PPLR

The dimensional structure of posttraumatic stress symptomatology in 323,903 U.S. veterans Ilan Harpaz-Rotem a,*, Jack Tsai b, Robert H. Pietrzak a, Rani Hoff c aNational Center for PTSD, VA Connecticut Healthcare System and Yale Department of Psychiatry, United StatesbThe New England Mental Illness Research Education and Clinical Center, VA Connecticut Healthcare System and Yale Department of Psychiatry, United States cThe Northeast Evaluation Program (NEPEC), the National Center for PTSD, VA Connecticut Healthcare System and Yale Department of Psychiatry, United States article info Article history:

Received 4 October 2013 Received in revised form 28 October 2013 Accepted 31 October 2013 Keywords:

PTSD Depression Anxiety Substance use disorder Psychopathology abstract There is ongoing debate regarding the optimal dimensional structure of posttraumatic stress disorder symptomatology. A better understanding of this structure has signiﬁcant implications, as it can provide more reﬁned phenotypic measures for use in studies of the etiology and neurobiology of PTSD, as well as for use as endpoints in treatment studies of this disorder. In this study we analyzed the dimensional structure of PTSD symptomatology, as assessed using the PTSD Symptom Checklist-Military Version in 323,903 Veterans. Conﬁrmatory factor analyses were used to compare two 4-factor models and a newly proposed 5-factor model to the 3-factor DSM-IV model of PTSD symptom dimensionality. To evaluate the external validity of the best-ﬁtting model, we then conducted a structural equation model examining how the symptom dimensions of this model related to diagnoses of depression, anxiety, and substance use disorder. Results indicated that a newly proposed 5-factor‘dysphoric arousal’model comprised of separate re-experiencing, avoidance, numbing, dysphoric arousal, and anxious arousal symptom clusters provided a signiﬁcantly betterﬁt to the data compared to the DSM-IV and the two alternative four-factor models. External validity analyses revealed that numbing symptoms were most strongly related to di- agnoses of depression and substance use disorder, and that dysphoric arousal symptoms were most strongly related to a diagnosis of anxiety disorder. Thus the dimensional structure of PTSD may be best represented byﬁve symptom dimensions. The clinical implications of these results and implications for further reﬁnement of extant PTSD assessment instruments are discussed.

Published by Elsevier Ltd. In recent years, there has been ongoing debate regarding the optimal characterization of the structure of posttraumatic stress disorder (PTSD) symptomatology (Armour et al., 2013a; Elhai et al., 2011;Elhai and Palmieri, 2011; Friedman et al., 2011; King et al., 1998; Shevlin et al., 2009; Simms et al., 2002). This debate was relevant to the reformulation of diagnostic criteria for PTSD in the recently published 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-V), and will likely continue to inform the evolving conceptualization of the component symptom clusters that comprise the heterogeneous phenotype that charac- terizes this disorder (Friedman et al., 2011).

PTSD is one of the most prevalent and disabling psychiatric disorders among U.S. Veterans (Harpaz-Rotem and Rosenheck,2011; Tanielian and Jaycox, 2008; Thomas et al., 2010). It is also prevalent among the general adult population, with 20% of the US adult population experiencing a traumatic event in a given year (Kessler et al., 2005; Kessler et al., 1995). A recent examination of the delivery of psychotherapy among the privately insured US population found that individuals diagnosed with PTSD were the most likely to receive psychotherapy compared to individuals diagnosed with other psychiatric disorders, thereby highlighting the psychological burden of this disorder (Harpaz-Rotem et al., 2012). To date, however, all studies that have examined the struc- ture of PTSD symptomatology in Veterans have employed relatively small sample sizes and thus, the generalizability of these results to the broader population of U.S. Veterans is unknown.

Several studies have challenged that the structure of PTSD symptoms is comprised of three symptom clusters, as speciﬁed in DSM-IV: re-experiencing (Criterion B), avoidance/numbing (Crite- rion C), and hyperarousal (Criterion D). Two alternative four-factor models have been proposed and conﬁrmatory factor analytic (CFA) * Corresponding author. The National Center for PTSD and Yale Department of Psychiatry, VACHS/116B, 950 Campbell, West Haven, CT 06511, United States.

