Annotated Bibliography

Open Posted By: surajrudrajnv33 Date: 23/02/2021 Graduate Case Study Writing

Complete an annotated bibliography for the following 5 articles attached. 

I have attached an example of the layout of this paper. 


Follow this outline exactly---

Provide ALL three elements:

a. Citation (APA format)

b. Summary of article – usually 1 or 2 paragraphs

c. Your analysis – usually 1 or 2 sentences NOT a full critique

Topic: the influence of combat and trauma on war veterans in the U.S.

- discuss Comorbidity between PTSD AND anxiety, depression, and substance use

Category: Accounting & Finance Subjects: Behavioral Finance Deadline: 12 Hours Budget: $120 - $180 Pages: 2-3 Pages (Short Assignment)

Attachment 1

MILITARY MEDICINE, 182, 7/8:e1787, 2017

The Influence of Combat Experience on Psychologically Healthy Soldiers’ Attentiveness to Environmental Threats

Bethany Ranes, PhD*†; MAJ Chris P. Long, MS USAR*‡; CPT Stephanie Traynham, MS USA*; Amanda Hayes*§

ABSTRACT Introduction: In contrast to previous research that has primarily examined how psychological disorders (e.g., post-traumatic stress disorder [PTSD], anxiety) are affected by and affect individuals’ threat perceptions, this study examines the relationship between combat experience and threat-monitoring in psychologically healthy Soldiers. Existing research has established how prolonged or intense experiences with war-related stressors can lead individuals to undergo an unconscious fear-conditioning process that affects the circuitry of the prefrontal cortex, hippocampus, amygdala, and anterior cingulate cortex regions of the brain. We predict that the intensity of one’s combat experience positively influences Soldiers’ attention to environmental threats. Materials and Methods: Participants included U. S. Army Soldiers with a score of 50 or below on the PTSD Checklist-Military Version. Participants completed the Com- bat Exposure Scale and the State-Trait Anxiety Inventory. The experimental prediction task we employed assesses the expectation of an intrusively loud white noise sound that occurred on three variable patterns in a pseudorandomized order. Each tone pattern was used 20 times over a total of 60 trials. The experimental prediction task included two neutral tones (700 and 1,300 Hz) that were presented in a repeated pattern along with a 100-dB burst of white noise (0.5-second duration). In each trial, one of three possible tone combinations was presented. To assess their attentive- ness to threats, participants were asked to continuously rate their expectancy of the burst of white noise using a visual analogue scale (VAS) ranging from 0 to 100. VAS ratings were collected at controlled points throughout the task. Results: None of the participants reported scores on any of the diagnostic surveys that met standards for clinical signif- icance. A repeated-measures analysis of variance was conducted to assess the overall effect of the three prediction con- ditions on participants’ VAS ratings. There was a significant main effect for Combat Exposure Scale scores on VAS ratings [F(1, 27) = 5.19, p = 0.031], with high scorers demonstrating a generally higher expectancy of the white noise burst throughout the entire experimental sequence. Results suggest that within subclinical populations of Soldiers, the intensity of one’s combat experience is positively associated with their attention to threats. Conclusion: These findings suggest that Soldiers who experience combat should be observed for signs of increased threat-attention bias, as this may indicate that their capacities for information processing, decision-making, and emotion regulation could be compromised. The positive relationship we observe between a level of combat experience and attentional biases toward threatening stimuli may also help to explain why these veterans engage in “externalizing” behaviors that are risky, aggressive, or violent as well as relational problems and antisocial behaviors that are reported in higher-than- average rates among these populations of Soldiers. Acknowledging that increased threat attention may be a preclinical indication of developing PTSD or other related psychological conditions (e.g., depression, anxiety) should motivate clinicians to more actively diagnose and treat this condition.

INTRODUCTION The experiences of Soldiers in Operation Iraqi Freedom and Operation Enduring Freedom have renewed attention on the often intense psychological stressors associated with combat. As evidenced by the high incidence of post-traumatic stress disorder (PTSD) and related psychological conditions (anxi- ety, depression, etc.), abundant research1,2 now shows how combat experience can significantly influence how individ- uals perceive and react to factors in their environments.

Although the most severe psychological injuries garner significant attention from both clinicians and researchers,

combat experiences may also generate more subtle but important and lasting changes to a Soldier’s psychology. In many cases, these changes significantly affect how these individuals experience environmental cues, process informa- tion, and relate to others with whom they routinely interact. While these changes may manifest in Soldiers who other- wise appear psychologically healthy, they may leave these individuals with lingering questions about what these changes mean and why they have occurred.

In this article, we examine one such change that can occur in otherwise psychologically healthy soldiers: how prolonged or intense experiences with war-related stressors can increase the proclivity of individuals to exhibit threat- attention biases. Shin and Handwerger3 suggest that threat- attention biases develop as Pavlovian fear conditioning affects the circuitry in the prefrontal cortex, hippocampus, amygdala, and anterior cingulate cortex regions of the brain.4,5 In an exhaustive review of relevant research, Clark et al6 find that when participants who exhibit threat-attention bias encounter fear potentiated stimuli, they display exaggerated startle

*U.S. Army Aeromedical Research Laboratory, P.O. Box 320577, Fort Rucker, AL 36362.

†Butler Center for Research, Hazelden Betty Ford Foundation, P.O. Box 11, Center City, MN 55012.

‡McDonough School of Business, Georgetown University, Rafik B Hariri Building, 37th and O Streets NW, Washington, DC 20057.

§Oak Ridge Institute for Science and Education, MC-100-44, P.O. Box 37831-3146, Oak Ridge TN 37831.

doi: 10.7205/MILMED-D-16-00261

MILITARY MEDICINE, Vol. 182, July/August 2017 e1787

Downloaded from publications.amsus.org: AMSUS - Association of Military Surgeons of the U.S. IP: on Jul 06, 2017.

Copyright (c) Association of Military Surgeons of the U.S. All rights reserved.

responses in their electroencephalogram recordings.4,5,7–12

Fani et al8 also report that during a dot-probe task, individ- uals who exhibit threat-attention biases display exaggerated startle responses when shown both threatening and neutral facial expressions.

Individuals who exhibit threat-attentional biases often must manage significant psychological and physiological challenges.4,5,8–12 Eysenck et al’s4 exhaustive review of the literature on attentional control describes how threat-attention biases can compromise one’s ability to accurately process environmental information as these individuals can perceive even benign environmental conditions as threatening.5,13

Because individuals with threat-attention biases have diffi- culty shifting their attention away from perceived threats, they expend excessive cognitive resources in monitoring the real and potential threats that they perceive in their environ- ments.5–7,14 They also tend to devote significant physical resources to regulating their often strong and negative behav- ioral reactions to stimuli they perceive as threatening.4,15

This Study Although previous research has primarily examined the connection between manifest psychological disorders (e.g., PTSD, anxiety) and threat-attention biases, we examine the relationship between combat experience and threat-monitoring in otherwise psychologically healthy Soldiers.4,5,7,8,11,12,16 We specifically predict that the intensity of one’s combat experi- ence positively influences how vigilantly these combat vet- erans attend to environmental threats. We test this idea using an experimental task where participants predict when an intru- sively loud white noise stimulus will be administered. In our experiment, we evaluate whether individuals with more intense combat experiences anticipate white noise stimuli at higher levels when compared with individuals who had less intense combat experiences.

We focus our research on threat-attention biases in other- wise psychologically healthy Soldiers to assess a key facet of the psychology of combat veterans. If the combat experi- ences of these individuals promote exaggerated threat expec- tations or overly active fear-potentiated startle responses in the absence of PTSD or related disorders, it can help us explain more subtle but still observable psychological changes in this population of Soldiers.5,12,16 Moreover, if we can detect changes in how otherwise psychologically healthy combat veterans react to environmental cues, it may alter how these facets of Soldier psychology are identified and treated on the battlefield.

In addition to influencing Soldier performance, an impor- tant implication of this work is that it can help clinicians identify a condition that is often a precursor to PTSD. This idea builds from work by Fani et al8 who suggest that a hypervigilance to both threatening and neutral stimuli can foster the initial development or intensification of PTSD as well as related conditions such as anxiety and depres-

sion.4,5,7,8,11,12,16 They suggest that this occurs because an attention-related bias toward threats exacerbates changes in the brain that intensify PTSD symptoms and make it more difficult to ameliorate this condition. We, however, contend that if threat-attention biases can be identified as a Soldier’s psychological condition begins to change, we may be able to prevent or fully mitigate the development of PTSD and other more serious psychological conditions (i.e., anxiety, depression) in some affected warfighters.

In the next section, we outline our methodology and pro- vide details about the administration of our study. After presenting our results, we discuss our findings in relation to current research on attention, specify some limitations and extensions of our investigation, and evaluate the implications of this study for clinical practice.


