C for DSM-IV. Boston: National Center for PTSDÐBehavioral Science Division. Zinbarg, R. E., & Barlow, D. H. (1998). Structure of anxiety and the anxiety disorders. Journal of Abnormal Psychology, 105, 181±193. Zwick, W. R., & Velicer, W. F. (1986). Comparison of the ®ve rules for determining the number of components to retain. Psychological Bulletin, 99, 432±442.