E-mail address:[email protected](I. Harpaz-Rotem). Contents lists available atScienceDirect Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires 0022-3956/$esee front matter Published by Elsevier Ltd.

http://dx.doi.org/10.1016/j.jpsychires.2013.10.020 Journal of Psychiatric Research 49 (2014) 31e36 studies have suggested that they provide superiorﬁt to symptom- level PTSD data than the 3-factor DSM-IV model (King et al., 1998; Simms et al., 2002). These two four-factor models include the dysphoria model (Simms et al., 2002) and the emotional numbing model (King et al., 1998). In the dysphoria model, PTSD symptoms are separated into distinct four factors of re-experiencing, avoid- ance, dysphoria and hyperarousal symptoms. In the emotional numbing model, PTSD symptoms are separated into distinct four factors of re-experiencing, avoidance, emotional numbing and hy- perarousal symptoms (seeTable 1for details). Several CFA studies have demonstrated the superiority of the two 4-factor models of PTSD symptoms over the DSM-IV 3-factor model (Elhai et al., 2009; Elhai et al., 2008; Elhai et al., 2011; Engdahl et al., 2011; Grubaugh et al., 2010; Mansﬁeld et al., 2010; Palmieri et al., 2007; Yuﬁk and Simms, 2010). A recent meta-analysis suggested that the dysphoria model is only marginally superior than the numbing model in characterizing PTSD symptom structure, irrespective of the sample studied or PTSD assessment instrument employed (Yuﬁk and Simms, 2010). Given that each of the two 4-factors models have demonstrated superiority that varied based on sam- ple characteristics, testing conditions and instruments used, it is yet premature to determine which of these models better represents the latent structure of PTSD. The DSM-5 PTSD symptom clusters, however, more closely resemble the emotional numbing model in its inclusion of 4 symptom dimensions: intrusion symptoms, avoidance, negative alterations in cognitions and mood, and alter- ations in arousal and reactivity (Friedman et al., 2011).

One of the debates surrounding the latent structure of PTSD symptomatology involves three speciﬁc Hyperarousal symptoms (D1eD3) and whether they represent one of two constructse Dysphoria or Hyperarousal. In an attempt to address this issue, Shevlin et al. (2009)analyzed data from a large, nationally repre- sentative sample of civilian U.S. adults, and found that symptoms D1-D3 were not clear indicators of either the Dysphoria or the Hyperarousal factors but rather that they cross-loaded on both factors, thereby supporting the 4-factor dysphoria model. More recently,Elhai et al. (2011)have attempted to reconcile differences between the numbing and dysphoria models. In their investigation, they found support for a novel 5-factor model that separates fear- based panic symptoms (i.e., hypervigilance, exaggerated startle; “anxious arousal”) from hyperarousal symptoms characterized by dysphoria-related arousal abnormalities (symptoms D1eD3) rep- resented by anger/irritability, sleep difﬁculties, and concentrationproblems (i.e.,“dysphoric arousal”). This solution is in line with a theoretical model proposed byWatson (2005), which separates symptoms that involve restlessness and agitation (i.e., irritability) from more fear-based, panic-like symptoms (i.e., exaggerated startle response).

A growing number of CFA studies, which have been conducted in nationally representative civilian samples, general adult samples of medical patients, survivors of domestic violence, natural di- sasters, a violent riot, and military veterans have found that this newly proposed 5-factor model provides a signiﬁcantly better representation of PTSD symptom structure than the DSM-IV and both of the four-factor models (Armour et al., 2012, 2013a, 2013b; Elhai et al., 2011; Pietrzak et al., 2012a; Pietrzak et al., 2012b; Wang et al., 2011a, 2011c). Some studies have also examined how the 5-factor model relates to external measures of psychopathol- ogy, such as depression and anxiety (Pietrzak et al., 2012b; Wang et al., 2012; Wang et al., 2011b). Results of these studies have demonstrated that re-experiencing, avoidance, and anxious arousal symptoms are most strongly linked to anxiety, numbing symptoms to depression, and dysphoric arousal symptoms to both anxiety and depression. Results of these studies provide preliminary support for the external validity of the 5-factor PTSD model.

In the current study, we examined the dimensional structure of PTSD symptoms from more than 320,000 U.S. veterans who pre- sented for treatment at any VA medical center in the United States.

To our knowledge, this is the largest dataset of symptom-level PTSD data ever assembled and thus provides a unique opportunity to assess the dimensional structure of PTSD symptomatology. Our primary aim was to evaluate the potential robustness of the 5- factor model in characterizing PTSD symptom dimensionality relative to two 4- factor models (numbing and dysphoria) and the 3-factor DSM-IV model. Based on a growing body of CFA research that has highlighted the superiority of the 5-factor model relative to the two 4- and 3-factor DSM-IV models in characterizing the structure of PTSD symptoms (Armour et al., 2013a; Elhai et al., 2011; Pietrzak et al., 2012a), we hypothesized that the 5-factor model would provide the best structural representation of PTSD symp- tomatology in this sample. We then repeated the analyses with two restricted subsamples,ﬁrst with women only and then with only veterans diagnosed with PTSD to evaluate the stability of the factor structure among these important veteran subsamples. As a sec- ondary aim, we evaluated how the 5-factor Dysphoric Arousal model relates to other common psychiatric conditions in veter- ansddepression, other anxiety disorder, and substance use disorder.