Participants The study protocol was approved by the Headquarters, U.S. Army Medical Research and Materiel Command Institu- tional Review Board. A convenience sample of U.S. Army Active Duty and National Guard/Reserve Soldiers in the greater Fort Rucker, Alabama, area were recruited as partici- pants. Participants were recruited through word-of-mouth, flyers, posters, and e-mail. To claim compensation, partici- pants completed the study during their “off-duty” time. Written informed consent was obtained from all volunteers. All 40 participants who consented completed the study examining the relationship between combat exposure and vigilance to threats. Cohen’s conventional criteria (medium, d = 0.5) in a within-group analyses of variance (ANOVA) confirmed that 40 participants would provide sufficient statistical power at 80% to detect statistically significant medium effects.

The mean age of participants was 28.54 years (standard deviation [SD] = 8.31 years). Eight participants were female, and 32 were male. Eighteen participants, who had never been deployed to a combat zone served as the con- trol group. Participants’ time-in-service ranged from 3 to 288 months, with a mean of 83.18 months (SD = 70.20 months). Participants included both officers (n = 6) and enlisted (n = 34) Soldiers.

As the stimuli used in the experiment were auditory, exclusion criteria for all participants included a hearing pro- file per Army Regulation 40-501 and a diagnosis of tinnitus. In addition, because of the unpleasant and potentially star- tling nature of some of the experimental stimuli, participants with diagnoses of PTSD and/or psychotic disorders were disqualified. Finally, participants reporting clinically signifi- cant PTSD symptoms at the time of the experiment were not allowed to participate (i.e., a score of 50 or higher on the PTSD Checklist–Military Version [PCL-M]). A similar exclu- sion criterion was used by Stetz et al.17 The mean PCL-M score for participants in this study was 20.44 (SD = 0.65).

MILITARY MEDICINE, Vol. 182, July/August 2017e1788

Combat Experience and Attentiveness to Threat

Downloaded from publications.amsus.org: AMSUS - Association of Military Surgeons of the U.S. IP: on Jul 06, 2017.

Copyright (c) Association of Military Surgeons of the U.S. All rights reserved.

Experimental Prediction Task The experimental prediction task we employed is founded on the methods used by Knight, Nguyen, and Bandettini.18

The task assesses a participant’s expectation of an intru- sively loud white noise sound. Participants were exposed to three tone patterns with each tone pattern being used 20 times over a total of 60 trials. The experimental predic- tion task included two neutral tones (700 and 1,300 Hz; 10-second duration) and a 100-dB burst of white noise (0.5-second duration) that were presented in a repeated pat- tern. In each trial, participants encountered 1 of 3 possible tone combinations. In one tone combination, the white noise burst immediately followed the 700-Hz tone. In the second combination, the 1,300-Hz tone was always presented alone (not preceded or followed by a white noise burst). In the third combination, the white noise burst was presented to participants after a 20-second pause. Six trials (two sets of each combination) were grouped in a pseudorandomized order. All trial groups were separated by a 20-second pause. Table I presents an exemplar trial group schedule.

Criterion Variable

Visual Analogue Scale Ratings

To assess their attentiveness to threats, participants were asked to continuously rate their expectancy of a burst of white noise using a visual analogue scale (VAS) ranging from 0 to 100.16 This VAS rating provided the primary dependent variable examined in this study. A visual repre- sentation of the VAS is provided in Figure 1. Using pre- sentation software, the VAS was displayed on a computer screen throughout the entire duration of the prediction task. A computer mouse was used to control the rating bar on the VAS in real time, and participants were asked to contin- uously monitor their level of expectancy for the white noise.

They were instructed to give a rating between 0 and 100 on the basis of their level of certainty that the white noise would sound. Zero on the scale indicated that the participant was certain the stimulus would not occur; 100 indicated that the participant was certain the stimulus would occur. If par- ticipants were entirely unsure of whether the stimulus would occur, they were instructed to choose a VAS score of 50. For computational purposes, final VAS ratings were trans- formed by subtracting 50 from the raw VAS rating so nega- tive values indicated that the participant did not expect the stimulus and the absolute value of the score indicated the magnitude at which they rated their expectancy. A trans- formed rating of 0 indicates that the participant was unsure of whether the stimulus would occur.

VAS ratings were collected at controlled points through- out the task. The presentation software recorded VAS ratings at intervals that were either 9 seconds before the white noise, 19 seconds before the white noise, and a control point with no immediate discernible white noise following (69 seconds before the white noise). These points were classified to rep- resent short-term prediction ratings, long-term prediction ratings, and control ratings, respectively. Ratings were used to determine how well participants were able to predict the occurrence of the white noise after continuous exposure to the repeating tone patterns and intertone pauses. The participants were unaware of when these collection points occurred and simply updated the VAS rating bar in real time as they attempted to predict when the white noise would sound. We use our analysis of the VAS ratings to evaluate our general prediction that Soldiers with more exposure to combat will present significantly higher predic- tion rates of white noise at 9 and/or 19 seconds than Soldiers with less combat experience.

Predictor Variable

Combat Exposure Scale

Combat experiences were measured using the Combat Exposure Scale (CES). The CES is a seven-item measure allowing for Likert-type responses on a 5-point scale, designed to measure combat experiences. The items on the CES ask questions specifically related to the frequency and intensity of situations of enemy contact during deploy- ments (e.g., “Were you ever surrounded by the enemy?”). Higher scores on this scale indicate a higher intensity of combat experience.19 The Cronbach’s alpha for this measure was 0.92.

TABLE I. Schedule of Events

Session Activities

1. In-processing Informed Consent PTSD Screening

2. Surveys Demographic Survey CES PCL-M SAI TAI

3. Testing Prediction Task Schedule of Trial Group Example

Tone Combination Duration Pause 20 Seconds White Noise Only 20 Seconds 1,300-Hz Tone 30 Seconds 700-Hz Tone and White Noise 30 Seconds White Noise Only 20 Seconds 700-Hz Tone and White Noise 30 Seconds 1,300-Hz Tone 30 Seconds

FIGURE 1. Experimental prediction task VAS example.

MILITARY MEDICINE, Vol. 182, July/August 2017 e1789

Combat Experience and Attentiveness to Threat

Downloaded from publications.amsus.org: AMSUS - Association of Military Surgeons of the U.S. IP: on Jul 06, 2017.

Copyright (c) Association of Military Surgeons of the U.S. All rights reserved.

Control Variables

PTSD Checklist–Military Version

The PCL-M is a widely used self-administered questionnaire with 17 questions assessing trauma-related stress.20 Response options range from “Not at All” to “Extremely” (1 to 5), with higher numbers indicating greater stress. Possible scores range from 17 to 85. Each item is relevant to symptoms that have been observed in service members clinically diagnosed with PTSD (e.g., “Loss of interest in the things you used to enjoy?”). Higher cumulative scores on all items indicate a higher level of PTSD-related symptoms. The Cronbach’s alpha for this mea- sure was 0.75.

The State-Trait Anxiety Inventory

The Spielberger Trait Anxiety (TAI) and State Anxiety Inven- tory (SAI)21 ask respondents to rate 40 items on a 4-point scale. Higher scores on each dimension indicate the presence of higher levels of state or trait anxiety. Trait anxiety is described as relatively stable anxiety an individual experi- ences at all times, whereas state anxiety is described as situa- tional anxiety.22 The 20 items that describe aspects of trait anxiety (e.g., “I worry too much over something that doesn’t matter.”) achieved a Cronbach’s alpha of 0.86. The 20 items that describe state anxiety (e.g., “I am tense.”; “I am worried.”) achieved a Cronbach’s alpha of 0.87.

Demographic Information

A 10-item demographic survey reporting age, gender, ethnicity, rank, time in service, and deployment charac- teristics was administered to each participant.

Procedure Complete participation required approximately 2 hours of the volunteers’ time. The schedule of events is presented in Table I.

All interested participants attended an information session and completed a series of in-processing procedures. Next, par- ticipants were placed at the testing station, which was equipped with a laptop computer fitted with an external mouse. They were asked to complete the surveys on the laptop. Participants were then given instructions on the experimental prediction task, and were offered the opportunity to ask any questions. Once participants signaled that they understood the task instructions, they completed the experimental prediction task.

Data Analysis Corrected prediction ratings collected by the VAS (raw VAS scores minus 50) were the primary dependent variables for all statistical analyses. A combination of repeated measures and between-group analyses were used to determine (1) pat- terns and differences within subjects between the short-term prediction, long-term prediction, and control points; (2) impacts of anxiety, combat, and/or PTSD on within-subject predic- tion ratings; and (3) impacts of anxiety, combat, and/or PTSD on overall levels of white noise expectancy, regard- less of when VAS ratings were collected. All data analyses were completed using IBM’s Statistical Package for the Social Sciences (SPSS), version 19.