1. Method Since the beginning of Fiscal Year 2008, the U.S. Department of Veterans Affairs (VA) has mandated mental health providers to assess PTSD symptoms using the PTSD Checklist Military Version (PCL-M; (Weathers et al., 1991) during their initial contact with Veterans who have served in the conﬂicts in Iraq and/or Afghanistan and who present for mental health assessment or treatment at a VA medical center. In 2010, the required reporting expanded to all patients with a diagnosis of PTSD. The PCL-M is a 17-item self-report instrument developed to assess the presence and severity of military-related PTSD symptoms that is based on the DSM-IV diagnostic criteria for PTSD. The PCL-M screening re- sults are collected by the VA mental health provider and are entered by the provider into the electronic medical record. Using the VA electronic medical record databases that capture outpatient care and test results, we identiﬁed PCL-M scores that were completed for every Veteran who received mental health care be- tween October 1, 2008 and September 31, 2012. The data were Table 1 Item mappings for each of the PTSD factor models.

PTSD symptom Model DSM-IV Dysphoria Numbing 5-Factor B1. Intrusive thoughts R R R R B2. Recurrent dreams R R R R B3. Flashbacks R R R R B4. Emotional reactivity R R R R B5. Physiological reactivity R R R R C1. Avoiding thoughts of trauma A A A A C2. Avoiding reminders of trauma A A A A C3. Inability to recall aspects of trauma A D N N C4. Loss of Interest A D N N C5. Detachment A D N N C6. Restricted affect A D N N C7. Sense of foreshortened future A D N N D1. Sleep disturbance H D H DA D2. Irritability/anger H D H DA D3. Difﬁculties concentrating H D H DA D4. Hypervigilance H H H AA D5. Exaggerated startle response H H H AA R¼Re-experiencing; A¼Avoidance; H¼Hyperarousal; D¼Dysphoria; N¼Emotional numbing; DA¼Dysphoric Arousal; AA¼Anxious Arousal. I. Harpaz-Rotem et al. / Journal of Psychiatric Research 49 (2014) 31e36 32 unduplicated to include only theﬁrst PCL-M on record, so that there are no repeated measurements for any individual veteran included in the current study. All PCL-M assessments with incom- plete data were also excluded. Theﬁnal sample included PCL-M scores from 323,903 unique Veterans. From the electronic medi- cal record, we additionally recorded participants’demographic data and all mental health diagnoses given to each individual within 365 days from the initial administration of the PCL-M. These diagnoses however, represent providers’clinical impression and judgment, and are not based on a structural diagnostic interview.

2. Sample characteristics Table 2shows the general demographic and clinical character- istics of our sample. The mean age of the sample was 44.2 and 59.2% were white. As expected, the majority of veterans were men (90.3%). The most commonly assigned clinician-assigned diagnoses within 12 months of veterans’initial PCL-M screening were PTSD (40.1%) and mood disorder (42.9%; Major Depressive disorder 12.5% and 30.4% dysthymia). Mean annual income of participants was $25,337 49,941. The majority of Veteran (58.5%) did not receive any disability income within 12 months of their initial PCL-M assessment. Only 12,851 (4%) of the veterans in the sample were classiﬁed as unemployable and were receiving 100% disability compensation from the VA. 65% of the veterans in the sample resided in a large urban area and 40.2% had served in the recent conﬂict in Iraq and/or Afghanistan.

3. Data analysis PCL-M scores were non-normally distributed, as evidenced by Mardia’s coefﬁcient for multivariate kurtosis>1.96. Thus, CFAs were conducted using robust maximum likelihood estimation with the Satorra and Bentler (SeB) c2scaling correction (Satorra and Bentler, 2001). This procedure estimates standard errors under conditions of multivariate non-normality and calculates other c2- dependentﬁt statistics based on the SeB c2statistic. In the CFA models, we speciﬁed PCL-M items to load only on one of the pro- posed factors each, all factors were allowed to correlate, all errorcovariances wereﬁxed to zero, and all tests were 2-tailed. In addition to the SeB c2, modelﬁt was evaluated using the comparativeﬁt index (CFI), Tucker Lewis Index (TLI), Akaike In- formation Criterion (AIC), Bayesian Information Criterion (BIC), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR) values. Higher CFI and TLI values and lower SeB c2, AIC, BIC, RMSEA, and SRMR values indicate a betterﬁtting model. Fit was also determined by empirically- deﬁned benchmarks, as follows: CFI and TLI 0.90 indicate adequateﬁt and 0.95 an excellentﬁt; RMSEA 0.08 as an adequateﬁt and 0.06 as indicative of excellentﬁt; and SRMR 0.08 generally considered as goodﬁt(Hu and Bentler, 1999). Finally, we calculated c2difference tests for nested models with a correction factor (given the use of the SeB c2statistic) to compare the relativeﬁt of theﬁve-factor model to the four-factor dysphoria and numbing models and the three-factor DSM-IV model (Fan and Sivo, 2009).

Last, we evaluated the external validity of theﬁve-factor model in relation to clinician-assigned diagnoses of depression, other anxiety disorder, and substance use disorder, which were diag- nosed within 365 days from the administration of the PCL-M.

Because the factors that comprise the 5-factor model are inher- ently highly correlated, we conducted these analyses using a structural equation model (SEM; seeFig. 1).

4. Results Table 3shows the results of the CFAs for the three-factor DSM-IV model, the four-factor dysphoria and numbing models, and the ﬁve-factor dysphoric arousal model. As indicated by lower SeB c2, AIC, BIC, RMSEA, and SRMR values, as well as higher CFI and TLI values, theﬁve-factor model provided a betterﬁt to the data than the three- and four-factor models. Most notably and relevant for the comparisons was theﬁve-factor lower BIC score compared to the other models. Compared to empirically deﬁned benchmarks, theﬁve-factor model indicated excellentﬁt (CFI and TLI 0.95; RMSEA 0.06; SRMR 0.08). c2difference tests further demon- strated that theﬁve-factor modelﬁt signiﬁcantly better than the three-factor DSM-IV model, Dc 2(7)¼11110 0 . 9 1,p<.001; the four- factor dysphoria model, Dc 2(4)¼34150.88,p<.001; and the four- factor numbing model, Dc 2(4)¼39172.59,p<.001. Moreover, Table 3demonstrates that theﬁve-factor’s CFI difference from the alternative models is greater than 0.01 indicating a betterﬁt(Fan and Sivo, 2009).

To assess the factor structure among important veteran sub- samples, we repeated the CFA analyses in two subgroups, women and veterans with a diagnosis of PTSD. When these CFAs were repeated on only female veterans (n¼31,259), the results remained the same. Theﬁve-factor model provided a betterﬁt than the three- and four-factor models as shown by lower SeB c2, AIC, BIC, RMSEA, and SRMR values; higher CFI and TLI values (CFI and TLI 0.95); and signiﬁcant c2difference tests.

When analyses were repeated on only veterans with PTSD (n¼129,808), the results also remained the same. Theﬁve-factor model provided a betterﬁt than the three- and four-factor models, as evidenced by lower SeB c2, AIC, BIC, RMSEA, and SRMR values; higher CFI and TLI values (CFI and TLI 0.95); and signiﬁcant c2difference tests.

To assess the external validity of theﬁve-factor model while accounting for the high correlations among theﬁve factors, we conducted a structural equation model to assess association be- tween theﬁve-factor model and clinician-assigned diagnoses of depression, other anxiety disorder, and substance use disorder, which highly prevalent among veterans diagnosed with PTSD (see Fig. 1). Results indicated a unique and signiﬁcant contribution of Table 2 Characteristics of study sample (n¼323,903).

Variable Mean (SD) orn(%) Mean age (SD) 44.16 (15.50) Male 292,644 (90.35%) Married 156,132 (48.20%) Race White (Non-Hispanic) 191,641(59.17%) Black 50,379 (15.55%) Hispanic 27,128 (8.38%) Other 54,755 (16.90%) Urban (vs. Rural) 210,632 (65.03%) Service-connection 0% or not service-connected 189,682 (58.56%) 1e99% 121,370 (37.47%) 100% 12,851 (3.97%) Psychiatric diagnoses PTSD 129,808 (40.08%) Dysthymia 98,484 (30.41%) Other anxiety disorder 58,584 (18.09%) Alcohol use disorder 41,771 (12.90%) Major depression 40,436 (12.48%) Adjustment disorder 32,147 (9.92%) Drug use disorder 28,114 (8.68%) Bipolar disorder 11,066 (3.42%) Personality disorder 6339 (1.96%) Other psychotic disorder 4295 (1.33%) Schizophrenia 4970 (1.53%) OEF/OIF status 130,263 (40.22%) I. Harpaz-Rotem et al. / Journal of Psychiatric Research 49 (2014) 31e3633 each factor to each of these diagnoses. The Model’s c2(148)¼222,294, CFI¼0.95 and RMSEA¼0.031 indicated a relatively goodﬁt. Standardized estimates are presented inTable 4.