RESULTS Of those participants who completed the experimental pre- diction task (N = 40), 39 completed the PCL-M and CES,

TABLE II. Mean Survey Scores With Standard Error

Survey Mean Score SE Clinical Significance Cutoff Score

CES 10.13 1.17 Not Applicable PCL-M 20.44 1.16 50 SAI 27.5 1.17 40 TAI 30.05 0.65 40

Cases by Survey Score Groups

Survey No. of Participants in

High Score Group No. of Participants in

Low Score Group Median Score

CES 18 21 7 PCL-M 19 20 19 SAI 17 22 25 TAI 18 21 29

Mean and SD VAS Ratings by Survey Score Groups

Survey Group High Score Group Mean

VAS Rating (SE) Low Score Group Mean

VAS Rating (SE)

CES 7.49 (3.98) −7.50 (4.45) PCL-M −3.50 (4.36) 3.48 (4.08) SAI 0.43 (4.54) −0.44 (3.88) TAI −8.49 (4.42) 8.48 (4.01)

MILITARY MEDICINE, Vol. 182, July/August 2017e1790

Combat Experience and Attentiveness to Threat

Downloaded from publications.amsus.org: AMSUS - Association of Military Surgeons of the U.S. IP: on Jul 06, 2017.

Copyright (c) Association of Military Surgeons of the U.S. All rights reserved.

and all 40 completed the SAI and Trait Anxiety Inventory (TAI). CES scores ranged from 0 (light combat exposure) to 26 (moderate-heavy combat exposure), with a mean consis- tent with light to moderate combat exposure. None of the participants reported scores on any of the clinical surveys that met standards for clinical significance. Mean scores, standard error (SE) values, and clinical significance thresh- olds for each survey are included in Table II.

A repeated-measures ANOVA was conducted to assess the overall effect of the three prediction conditions on par- ticipants’ VAS ratings. A Mauchley’s test indicated that the VAS ratings did not meet the assumption of sphericity. Since the Greenhouse–Geisser ratio was below 0.75 (0.575), a Greenhouse–Geisser correction was used to adjust degrees of freedom for the within-subject analysis. A significant main effect for condition was identified within subjects [F(1.15, 31.07) = 10.56, p = 0.002]. This effect is presented in Figure 2. A test of within-subject contrasts demonstrated significance for both the linear [F(1, 27) = 8.89, p = 0.006] and quadratic [F(1, 27) = 23.69, p < 0.001] relationships, suggesting that the difference between the long-term predic- tion ratings and control ratings was significantly less robust than their respective differences from the short-term predic- tion ratings.

Within-Subjects Analyses Between Survey Score Groups To more clearly compare experimental task scores between participants, groups were created to delineate between high and low scores for each survey. Since there were no clini- cally significant cases included in the participant sample, high and low scores were sample specific, and were deter- mined on the basis of the median scores for each survey. High scores were above the median score; low scores were at or below the median score. A case breakdown of each group, including median score used for high/low cutoff, is included in Table II.

Survey score groups were included in the repeated- measures ANOVA to determine whether significant interac- tion effects of any of the survey measures and prediction conditions were present. After correcting for a lack of sphe- ricity using the Greenhouse–Geisser correction, it was deter-

mined that none of the survey score groups (CES, PCL-M, SAI, or TAI) had any significant interaction with the VAS ratings among the three prediction conditions. These find- ings indicate that the overall VAS rating patterns between prediction conditions (as illustrated in Fig. 2) were generally consistent for all survey score groups.

Between-Group Differences in Overall VAS Ratings A between-group analysis was also conducted to investigate differences in overall white noise expectancy between high- and low-scoring participants for all four surveys, regard- less of when the prediction was made during the sequence (pooling prediction ratings for control, long-term, and short- term conditions). Mean corrected VAS ratings (with SE values) for each survey score group are listed in Table II.

After assuring that the data met the assumptions for a parametric analysis, mean VAS ratings were compared for all groups with a 2 × 2 × 2 × 2 factorial ANOVA. There was a significant main effect for CES scores on VAS ratings [F(1, 27) = 5.19, p = 0.031], with high scorers demonstrat- ing generally higher expectancy of the white noise burst throughout the entire experimental sequence. This result sug- gest that an individual’s level of combat experience influ- ences the ways that they attend to and process threats.

DISCUSSION The results of this study support our hypothesis that, within subclinical populations of Soldiers, the intensity of one’s combat experience increases their tendency to exhibit an attention-bias toward threats. These findings suggest that Soldiers who experience combat should be observed for signs of increased threat attention because this may indicate that their information processing and decision-making capac- ities are being compromised in ways that may negatively impact their performance. This may also signal the develop- ment of more serious psychological conditions. Over the next section, we outline some implications of the findings we observe here.

The positive relationship we observe between an individ- ual’s level of combat experience and their attentional biases toward threatening stimuli is important because it may help to explain why these veterans exhibit a variety of related and often concerning emotions, attitudes and behaviors. For example, previous researchers have observed that the inten- sity of Soldiers’ combat experiences increase the extent to which they engage in “externalizing” behaviors that are risky, aggressive, violent, or even criminal in nature.23–26 In addition, the tendencies of combat veterans to exhibit threat- attention biases may also partially account for the higher- than-average rates of vocational challenges, marital conflicts, parenting problems, or antisocial behaviors such as vagrancy, indebtedness, and pathological lying that are reported among these populations of Soldiers.23,25,26

FIGURE 2. Mean and SE for corrected VAS by condition.

MILITARY MEDICINE, Vol. 182, July/August 2017 e1791

Combat Experience and Attentiveness to Threat

Downloaded from publications.amsus.org: AMSUS - Association of Military Surgeons of the U.S. IP: on Jul 06, 2017.

Copyright (c) Association of Military Surgeons of the U.S. All rights reserved.

As a potential, early indication of developing PTSD or other related conditions, evidence of increasing threat-attention biases should provide clinicians with important diagnostic information.23 This information can be used to initiate poten- tially therapeutic psychological interventions. For instance, clinicians who treat observed threat-attention biases as a component of a “stress response syndrome” may prevent combat veterans from manifesting feelings of alienation, guilt, anger, or frustration that often accompany more seri- ous psychological disorders.4,8,26 As an example, by training individuals to control how they direct their attention and reg- ulate their emotions,27,28 clinicians may be able to assist these individuals in regulating their hypervigilance to threat- ening stimuli in ways that prevent or mitigate potentially harmful changes in the prefrontal cortex, amygdala, and anterior cingulate cortex regions of the brain.4,5,7–9,11,12,16

An inherent challenge in attempting to identify and treat individuals who suffer with threat-attention bias is that they may tend to view clinicians and the prospect of seeking psychological treatments as salient threats that they should actively seek to avoid. Although this is beginning to change, Soldiers seeking treatments for manifest psychological dis- orders have traditionally felt stigmatized.29–31 As a result, individuals who are most in need of treatment for threat- attention biases may be more prone to internalize negative stereotypes and exhibit heightened concerns that they will be discriminated against for their mental health conditions. This when combined with the presence of comorbid psycho- logical conditions such as a compromised sense of their self- esteem or an enhanced feeling of their own invincibility may decrease their motivation to seek the type of mental health interventions that may most help them.24,29

While the prevention of psychological disorders and com- promised decision-making should motivate leaders to evalu- ate their Soldiers, it should also be noted that a preclinical increase in Soldiers’ attention to threats is not always prob- lematic. For many Soldiers, gaining experience in combat is a necessary and integral aspect of their role as Warfighters.30

For these individuals, a moderate increase in attention to environmental threats may equate to a higher level of situa- tional awareness, which can assist Soldiers both in general force protection and in performing combat-related jobs where vigilant monitoring of threats is required.32 Some preliminary evidence for this idea is presented in recent research by Akinola and Mendes.33 They find that police officers who maintain higher cortisol levels after a social-stress task per- form better in shoot/don’t shoot scenarios. Future studies, however, are needed to further examine the relationship between an individual’s exposure to high-stress situations, their threat-attention biases, and their task performance in both military and paramilitary contexts.

These observations highlight some limitations of this study and ideas for …

Attachment 2


JRRDJRRD Volume 53, Number 6, 2016Pages 781–796

The influence of physical and mental health symptoms on Veterans’ functional health status

Tong Sheng, PhD;1–2* J. Kaci Fairchild, PhD;1–2 Jennifer Y. Kong, MSW;3 Lisa M. Kinoshita, PhD;1 Jauhtai J. Cheng, MD;1–2 Jerome A. Yesavage, MD;1–2 Drew A. Helmer, MD, MS;4 Matthew J. Reinhard, PsyD;5 J. Wesson Ashford, MD, PhD;1–2 Maheen M. Adamson, PhD1–2,6 1War Related Illness and Injury Study Center, Department of Veterans Affairs (VA) Palo Alto Health Care System, Palo Alto, CA; 2Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA; 3School of Social Work, Boston University, Boston, MA; 4War Related Illness and Injury Study Center, VA New Jersey Health Care System, East Orange, NJ; 5War Related Illness and Injury Study Center, VA Medical Center, Washington DC; 6Defense and Veterans Brain Injury Center, VA Palo Alto Health Care System, Palo Alto, CA

Abstract—Veterans who have been deployed to combat often have complex medical histories, including some combination of traumatic brain injury (TBI); mental health problems; and other chronic, medically unexplained symptoms (i.e., chronic multisymptom illness [CMI] clusters). How these multiple pathologies relate to functional health is unclear. In the current study, 120 Veterans (across multiple combat cohorts) under- went comprehensive clinical evaluations and completed self- report assessments of mental health symptoms (Patient Health Questionnaire-2 [PHQ-2], Posttraumatic Stress Disorder Checklist–Civilian Version [PCL-C]) and functional health (Veterans Rand 36-Item Health Survey). Canonical correlation and regression modeling using split-sample permutation tests revealed that the PHQ-2/PCL-C composite variable (among TBI severity and number of problematic CMI clusters) was the primary predictor of multiple functional health domains. Two subscales, Bodily Pain and General Health, were associated with multiple predictors (TBI, PHQ-2/PCL-C, and CMI; and PHQ-2/PCL-C and CMI, respectively), demonstrating the mul- tifaceted nature of how distinct medical problems might uniquely and collectively impair aspects of functional health. Apart from these findings, however, TBI and CMI were not predictors of any other aspects of functional health. Taken together, our findings suggest that mental health problems might exert ubiquitous influence over multiple domains of functional health. Thus, screening of mental health problems and education and promotion of mental health resources can be important to the treatment and care of Veterans.

Key words: chronic multisymptom illness, daily functioning, functional health, mental health, posttraumatic stress, quality of life, symptoms, traumatic brain injury, Veterans, war-related illness.


Veterans who have been deployed to combat often exhibit many physical and mental health symptoms as a consequence of their experiences in service. However, how these multiple war-related illnesses affect Veterans’

Abbreviations: CMI = chronic multisymptom illness, OEF = Operation Enduring Freedom, OIF = Operation Iraqi Freedom, OND = Operation New Dawn, PCL-C = Posttraumatic Stress Disorder Checklist–Civilian Version, PHQ-2 = Patient Health Questionnaire-2, PTSD = posttraumatic stress disorder, TBI = traumatic brain injury, VA = Department of Veterans Affairs, VR-36 = Veterans RAND 36-Item Health Survey, WRIISC = War Related Illness and Injury Study Center. *Address all correspondence to Tong Sheng, PhD; 3801 Miranda Ave, Palo Alto, CA 94303; 650-493-5000, ext 67238. Email: [email protected] http://dx.doi.org/10.1682/JRRD.2015.07.0146


JRRD, Volume 53, Number 6, 2016

functional health and health-related quality of life is cur- rently not well understood. The California campus of the War Related Illness and Injury Study Center (WRIISC) at the Department of Veterans Affairs (VA) Palo Alto Health Care System is one of three WRIISC sites nationwide (including East Orange, New Jersey, and Washington, DC) that focus on the postdeployment health concerns of Veterans by providing consultation services as part of a national effort to provide clinical care to Veterans with complex health problems [1].

Some Veteran patients are eligible to be referred to the WRIISC program by their local VA providers to seek a second opinion on their conditions, such as those who have complex health conditions and medically unexplained symptoms, show little to no symptom improvement fol- lowing several tests and/or treatment or follow-up, or have possible service-related problems or concerns related to environmental exposures. (For referral and enrollment information, see http://www.warrelatedillness.va.gov/ WARRELATEDILLNESS/referral/.) As such, the WRIISC program evaluates Veterans who served in all combat eras from the Vietnam War to Operation Iraqi Freedom/Operation Enduring Freedom/Operation New Dawn (OIF/OEF/OND). Enrolled Veterans are brought to the designated service-area WRIISC site for a multiday assessment, during which they are evaluated by a special- ized team of clinicians, researchers, and educators who offer recommendations for further testing and treatment.

Typical WRIISC patients have a complex medical history with problems related to the chronic sequelae of traumatic brain injury (TBI), mental health disorders (e.g., depression, posttraumatic stress disorder [PTSD]), and other chronic medical conditions that involve multi- ple symptoms without clear explanations. These chronic, medically unexplained cases span multiple domains (i.e., clusters), including fatigue; pain; problems in pulmonary, dermatologic, and gastrointestinal systems; and problems with sleep, mood, cognition, and memory [2–3]. These cases, collectively described as chronic multisymptom illness (CMI), are frequently observed among Veterans who have served in the Gulf War [4]. However, in more recent reports, CMI-like cases are also described in Vet- erans deployed to other conflicts (e.g., OIF/OEF/OND) as well as in nondeployed Veterans [5–6].

WRIISC patients are more likely to exhibit TBI, mental health problems, and problems in CMI symptom clusters at higher rates than are epidemiological Veteran samples because patients are referred to the WRIISC pro- gram specifically because they presented chronic and

complex medical problems. In representative Veteran samples, TBI is observed in an estimated 10 to 30 percent of Veterans [7–8]. Similar proportions are reported for depression (14%; National Alliance on Mental Illness; http://www.nami.org) and PTSD (4–20% in Gulf War-era and OIF/OEF/OND Veteran samples) [9–11]. Estimated numbers of Veterans affected by CMI are more variable because of multiple working case definitions. Nonethe- less, an estimated 8 to 62 percent of Gulf War-era Veteran samples are thought to be affected by CMI problems [3– 4]. The higher rates of TBI, mental health problems, CMI problems, and comorbidities in WRIISC patients intro- duce an analytical challenge and potentially obfuscate comparisons with studies involving other Veteran cohorts (e.g., the Vietnam War, Gulf War, and OIF/OEF/OND conflict eras [12–16]). However, such patients provide a unique opportunity to gain insights into the potential effect that multiple medical problems and comorbidities might have on other health outcomes.

Because postdeployed Veterans typically experience some combination of TBI, mental health, and CMI symp- toms, a better understanding of how multiple medical fac- tors are potentially associated with impaired functional health and health-related quality of life may inform more effective treatment approaches [16–21]. Whereas contem- porary medical practice generally treats individual medi- cal conditions independently from comorbidities, different pathologies can have overlapping or compound- ing effects. Thus, a more integrated treatment approach addressing multiple aspects of a Veteran’s health might be more beneficial. In the current study, we aimed to charac- terize the extent to which commonly observed war-related pathologies (i.e., TBI, depression, PTSD, and problems in CMI clusters) were associated with various aspects of functional health and health-related quality of life in Vet- erans evaluated at the California WRIISC program.

We expected different domains of functional health to be uniquely associated with different medical factors. For example, TBI and problems in CMI symptom clus- ters were expected to be more strongly associated with physical, bodily, and pain-related impairments (e.g., lim- ited ability in moderate and intense physical activities, difficulty in performing work, limitations as a result of severe pain), whereas mental health symptoms related to depression and PTSD might be more strongly associated with psychological, emotional, and social impairments (e.g., limitations from feeling low energy and anxious, limitations in social activities). However, the relation- ships between medical factors and functional health


SHENG et al. Multiple medical problems and functioning

might be substantially more complex, with multiple med- ical factors (TBI, depression, PTSD, problems in CMI clusters) influencing various aspects of functional health. Thus, the current investigation represented a preliminary effort toward elucidating these relationships in a Veteran sample characterized by multiple comorbid pathologies.


Sample Participants in this study (N = 120; mean age, 47.8 yr,

range 27–78 yr; 15 females) were Veterans seen at the California WRIISC. They are a sample of patients evalu- ated at the WRIISC program who met the following cri- teria: (1) completed a self-report questionnaire packet assessing demographic and health-related information, (2) underwent the comprehensive on-site clinical evalua- tion, and (3) gave written consent to making their data available for research purposes.

The Veterans included in the study were part of a mixed cohort sample, the majority of whom had served in combat. Most Veterans had been deployed to the Gulf War (46%), but a substantial proportion had been deployed to OIF/OEF/OND (38%). Smaller groups were deployed to the Vietnam War (14%) and other conflicts (6%). Fifteen percent have been deployed to multiple combat theaters, and 6 percent have never been deployed. This small number of nondeployed Veterans demon- strates the fact that chronic illnesses and injuries can also be incurred from training-related activities and exposures outside of combat. Although the WRIISC programs typi- cally treat combat Veterans, the programs also receive and enroll referrals involving noncombat Veterans who present persistent complex symptoms. For sample char- acteristics, see Table 1 and the Figure.

Clinical Assessments and Determination of Clinical Conditions

Neurological Examination for Assessment of Traumatic Brain Injury

All patients included in the study underwent a complete neurological examination by a neurologist for the assess- ment of TBI. Each patient was examined

Demographic* Description

Age, yr (range) 47.8 (27–78) Education, yr (mean ± SD) 14.3 ± 3.5 Sex, Female, % 12.5 Combat History/Theater, %

14.2 45.8 38.3 5.8

No. Deployments, % 5.8 79.2 15.0

Medical History, % 57.5 42.5 52.5 5.0 59.2 65.0

Pain 95.8 Sleep 85.8 Gastrointestinal 60.0 Cognitive NOS 55.8

49.2 34.3 24.2

Problematic CMI Clusters, No. (cumulative %) 0 3.3 (100.0) 1 1.7 (96.7) 2 5.0 (95.0) 3 23.3 (90.0) 4 29.2 (66.7) 5 21.7 (37.5) 6 12.5 (15.8) 7 3.3 (3.3)

and diagnosed (with determination of severity) in accordance with the guidelines set forth by the VA and the Department of Defense [22–23]. A TBI was defined as any period of loss or

Table 1. Complete demographic and medical history information for the Veterans evaluated in the California WRIISC program (N = 120).