The results indicated that while each of theﬁve factors were signiﬁcantly associated with depression, other anxiety disorder, and substance use disorder, the strongest associations were observed between numbing symptoms and depression and sub- stance use disorder; and between dysphoric arousal and other anxiety disorder.

5. Discussion In this study, we evaluated the dimensional structure of PTSD in more than 300,000 U.S. veterans, one of the largest samples to have ever been studied in the CFA literature on PTSD. We further eval- uated the external validity of theﬁve-factor model by examining its association with common comorbid diagnoses using structural equation modeling, which accounts for the high intercorrelations among theﬁve factors. Results of CFAs demonstrated the superi- ority of the 5-factor model of re-experiencing, avoidance, numbing, dysphoric arousal, and anxious arousal (Elhai et al., 2011) in com- parison to the three-factor DSM-IV and 4-factor numbing and dysphoria models in characterizing PTSD symptom dimensionality.

Theseﬁndings extend prior studies that have recently beenpublished from relatively smaller samples of OIF/OEF/OND veterans (Pietrzak et al., 2012a), survivors of domestic violence (Elhai et al., 2011), an earthquake (Wang et al., 2011a), a violent riot (Wang et al., 2011a), and of the general population (Armour et al., 2013a), which similarly found that the dimensional structure of PTSD symptom- atology is best represented by the 5-factor dysphoric arousal model.

External validity analysis using SEM revealed signiﬁcant asso- ciations between each of theﬁve factors, and clinician-based di- agnoses of depression, other anxiety disorder, and substance use disorder. Numbing symptoms were most strongly related to di- agnoses of depression and substance use disorder, while dysphoric arousal symptoms were most strongly related to a diagnosis of other anxiety disorder. Theseﬁndings extend on work, which has highlighted the association between PTSD numbing symptoms and depression and, hyperarousal symptoms and anxiety (Pietrzak et al., 2010; Wang et al., 2011a, 2011b) to suggest unique associa- tions between numbing symptoms and substance use disorder, and between dysphoric arousal and other anxiety disorder.

Results of our study provide further empirical support for Wat- son’s theoretical model where general distress/dysphoric symp- toms are described as being distinct from more fear-based, panic symptoms that characterize this disorder (Watson, 2005). In this respect, symptoms associated with the dysphoric arousal cluster, Fig. 1.SEM model assessing external correlates of the PTSD 5-factor model.*All 5 factors were allowed to correlate with each other in the model. Table 3 Fit indices of conﬁrmatory factor analyses of the 17-item PTSD Checklist (N¼323,903).

Model SeB c2 df CFI TLI RMSEA AIC BIC SRMR 3-factor DSM-IV 196,842.06 116 0.93 0.92 0.07 15,555,200.23 15,555,777.39 0.04 4-factor dysphoria 107,435.52 113 0.96 0.96 0.05 15,435,737.75 15,436346.97 0.03 4-factor numbing 112,457.23 113 0.96 0.95 0.06 15,442,706.92 15,443,316.15 0.03 5-factor 73,284.64 109 0.98 0.97 0.05 15,390,605.02 15,391,257.00 0.02 I. Harpaz-Rotem et al. / Journal of Psychiatric Research 49 (2014) 31e36 34 which are primarily characterized by restlessness and agitation (e.g., irritability) are considered to be conceptually distinct from the numbing cluster symptoms, which are characterized by numbing of responsiveness and anhedonia. The symptoms of the dysphoric arousal cluster are further considered distinct from the two other symptoms that comprise the DSM-IV hyperarousal clustere hypervigilance and exaggerated startle, which are physiological and fear-based symptoms (Elhai et al., 2011).

Ourﬁndings have several potential clinical implications. First, they suggest that the dysphoric and anxious arousal symptoms may constitute independent, theoretically distinct symptom clus- ters. This separation of the hyperarousal symptom cluster is important, as it helps to inform understanding of the optimal phenotypic model of this aspect of PTSD and accordingly, can inform studies of the underlying neurobiology of PTSD. Second, characterization of the optimal symptom structure of PTSD can also increase the reliability of clinical evaluation and assessment of component elements of PTSD, and provide more reﬁned endpoints for use as treatment outcome measures. Third, hyperarousal symptoms that are considered to play a major role in the persis- tence of PTSD (Marshall et al., 2006; Schell et al., 2004; Solomon et al., 2009). Thus, better understanding of which aspect of hy- perarousaleanxious arousal or dysphoric arousaleis most strongly related to the development and maintenance of other, more disabling symptom clusters, such as emotional numbing, may help guide the development of more reﬁned prevention and treatment approaches. Fourth, the proposedﬁve-factor model of PTSD symptom classiﬁcation may also provide a more nuanced understanding of the nature and etiology of the high levels of co- morbid psychopathology associated with PTSD (Biehn et al., 2013).

The distinction of dysphoric and anxious arousal may also assist researchers and clinicians in better assessing the effectiveness of PTSD treatment by providing more reﬁned domains to be used in assessing treatment outcomes. Both prolonged exposure therapy (PE) and cognitive processing therapy (CPT) have demonstrated promising results in mitigating PTSD symptoms, primarily reex- periencing, hypervigilance, avoidance, and numbing symptoms, but has had only limited success in reducing the dysphoric arousal symptoms, particularly sleep problems. Thisﬁnding provides clinical support to the independence nature of this symptom cluster (Galovski et al., 2009; Resick et al., 2012). For instance, when assessing the efﬁcacy of CPT and PE on sleep, it was documented that although participants in the two treatment groups experi- enced improvements in their sleep, irrespective of the treatment condition and treatment gains, participants never reached what authors deﬁned as“good sleeper status”or a“normal sleep func- tioning”(Galovski et al., 2009). Interestingly, CPT non-responsive clients were found to be presenting with signiﬁcantly higherhyperarousal symptoms then responders (Stein et al., 2012), thereby suggesting potential relationships between the distinct anxious arousal cluster and the mechanism of change in CPT.

Additional research is needed to evaluate this possibility.

Given the VA’s use of PCL-Ms to screen new veterans presenting for services, the ﬁve-factor model may also be useful in providing recommendations for speciﬁc targeted areas for treatment for veterans who screen positive for PTSD. Currently, scores above a total threshold PCL-M score trigger a positive screen, but subscale scores from a more reﬁned phenotypic model could yield more detailed, patient-speciﬁc clinical information that may be useful for treatment providers.

This study has several limitations. First, theﬁnal factor solution was based on comparingﬁt statistics between models using conﬁr- matory factor analysis, and the differences, while consistent with emerging literature, were not dramatic, which may be due to vari- ability in measurement and administration of the PCL-M in a national healthcare system. Moreover, there was high inter-factor correlation which is not a problem in conﬁrmatory factor analysis but multi- collinearity needs to be considered if the factors are used together to predict other variables. Second, all study participants were U.S.

Armed Forces personnel or veterans of the U.S. Armed Forces pre- senting for treatment or evaluation and thus it is not a representative sample of the broader U.S. veteran population. In addition, it is un- clear which veterans in our sample were compensation-seeking versus treatment-seeking. Replication of these results with other samples is needed. Nevertheless, one of the strength associated with the collection of PCL scores in the VA Healthcare System is that veterans completed PCL-Ms with their mental health providers and thus, the PCL-Ms were not based purely on self-report. Third, both the avoidance and anxious arousal factors are comprised of only two PCL-M items. Thus it is difﬁcult to determine their validity and reli- ability in assessing these broader latent constructs.

6. Conclusion This study utilized data from over 320,000 veterans who responded to a widely used self-report measure of military-related PTSD during their initial VA mental health appointment. The data provide strong support for a theory-drivenﬁve-factor model of PTSD symptoms. Additional research is needed to evaluate the clinical utility of this model; its relation to neurobiological and genetic markers; and how this model may help to inform the reﬁnement of PTSD assessment instruments.

Conﬂict of interest none to report.

Contributors Harpaz-Rotem, IlaneDesign, data collection, data analyses and writing.

Tsai, Jack: data analyses and writing.

Pietrzak, Robert: Assist in data analyses interpretation and writing.

Hoff Rani: Data collection and writing.

Role of funding source No funding was provided.

Disclosures and acknowledgments None to disclose. Table 4 Standardized coefﬁcients in SEM of 5-factor model and co-occurring psychopathologies.

Factors Depression Anxiety SUD B (95% CI) B (95% CI) B (95% CI) Reexperiencing 0.143 (0.141e0.145) 0.080 (0.077e0.083) 0.107 (0.104e0.110) Avoidance 0.145 (0.143e0.147) 0.073 (0.070e0.076) 0.095 (0.092e0.098) Numbing 0.219 (0.217e0.221) 0.085 (0.082e0.088) 0.146 (0.143e0.149) Dysphoric arousal0.171 (0.169e0.173) 0.102 (0.099e0.105) 0.095 (0.092e0.098) Anxious arousal0.113 (0.111e0.115) 0.091 (0.088e0.094) 0.098 (0.095e0.101) p<.0001 in all correlations between the 5 factors and diagnosis.