Vietnam GW1 OIF/OEF/OND Other

0 1 2+

TBI Diagnosis†

None Mild Moderate/Severe

Probable Depression‡

PTSD Diagnosis†

Affected CMI Clusters, %§

Fatigue Dermatologic Pulmonary

*Demographic and combat history information were collected from Veterans during intake into the California WRIISC program. †Diagnoses of TBI and PTSD were made by the staff neurologist, clinical psy- chologist, and psychiatrist. ‡Likelihood of depression was based on the self-report screening measurement Patient Health Questionnaire-2, using a cutoff of 3. §Diagnoses of problems in CMI symptom clusters were made by providers at other various specialty clinics. CMI = chronic multisymptom illness, GW1 = Gulf War 1, No. = number, NOS = not otherwise specified, OEF = Operation Enduring Freedom, OIF = Operation Iraqi Freedom, OND = Operation New Dawn, PTSD = posttraumatic stress dis- order, SD = standard deviation, TBI = traumatic brain injury, WRIISC = War Related Illness and Injury Study Center.

Table 1. Complete demographic and medical history information for the Veterans evaluated in the California WRIISC program (N = 120).


JRRD, Volume 53, Number 6, 2016

decreased level of consciousness, any loss of memory imme- diately prior to or following injury, any alteration in mental state (e.g., confusion, disorientation, slowed thinking), neurological deficits

Figure. War Related Illness and Injury Study Center (WRIISC) at the Department of Veterans Affairs (VA) Palo Alto Health Care System pro-

vides consultation for patients with complex medical problems. Common among these Veterans are traumatic brain injury (TBI), post-

traumatic stress disorder (PTSD), probable depression, and problems in chronic multisymptom illness (CMI) clusters. Shown are the

percentages of individuals seen at the VA Palo Alto WRIISC reporting symptoms in CMI clusters and mental health and TBI problem

areas. GI = gastrointestinal.

(e.g., weakness, loss of balance, loss in vision or other senses, aphasia), or intracranial lesion as the result of an external force. Although Veterans may have possibly (or even likely) sustained multiple TBIs in the past, the neurologist made all diagnoses based on one self- reported incident at the patient’s disclosure. Accurate and reliable diagnosis of TBI was limited because it relied on self-reporting without consistently available supporting doc- umentation or medical information at the time of injury.

However, self-reporting and sparse medical records were the only information sources available [24]. Based on the retrospective self-report (corroborated with a chart review of past medical records in the VA’s centralized charting sys- tem prior to the interview), patients were diagnosed as hav- ing no TBI, mild TBI (normal brain scan, loss of consciousness <30 min, altered mental state <24 h, posttrau- matic amnesia <1 d), or moderate-severe TBI (normal or abnormal brain scan, loss of consciousness >30 min, altered mental state >24 h, posttraumatic amnesia >1 d).


SHENG et al. Multiple medical problems and functioning

Assessment of Chronic Multisymptom Illness Veterans’ CMI symptom profiles were characterized

by noting active problems within individual CMI clusters. A chart review of patients’ on-site evaluations was per- formed via the VA’s Computerized Patient Record System, and current diagnoses in the following clusters were noted: pain, fatigue, dermatologic, gastrointestinal, pulmo- nary, sleep, and cognitive disorders not otherwise speci- fied. Only diagnoses charted during the patients’ WRIISC evaluations were included in this study. A CMI symptom cluster was considered problematic if one or more medical conditions within the cluster were noted by a provider. Cognitive disorders not otherwise specified were diag- nosed by consensus of the psychiatrist, neurologist, and clinical psychologist. The total number of unique prob- lematic CMI clusters was tallied for each patient.

Self-Report Measures

Probable Depression Patients completed the Patient Health Questionnaire-

2 (PHQ-2), a brief, two-item measure that screens for symptoms of depression. It was shown to have acceptable sensitivity (83%–100%) and specificity (77%–92%) for major depression in civilian and Veteran samples and is routinely used in the VA as part of the Primary Care Manual [25–27].

Posttraumatic Stress Patients completed the PTSD Checklist–Civilian

Version (PCL-C), a 17-item self-report measure that assesses severity of symptoms resulting from stressful life experiences (not restricted to military-specific trauma) across different symptom domains of PTSD (re- experiencing, avoidance, and hyper arousal; criteria from the Diagnostic and Statistical Manual of Mental Disor- ders-4th Edition) [28–29]. The sum of all 17 responses was used as an indicator of global PTSD symptom sever- ity. Psychometric properties of the PCL-C have been reported as favorable, with good test-retest reliability (>0.75), internal consistency (>0.83), and strong correla- tion with clinical standards [29–30].

Functional Health Outcomes The Veterans RAND 36-Item Health Survey (VR-36)

was administered to assess constructs affecting health- related quality of life. The VR-36 is a 36-item, Veteran- specific inventory based on the functional health measure

36-Item Short Form Survey [31–33]. The VR-36 assesses different aspects of functioning and well-being across eight subscales (domains): Physical Functioning, Role Limitations: Physical, Bodily Pain, General Health, Vitality, Social Functioning, Role Limitations: Emo- tional, and Mental Health. Descriptions and sample items of each subscale are presented in Table 2. These Likert- style items were computed into summary scores for each domain (scaled 0–100), with greater scores indicating better functioning. Internal reliability is generally good, with seven of the eight individual domains having Cron- bach alpha between 0.76 and 0.91 [34].

Predictors of Functional Health Outcomes To quantify the associations between medical factors

(TBI severity, PHQ-2, PCL-C, and number of problem- atic CMI clusters) and functional health measures (the eight VR-36 subscales), we performed primary analyses to observe the overall multivariate relationship between medical factors and functional health measures, followed by confirmatory regression modeling and cross-validation using split-sample permutation tests.

Data Transformation Prior to data analyses, we inspected all self-report

measures for outliers. Outlier detection was performed using the Tukey procedure on the PHQ-2, PCL-C, and VR-36. Outliers were defined as >1.5 times the interquar- tile range beyond the first and third quartiles. No outliers were identified based on the PHQ-2 and PCL-C, but four patients were flagged as potential outliers among the eight VR-36 subscales. On further inspection, however, we found that distributions of many of the VR-36 sub- scales were positively skewed (overall tendency to report lower scores on nearly all subscales; Table 2. Because our sample was expected to exhibit greater disability in functional health, rather than exclude the flagged patients with lower subscale scores as outliers, we transformed all VR-36 subscales scores using Box-Cox transformations to give each subscale a more normal distribution.

In addition, because we expected depression and PTSD to be correlated, we combined the PHQ-2 and PCL- C scores to form a mental health composite (by summing the Z-transforms of the two) [35–36]. Inclusion of collinear terms in regression modeling would have negatively influ- enced interpretation of results and reduced statistical power. Thus, although we were unable to distinguish the relative contributions of depression and PTSD symptoms

Measurement Description Sample Items Mean ± Standard


PHQ-2 (2 items) Frequency of depressed mood and anhedonia over the past 2 wk

Little interest or pleasure in doing things? Feeling down, depressed, or hopeless?

3.2 ± 2.1

PCL-C (17 items) DSM-IV symptoms of PTSD Repeated, disturbing memories, thoughts, or images of a stressful experience from the past?

Avoid activities or situations because they remind you of a stressful experience from the past?

Feeling irritable or having angry outbursts?

52.5 ± 17.9

VR-36 Health-related quality of life, disease burden, and effect on daily functioning

— —

Physical Functioning (9 items)

Limitations in range of physical activities

Moderate activities, such as moving a table, pushing a vacuum cleaner, bowling, or playing golf?

Walking several blocks?

41.7 ± 26.4

Role Limitations: Physical (5 items)

Limitations in performing daily activities because of physical problems

Had difficulty performing the work or other activities (for example, it took extra effort)?

Were limited in the kind of work or other activities?

24.9 ± 25.7

Bodily Pain (2 items)

Intensity of pain and extent to which pain interferes with daily activities

During the past 4 wk, how much did pain interfere with your normal work (including both work outside the home and housework)?

26.3± 22.9

General Health (4 items)

Perception of general health In general, would you say your health is [poor–excellent]? Compared to one year ago, how would you rate your health

in general now?

29.1 ± 19.0

Vitality (4 items)

Subjective rating of energy and fatigue

Did you have a lot of energy? Did you feel worn out?

20.3 ± 19.9

Social Functioning (1 item)

Limitations in social activities During the past 4 wk, to what extent have your physical health or emotional problems interfered with your nor- mal social activities with family, friends, neighbors, or groups?

26.6 ± 22.9

Role Limitations: Emotional (3 items)

Limitations in performing daily activities due to emotional problems

Didn’t do work or other activities as carefully as usual? Accomplished less than you would like?

48.8 ± 33.5

Mental Health (5 items)

Subjective rating of mental health and emotional well-being

Have you been a very nervous person? Have you felt so down in the dumps that nothing could

cheer you up?

47.9 ± 24.6


JRRD, Volume 53, Number 6, 2016

to functional health, the use of a PHQ-2/PCL-C composite variable allowed us to investigate the potential effects of a broader mental health construct.