95%CI¼95% conﬁdence interval. I. Harpaz-Rotem et al. / Journal of Psychiatric Research 49 (2014) 31e3635 Acknowledgment None to report References Armour C, Carragher N, Elhai JD. Assessing theﬁt of the Dysphoric Arousal model across two nationally representative epidemiological surveys: the Australian NSMHWB and the United States NESARC. J Anxiety Disord 2013a;27:109e15.

Armour C, Elhai JD, Richardson D, Ractliffe K, Wang L, Elklit A. Assessing aﬁve factor model of PTSD: is dysphoric arousal a unique PTSD construct showing differ- ential relationships with anxiety and depression? J Anxiety Disord 2012;26:

368e76.

Armour C, Raudzah Ghazali S, Elklit A. PTSD’s latent structure in Malaysian tsunami victims: assessing the newly proposed Dysphoric Arousal model. Psychiatry Res 2013b;206:26e32.

Biehn TL, Contractor A, Elhai JD, Tamburrino M, Fine TH, Prescott MR, et al. Relations between the underlying dimensions of PTSD and major depression using an epidemiological survey of deployed Ohio National Guard soldiers. J Affect Dis- ord 2013;144:106e11.

Elhai JD, Biehn TL, Armour C, Klopper JJ, Frueh BC, Palmieri PA. Evidence for a unique PTSD construct represented by PTSD’s D1-D3 symptoms. J Anxiety Disord 2011;25:340e5.

Elhai JD, Engdahl RM, Palmieri PA, Naifeh JA, Schweinle A, Jacobs GA. Assessing posttraumatic stress disorder with or without reference to a single, worst traumatic event: examining differences in factor structure. Psychol Assess 2009;21:629e34.

Elhai JD, Grubaugh AL, Kashdan TB, Frueh BC. Empirical examination of a pro- posed reﬁnement to DSM-IV posttraumatic stress disorder symptom criteria using the National Comorbidity Survey Replication data. J Clin Psychiatry 2008;69:597e602.

Elhai JD, Palmieri PA. The factor structure of posttraumatic stress disorder: a liter- ature update, critique of methodology, and agenda for future research. J Anxiety Disord 2011;25:849e54.

Engdahl RM, Elhai JD, Richardson JD, Frueh BC. Comparing posttraumatic stress disorder’s symptom structure between deployed and nondeployed veterans.

Psychol Assess 2011;23:1e6.

Fan X, Sivo SA. Using goodness-of-ﬁt indexes in assessing mean structure invari- ance. Struc Equat Model 2009;16:54e67.

Friedman MJ, Resick PA, Bryant RA, Brewin CR. Considering PTSD for DSM-5.

Depress Anxiety 2011;28:750e69.

Galovski TE, Monson C, Bruce SE, Resick PA. Does cognitive-behavioral therapy for PTSD improve perceived health and sleep impairment? J Trauma Stress 2009;22:197e204.

Grubaugh AL, Long ME, Elhai JD, Frueh BC, Magruder KM. An examination of the construct validity of posttraumatic stress disorder with veterans using a revised criterion set. Behav Res Ther 2010;48:909e14.

Harpaz-Rotem I, Libby D, Rosenheck RA. Psychotherapy use in a privately insured population of patients diagnosed with a mental disorder. Soc Psychiatry Psy- chiatr Epidemiol 2012;47:1837e44.

Harpaz-Rotem I, Rosenheck RA. Serving those who served: retention of newly returning veterans from Iraq and Afghanistan in mental health treatment.

Psychiatr Serv 2011;62:22e 7.

Hu L, Bentler PM. Cutoff criteria forﬁt indexes in covariance structure analysis:

conventional criteria versus new alternatives. Struct Equat Model 1999;6:1e55.

Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 2005;62:593e602.

Kessler RC, Sonnega A, Bromet E, Hughes M, Nelson CB. Posttraumatic stress disorder in the National Comorbidity Survey. Arch Gen Psychiatry 1995;52:

10 4 8e60.

King DW, Leskin GA, King LA, Weathers FW. Conﬁrmatory factor analysis of the Clinician-Administered PTSD Scale: evidence for the dimensionality of post- traumatic stress disorder. Psychol Assess 1998;10:90e6.Mansﬁeld AJ, Williams J, Hourani LL, Babeu LA. Measurement invariance of post- traumatic stress disorder symptoms among U.S. military personnel. J Trauma Stress 2010;23:91e9.

Marshall GN, Schell TL, Glynn SM, Shetty V. The role of hyperarousal in the mani- festation of posttraumatic psychological distress following injury. J Abnorm Psychol 2006;115:624e8.