Primary Analyses Our primary objective was to describe potential rela-

tionships between medical factors and multiple func- tional health domains, so we performed a series of multivariate and multiple regression analyses.

Canonical correlation analysis. To gain an appreci- ation of the overall relationship between medical factors and functional health measures, we performed a canoni- cal correlation analysis with TBI severity, PHQ-2/PCL-C composite, and number of problematic CMI clusters as predictor variables, and the eight subscales of the VR-36 as outcome variables.

Table 2. Descriptions of self-report measures an overall sample mean ± standard deviation (n = 120).

Note: Raw VR-36 scores are described prior to Box-Cox transformation. DSM–IV = Diagnostic and Statistical Manual of Mental Disorders-4th Edition, PCL-C = Posttraumatic Stress Disorder Checklist–Civilian Version, PHQ-2 = Patient Health Questionnaire-2, PTSD = posttraumatic stress disorder, VR-36 = Veterans RAND 36-Item Health Survey.


SHENG et al. Multiple medical problems and functioning

Split-sample permutation testing and cross- validation of multiple regression models. To quantify specific associations between medical factors and functional health measures (especially those indicated by the canonical correlation analysis), we tested standard regression models using split-sample permutation tests and cross-validation. Permutation tests using split samples offered several ana- lytic advantages: precise estimates of effect sizes and inter- quartile ranges, insights into the overall distributions of effects, and metrics of generalizability based on cross- validations between the tested (discovery) and withheld (generalizability) subsamples.

The sample of 120 subjects was split into two ran- domly selected groups of 60 across 10,000 iterations. During each iterative process, for each VR-36 subscale, a standard linear regression model with medical factors and demographic variables (age and education) as predictor variables and the VR-36 subscale scores as the outcome variables was tested using the discovery subsample. Regression coefficients from the discovery model were then applied to the generalizability subsample data to obtain predicted VR-36 subscale scores. Cross-validation reliability was quantified as a Pearson correlation between the predicted and actual subscale scores in the generaliz- ability subsample. After 10,000 iterations, distributions of the overall model fit (adjusted R2) and its cross-validation (R) were obtained for each VR-36 subscale.

Contribution of individual medical factors to func- tional outcomes. Using permutation tests, we also com- puted distributions of estimated effect sizes (beta, t- statistic) for each regression coefficient (i.e., for each medi- cal factor, accounting for age and education as covariates), for each VR-36 subscale. By examining these distributions, we inferred the relative importance of each medical factor to each functional health domain.

Secondary Analyses The medical factor variables in the primary analyses

varied in form, so we ruled out potential effects of vari- able structure on the medical factor/functional outcome relationships by performing the same analyses on binary forms of the medical factor variables. Medical factors were binarized as follows: TBI severity (no TBI versus mild TBI; moderate-severe TBI cases were excluded from this analysis because of potential construct hetero- geneity if combined with mild TBI cases), PHQ-2 (using a cutoff of 3+), PCL-C (using a cutoff of 55+), and CMI variables (using a median split of 4+ symptom clusters)

[25,27,37–38]. The cutoff points for the PHQ-2 and PCL- C were chosen because the WRIISC Veteran sample had more complex medical histories than those reported in typical Veteran samples.

We also performed post hoc zero-order correlation analyses between each medical factor and each VR-36 subscale to observe relative associations without regress- ing out the potential influences of the other medical fac- tors or demographic variables.

Although we considered depression and PTSD symp- toms jointly in the primary analyses because of potential collinearity issues, we also carried out the same canonical correlation and split-sample permutation modeling analy- ses while considering PHQ-2 and PCL-C independently.

We had hoped to gain a more nuanced understanding of potential medical factor/functional health domain rela- tionships through these secondary analyses, but because these analyses are exploratory we have refrained from dis- cussing them in detail in this article.


Medical History and Self-Report Measures More than half of California WRIISC Veterans met

diagnostic criteria for mild (52.5%) or moderate TBI (5%). An even larger proportion met diagnostic criteria for PTSD (65%). Nearly all (97%) reported problems from at least one CMI symptom cluster; two-thirds had problems in four or more unique CMI symptom clusters. Complete demographic and medical history information are shown in Table 1.

Mean response on the PHQ-2 was 3.2, and the mean response on the PCL-C was 52.5. Mean raw scores on VR-36 subscales ranged from 20.3 to 48.8. Descriptive data for all self-report measures are shown in Table 2.

Primary Analyses

Canonical Correlation Analysis Results A significant relationship was observed between the

first canonical variate pair (Wilks lambda = 0.24, F = 8.43, p < 0.001). The pooled squared canonical correlation for the first canonical pair was 0.88. In the first canonical root, the PHQ-2/PCL-C composite was the most important (standardized canonical coefficient = 0.53) among predic- tor variables, and Mental Health (standardized canonical coefficient = 0.61), Role Limitations: Emotional (stan-


JRRD, Volume 53, Number 6, 2016

dardized canonical coefficient = 0.37), and Social Func- tioning (standardized canonical coefficient = 0.20) were the most important among outcome variables. All canoni- cal correlation results are presented in Table 3.

Split-Sample Permutation Tests and Cross-Validation Split-sample permutation tests of regression models

predicting VR-36 subscale scores indicated that the Men- tal Health, Role Limitations: Emotional, and Social Functioning subscales were substantially accounted for by the medical factors and covariates (mean adjusted R2 = 0.61, 0.52, 0.26, respectively). The Bodily Pain, Vitality, and General Health subscales were also well modeled by medical factors and covariates to lesser extents (mean adjusted R2 = 0.19, 0.14, 0.13, respectively). The Physical Functioning and Role Limitations: Physical subscales were not well explained by the medical factors included in the analyses. All split-sample permutation test results and 95 percent interquartile ranges are shown in Table 4.

The Mental Health, Role Limitations: Emotional, and Social Functioning subscales also had the highest cross- validation reliability (adjusted mean R = 0.78, 0.71, 0.48, respectively). The Bodily Pain, Vitality, General Health, and Physical Functioning

Variables Canonical Roots

1 2 3 Independent

0.016 0.772 0.675 0.533 0.173 0.014 0.069 0.611 0.808

Dependent 0.070 0.007 0.445 0.607 …

Attachment 3

Psychological Trauma: Theory, Research, Practice, and Policy Coping Strategy Utilization Among Posttraumatic Stress Disorder Symptom Severity and Substance Use Co- Occurrence Typologies: A Latent Class Analysis Nathan T. Kearns, Ateka A. Contractor, Nicole H. Weiss, and Heidemarie Blumenthal Online First Publication, September 7, 2020. http://dx.doi.org/10.1037/tra0000964

CITATION Kearns, N. T., Contractor, A. A., Weiss, N. H., & Blumenthal, H. (2020, September 7). Coping Strategy Utilization Among Posttraumatic Stress Disorder Symptom Severity and Substance Use Co-Occurrence Typologies: A Latent Class Analysis. Psychological Trauma: Theory, Research, Practice, and Policy. Advance online publication. http://dx.doi.org/10.1037/tra0000964

Coping Strategy Utilization Among Posttraumatic Stress Disorder Symptom Severity and Substance Use Co-Occurrence Typologies:

A Latent Class Analysis

Nathan T. Kearns Brown University and University of North Texas

Ateka A. Contractor University of North Texas

Nicole H. Weiss University of Rhode Island

Heidemarie Blumenthal University of North Texas

Objective: There is a lack of research on primary prevention of posttraumatic stress disorder (PTSD) symptoms and substance use among trauma-exposed populations. To guide the development of more effective prevention efforts, the current study sought to identify underlying coping mechanisms that impact PTSD–substance use co-occurrence. Method: A person-centered analytic approach (latent class analysis) examined PTSD–substance use co-occurrence typologies (classes) and identified theoretically adaptive (e.g., active coping) and maladaptive (e.g., denial) coping strategies that differentiated between classes among a sample of 1,270 trauma-exposed participants (Mage � 20.71, 73.5% female, 45.7% White). Results: Latent class analysis identified five distinct typologies, reflective of extant epidemio- logical and etiological work. Generally, behavioral disengagement and self-blame coping increased the likelihood of being in more severe PTSD–illicit substance use (e.g., cocaine) comorbidity classes. Positive reframing and planning differentiated between low and moderate illicit substance typologies with moderate PTSD severity. Venting, acceptance, and self-distraction differentiated between asymp- tomatic and moderate PTSD severity typologies with low illicit substance use. Conclusions: Findings identify general coping strategies associated with increased likelihood of being in more severe comor- bidity typologies, as well as several unique coping strategies associated with risk of transitioning between low/moderate PTSD and illicit substance use classes. Relevant interventions (e.g., trauma psychoedu- cation, guilt-reduction therapy, psychological first aid) that may be targets for future prevention-oriented work are discussed.

Clinical Impact Statement This study indicates that there are distinct PTSD–substance use co-occurrence typologies that utilize unique coping strategies for distress. Identification of these differentiating strategies may facilitate the development of more effective prevention efforts.