Palmieri PA, Weathers FW, Difede J, King DW. Conﬁrmatory factor analysis of the PTSD Checklist and the clinician-administered PTSD Scale in disaster workers exposed to the World Trade Center Ground Zero. J Abnorm Psychol 2007;116:

329e41.

Pietrzak RH, Goldstein MB, Malley JC, Rivers AJ, Southwick SM. Structure of posttraumatic stress disorder symptoms and psychosocial functioning in veterans of operations enduring freedom and Iraqi freedom. Psychiatry Res 2010;178:323e9.

Pietrzak RH, Tsai J, Harpaz-Rotem I, Whealin JM, Southwick SM. Support for a novel ﬁve-factor model of posttraumatic stress symptoms in three independent samples of Iraq/Afghanistan veterans: a conﬁrmatory factor analytic study.

J Psychiatr Res 2012a;46:317e22.

Pietrzak RH, Van Ness PH, Fried TR, Galea S, Norris F. Diagnostic utility and factor structure of the PTSD Checklist in older adults. Int Psychogeriatr 2012b;24:

1684e96.

Resick PA, Williams LF, Suvak MK, Monson CM, Gradus JL. Long-term outcomes of cognitive-behavioral treatments for posttraumatic stress disorder among fe- male rape survivors. J Consult Clin Psychol 2012;80:201e10.

Satorra A, Bentler PM. A scaled difference chi-square test statistic for moment structure analysis. Psychometrika 2001;66:507e14.

Schell TL, Marshall GN, Jaycox LH. All symptoms are not created equal: the prom- inent role of hyperarousal in the natural course of posttraumatic psychological distress. J Abnorm Psychol 2004;113:189e97.

Shevlin M, McBride O, Armour C, Adamson G. Reconciling the differences between the King et al. (1998) and Simms, et al. (2002) factor models of PTSD. J Anxiety Disord 2009;23:995e10 01.

Simms LJ, Watson D, Doebbeling BN. Conﬁrmatory factor analyses of posttraumatic stress symptoms in deployed and nondeployed veterans of the Gulf War.

J Abnorm Psychol 2002;111:637 e47.

Solomon Z, Horesh D, Ein-Dor T. The longitudinal course of posttraumatic stress disorder symptom clusters among war veterans. J Clin Psychiatry 2009;70:

837e43.

Stein NR, Dickstein BD, Schuster J, Litz BT, Resick PA. Trajectories of response to treatment for posttraumatic stress disorder. Behav Ther 2012;43:790e800.

Tanielian T, Jaycox LH. Invisible wounds of war: psychological and cognitive injuries, their consequences, and services to assist recovery. Santa Monica, CA: The RAND Center for Military Health Policy Research; 2008.

Thomas JL, Wilk JE, Riviere LA, McGurk D, Castro CA, Hoge CW. Prevalence of mental health problems and functional impairment among active component and National Guard soldiers 3 and 12 months following combat in Iraq. Arch Gen Psychiatry 2010;67:614e23.

Wang L, Cao C, Wang R, Zhang J, Li Z. The dimensionality of PTSD symptoms and their relationship to health-related quality of life in Chinese earthquake sur- vivors. J Anxiety Disord 2012;26:711e8.

Wang L, Li Z, Shi Z, Zhang J, Zhang K, Liu Z, et al. Testing the dimensionality of posttraumatic stress responses in young Chinese adult earthquake survivors:

further evidence for“dysphoric arousal”as a unique PTSD construct. Depress Anxiety 2011a;28:1097e10 4.

Wang L, Long D, Li Z, Armour C. Posttraumatic stress disorder symptom structure in Chinese adolescents exposed to a deadly earthquake. J Abnormal Child Psychol 2011:749e58.

Wang L, Zhang J, Shi Z, Zhou M, Li Z, Zhang K, et al. Comparing alternative factor models of PTSD symptoms across earthquake victims and violent riot witnesses in China: evidence for aﬁve-factor model proposed by Elhai et al. (2011).

J Anxiety Disord 2011c;25:771e6.

Watson D. Rethinking the mood and anxiety disorders: a quantitative hierarchical model for DSM-V. J Abnorm Psychol 2005;114:522e36.

Weathers FW, Litz BT, Huska JA, Keane TM. The PTSD checklist (PCL). p. Boston, Massachusetts: National Center for PTSD; 1991.

Yuﬁk T, Simms LJ. A meta-analytic investigation of the structure of posttraumatic stress disorder symptoms. J Abnorm Psychol 2010;119:764e76. I. Harpaz-Rotem et al. / Journal of Psychiatric Research 49 (2014) 31e36 36