Keywords: trauma, prevention, substance use, coping, posttraumatic stress

Growing evidence indicates robust associations between posttraumatic stress disorder (PTSD) symptoms and substance use (Debell et al., 2014; Jacobsen, Southwick, & Kosten, 2001; Kearns et al., 2018). Indeed, PTSD and substance use disorder

(SUD) comorbidity is particularly high in both clinical and nonclinical populations, with national epidemiological studies indicating that upward of 46.4% of individuals with PTSD also meet criteria for a SUD (Pietrzak, Goldstein, Southwick, &

X Nathan T. Kearns, Center for Alcohol and Addiction Studies, Brown University, and Department of Psychology, University of North Texas; Ateka A. Contractor, Department of Psychology, University of North Texas; Nicole H. Weiss, Department of Psychology, University of Rhode Island; Heidemarie Blumenthal, Department of Psychology, University of North Texas.

Work on this article by Nathan T. Kearns was supported by the National Institute on Alcohol Abuse and Alcoholism Grant

F31AA027142 and National Institute on Drug Abuse Grant T32DA016184, Nicole H. Weiss was supported by National Institute on Drug Abuse Grant K23DA039327, and Heidemarie Blumenthal was supported by National Institute on Alcohol Abuse and Alcoholism Grant R15AA026079.

Correspondence concerning this article should be addressed to Nathan T. Kearns, Center for Alcohol and Addiction Studies, Brown University, 121 South Main Street, Providence, RI 02912. Contact: E-mail: [email protected] gmail.com

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl


Psychological Trauma: Theory, Research, Practice, and Policy

© 2020 American Psychological Association 2020, Vol. 2, No. 999, 000 ISSN: 1942-9681 http://dx.doi.org/10.1037/tra0000964


Grant, 2011). This high co-occurrence rate is problematic, with extant work indicating that individuals with concurrent PTSD– SUD report elevated PTSD severity, greater psychiatric comor- bidities, and worse treatment outcomes than individuals with PTSD alone (McCauley, Killeen, Gros, Brady, & Back, 2012; Read, Brown, & Kahler, 2004).

To ameliorate these problematic outcomes, empirically based prac- tice guidelines for patients with comorbid PTSD–SUD have been established, recommending integrated treatments that incorporate el- ements of cognitive– behavioral therapy, motivational interviewing, and/or exposure therapies for PTSD (McCauley et al., 2012). The development of these treatments has been partially driven by extant clinical research emphasizing the utility of bolstering adaptive coping strategies (Najavits, 2002) to replace maladaptive coping strategies found to be associated with PTSD–SUD comorbidity (Ford & Russo, 2006). However, despite a burgeoning literature focused on identify- ing, understanding, and addressing coping strategies employed by PTSD–SUD patients in clinical settings, little work has evaluated coping mechanisms underlying PTSD and substance use co- occurrence in nonclinical populations—individuals not currently seeking treatment and with no history of treatment for PTSD or SUD.

Importantly, the limited existing literature focused on identification of coping strategy utilization for PTSD–substance use in nonclinical populations has generally indicated that adaptive coping strategies— such as social support (e.g., Bryant-Davis et al., 2015), acceptance (e.g., Kearns, Jackson, Elliott, Ryan, & Armstrong, 2018; Vujanovic, Bonn-Miller, & Marlatt, 2011), and cognitive reframing (e.g., Brief, Rubin, Enggasser, Roy, & Keane, 2011)—are associated with lesser PTSD symptom severity and lesser substance use; conversely, mal- adaptive coping strategies—such as avoidance (e.g., Bordieri, Tull, McDermott, & Gratz, 2014), self-blame (e.g., Startup, Makgek- genene, & Webster, 2007), and self-distraction (e.g., Hruska, Fallon, Spoonster, Sledjeski, & Delahanty, 2011; Kearns et al., 2018)— have been associated with more severe posttraumatic stress and substance use. However, reliance on this research base may be problematic due to at least two notable methodological limitations. More specifically, extant work has generally (a) examined potential coping strategies in isolation (i.e., focusing on a singular construct), disallowing for com- prehensive comparative evaluation of the impacts and magnitudes of influence of disparate adaptive and maladaptive coping strategies, and/or (b) presumed that a particular construct was being utilized as a coping strategy, without explicitly evaluating that construct as a means of coping with posttraumatic stress or problematic substance use (e.g., assuming a general measure of perceived social support was indicative of the frequency with which an individual utilizes that social support as a means of coping).

Appropriately identifying and comprehensively evaluating these adaptive and maladaptive coping mechanisms in nonclini- cal populations may be particular important for two reasons. First and foremost, the prevention-oriented interventions cur- rently being employed following trauma have not been effective (see Roberts et al., 2019 for review). For example, a meta- analysis indicated that psychological debriefing—a popular technique for managing psychological distress following trau- ma— did not prevent the onset of PTSD, nor did it reduce general psychological morbidity, depression, or anxiety (Rose, Bisson, Churchill, & Wessely, 2002). Second, despite the inef- fectiveness of these interventions, there is a lack of research on primary prevention of PTSD (see Skeffington, Rees, & Kane,

2013 for review) and broadly applicable substance use preven- tion strategies (e.g., environmental management; DeJong & Langford, 2002) and, subsequently, no work evaluating preven- tative interventions for PTSD–substance use co-occurrence fol- lowing trauma. As such, there is little information available to justify or guide the development of more effective prevention efforts to replace current nonefficacious interventions.

Additionally, although increasingly more common in re- search independently evaluating PTSD (e.g., Contractor, Roley- Roberts, Lagdon, & Armour, 2017) and substance use (e.g., Connor, Gullo, White, & Kelly, 2014), few studies evaluating both PTSD and substance use have acknowledged heterogeneity in the general population via use of appropriate person-centered statistical approaches (e.g., latent class analysis [LCA]; Ander- son, Hruska, Boros, Richardson, & Delahanty, 2018). These person-centered statistical approaches provide several advan- tages over standard analytic techniques (e.g., regression mod- eling), which typically assess broad, linear associations be- tween PTSD symptoms and substance use; alternatively, LCA first identifies distinct subgroups—termed typologies or class- es—that exist within the sample based on comorbid response patterns, then compares those typologies based on relevant health outcomes and/or population characteristics, such as cop- ing strategy utilization (Kline, 2011; McCutcheon, 1987). Given the lack of research aimed at understanding PTSD symptom–substance use co-occurrence in nonclinical popula- tions, identification of such typologies will be a critically important step in pinpointing coping mechanisms underlying or impacting PTSD–substance use comorbidity patterns within this trauma population, which, in turn, will provide information that may aid in the development of effective prevention efforts.

Addressing the aforementioned limitations, the current study aims to comprehensively examined coping strategy utilization among PTSD–substance use co-occurrence typologies in a large, nonclinical sample of trauma-exposed individuals using a recommended person-centered statistical approach. More spe- cifically, the current study examined the optimal latent class- solution based on endorsed PTSD symptom severity and type of substances used (e.g., alcohol, marijuana), then evaluated the extent to which 14 unique coping strategies differentiated be- tween identified PTSD–substance use classes. Given previous LCA findings regarding PTSD comorbidities (Anderson et al., 2018; Contractor et al., 2017), as well as strong associations between PTSD symptom severity and substance use (Debell et al., 2014; Jacobsen et al., 2001), it was hypothesized that the current LCA analyses would, at minimum, produce three par- allel low/medium/high PTSD–substance use co-occurrence classes. Further, given the focus on developing adaptive coping mechanisms in many empirically supported integrated treat- ments for PTSD–SUD (e.g., Seeking Safety; Najavits, 2002), it was generally hypothesized that theoretically adaptive coping strategies (e.g., active coping, social support, positive refram- ing) would be associated with less severe PTSD–substance use typologies, whereas theoretically maladaptive strategies (e.g., self-blame, denial) would be associated with more severe classes.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl




Participants and Procedure

The current study sample comprised 1,270 undergraduates (Mage � 20.71, 73.5% female; 45.7% White) attending a large university in the southwestern United States. Data were drawn from a larger study evaluating psychological well-being and sub- stance use from December 2016 through January 2019. Partici- pants were recruited via the university online research study pool and completed an online questionnaire battery via Qualtrics—an online data management software that complies with Health In- surance Portability and Accountability Act and Family Educa- tional Rights and Privacy Act regulations. Eligibility criteria in- cluded (a) being above the age of 18 and (b) experiencing at least

one Diagnostic and Statistical Manual of Mental Disorders (DSM)–5-defined PTSD Criterion A traumatic event (American Psychiatric Association, 2013) measured by the Life Events Checklist for DSM–5 (LEC-5; Weathers et al., 2013a). Participants were compensated via course credit. All procedures were approved by the institutional review board at the University of North Texas. See Table 1 for full demographic characteristics of the sample.


Traumatic experiences. The LEC-5 (Weathers et al., 2013a) assessed the presence of DSM–5 Criterion A traumatic events (American Psychiatric Association, 2013). The LEC-5 consists of 16 specified traumatic events (e.g., physical and sexual assault) and an option for an unspecified traumatic event. Only participants

Table 1 Demographics and Class Comparisons for Full Sample and Five Classes

Variable Full sample (N � 1,270)

Class 1 (n � 546)

Class 2 (n � 139)

Class 3 (n � 285)

Class 4 (n � 210)

Class 5 (n � 90)

Agea 20.71 � 3.19 20.35 � 2.43 22.20 � 4.27 20.54 � 3.05 20.74 � 3.35 21.09 � 4.48 Biological sex (Female)a 868 (73.5%) 379 (69.4%) 91 (65.5%) 230 (80.7%) 161 (76.7%) 73 (81.1%) Race/Ethnicityb

Asian 99 (7.8%) 46 (8.4%) 6 (4.3%) 24 (8.4%) 21 (10.0%) 2 (2.2%) African American 175 (13.8%) 80 (14.7%) 11 (7.9%) 40 (14.0%) 33 (15.7%) 11 (12.2%) White/Caucasian 580 (45.7%) 246 (45.1%) 75 (54.0%) 123 (43.2%) 92 (43.8%) 44 (48.9%) Hispanic/Latino 253 (19.9%) 114 (20.9%) 27 (19.4%) 61 (21.4%) 32 (15.2%) 19 (21.1%) Other 17 (1.3%) 11 (2.0%) 1 (0.7%) 1 (0.4%) 2 (1.0%) 2 (2.2%) Multiracial 144 (11.3%) 48 (8.8%) 19 (13.7%) 36 (12.6%) 29 (13.8%) 12 (13.3%)

Socioeconomic statusa

Less than $25,000 169 (13.3%) 68 (12.5%) 13 (9.4%) 33 (11.6%) 39 (18.6%) 16 (17.8%) $25,000 to $50,000 333 (26.2%) ‘145 (26.6%) 35 (25.2%) 80 (28.1%) 49 (23.3%) 24 (26.7%) $50,000 to $75,000 294 (23.1%) 134 (24.5%) 30 (21.6%) 61 (21.4%) 48 (22.9%) 21 (23.3%) More than $75,000 473 (37.2%) 199 (36.6%) 61 (43.9%) 111 (38.9%) 73 (34.8%) 29 (32.2%)

Substance use endorsement Alcohol 1,020 (80.3%) 406 (74.9%) 138 (99.3%) 229 (80.34%) 172 (81.9%) 72 (80.0%) AmED 547 (43.1%) 177 (32.4%) 115 (82.7%) 112 (39.3%) 100 (47.6%) 43 (47.8%) Cannabis 693 (54.6%) 237 (43.4%) 137 (98.6%) 142 (49.8%) 126 (60.0%) 51 (56.7%) Cocaine 157 (12.4%) 1 (0.2%) 80 (57.6%) 21 (7.4%) 38 (18.1%) 17 (18.9%) Prescription stimulants 171 (13.5%) 3 (0.5%) 81 (58.3%) 31 (10.9%) 37 (17.6%) 19 (21.1%) Schedule I/II hallucinogens 182 (14.3%) 2 (0.4%) 100 (71.9%) 22 (7.7%) 42 (20.0%) 16 (17.8%) Schedule III hallucinogens 144 (11.3%) 4 (0.7%) 71 (51.1%) 17 (6.0%) 40 (19.0%) 12 (13.3%)

Trauma type (worst)c

Natural disaster 67 (5.3%) 52 (9.5%) 1 (0.7%) 5 (1.8%) 8 (3.8%) 1 (1.1%) Fire or explosion 22 (1.7%) 14 (2.6%) 5 (3.6%) 2 (0.7%) 1 (0.5%) — Transportation accident 217 (17.1%) 140 (25.6%) 26 (18.7%) 39 (13.7%) 10 (4.8%) 2 (2.2%) Serious accident during activity 42 (3.3%) 29 (5.3%) 2 (1.4%) 7 (2.5%) 4 (1.9%) — Exposure to toxic substance 3 (0.2%) 1 (0.2%) — — 1 (0.5%) 1 (1.1%) Physical assault 83 (6.5%) 29 (5.3%) 9 (6.5%) 19 (6.7%) 17 (8.1%) 9 (10.0%) Assault with a weapon 27 (2.1%) 11 (2.0%) 1 (0.7) 11 (3.9%) 4 (1.9%) — Sexual assault 209 (16.5%) 49 (9.0%) 22 (15.8%) 60 (21.1%) 52 (24.8%) 26 (28.9%) Other unwanted sexual experience 120 (9.4%) 42 (7.7%) 15 (10.8%) 30 (10.5%) 25 (11.9%) 8 (52.2%) Combat or war exposure 8 (0.6%) 2 (0.4%) 1 (0.7%) 1 (0.4%) 2 (1.0%) 2 (54.4%) Captivity 7 (0.6%) 2 (0.4%) 1 (0.7%) 4 (1.4%) — — Life-threatening illness/injury 85 (6.7%) 46 (8.4%) 6 (4.3%) 17 (6.0%) 13 (6.2%) 3 (3.3.%) Severe human suffering 32 (2.5%) 9 (1.6%) 2 (1.4%) 9 (3.2%) 6 (2.9%) 6 (6.7%) Sudden violent death of loved one 71 (5.6%) 28 (5.1%) 3 (2.2%) 18 (6.3%) 13 (6.2%) 9 (10.0%) Sudden accidental death of loved one 86 (6.8%) 36 (6.6%) 19 (13.7%) 15 (5.3%) 13 (6.2%) 3 (3.3%) Serious injury to someone else 10 (0.8%) 1 (0.2%) 3 (2.2%) 2 (0.7%) 1 (0.5) 3 (3.3%) Unspecified traumatic experience 160 (12.6%) 39 (7.1%) 18 (12.9%) 46 (16.1%) 40 (19.0%) 17 (18.9%)

Note. Class 1 � asymptomatic PTSD–low illicit substance; Class 2 � asymptomatic PTSD– high illicit substance; Class 3 � low NACM/AAR–low illicit substance; Class 4 � moderate PTSD–moderate illicit substance; Class 5 � high PTSD–moderate illicit substance; AmED � alcohol mixed with energy drinks. a Data were not available for one participant (0.1%). b Data were not available for two participants (0.2%). c Designated by the participant as their “most stressful or traumatic” event; all participants endorsed that at least one of the 16 specified traumatic events “happened to them.”

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl



who endorsed that at least one of the 16 specified traumatic events “happened to me” (cf. “witnessed it”) were included in the anal- yses, consistent with the conservative approach used in trauma research (e.g., Kearns, Cloutier, Carey, Contractor, & Blumenthal, 2019; Paulus, Vujanovic, & Wardle, 2016; Thornley, Vorsten- bosch, & Frewen, 2016).

PTSD symptom severity. The PTSD Checklist for DSM–5 (PCL-5; Weathers et al., 2013b) is a 20-item self-report measure assessing past-month PTSD symptom severity. Participants were asked to complete the PCL-5 in response to their most stressful event from the LEC-5. The PCL-5 includes four subscales of Intrusions (Items 1–5), Avoidance (Items 6 –7), Negative Altera- tions in Cognition and Mood (NACM; Items 8 –14), and Altera- tions in Arousal and Reactivity (AAR; Items 15–20). Responses range from 0 (not at all) to 4 (extremely). The PCL-5 is a psycho- metrically sound measure (Blevins, Weathers, Davis, Witte, & Domino, 2015) and evidenced good reliability for the PTSD sub- scales in the current study (Cronbach’s �s � .89 to .92).

Substance use endorsement. Past-month substance use was assessed via a series of single-item questions derived from Barrett, Darredeau, and Pihl (2006). Specifically, participants were asked, “In the past month, how many times have you used [insert sub- stance] for the purpose of getting high, drunk, stoned, buzzed, or intoxicated?” for eight substances or categories of substances with similar neurological effects: alcohol, alcohol mixed with energy drinks (AmED), marijuana, cocaine, prescription stimulants (e.g., dextroamphetamine [Adderall], methylphenidate [Ritalin]), Sched- ule I/II hallucinogens (e.g., methylenedioxy-methamphetamine [MDMA], psilocybin [mushrooms]; Drug Enforcement Adminis- tration, 2017), and Schedule III hallucinogens (e.g., lysergic acid diethylamide [LSD] and ketamine; Drug Enforcement Adminis- tration, 2017). Responses to each substance use question ranged from zero occasions to 20 or more occasions. Participants report- ing zero occasions were coded as 0 (no), indicating no past-month use of that category of substances; conversely, participants report- ing one to two occasions or higher were coded as 1 (yes), indicat- ing past-month use of that category of substances (Kearns et al., 2019).

Coping strategies. The Brief COPE (Carver, 1997) is a 28-item self-report measure that evaluates 14 strategies (two questions for each strategy) for coping with stress: self-distraction, active coping, denial, substance use, emotional support, instrumental support, accep- tance, behavioral disengagement, venting, positive reframing, plan- ning, humor, religion, and self-blame. Participants rated the frequency of using each strategy on a 4-point scale from 1 (I haven’t been doing this at all) to 4 (I’ve been doing this a lot). Brief COPE subscales

evidence good psychometric properties (Carver, 1997) and adequate internal consistency: self-distraction (� � .62), active coping (� � .76), denial (� � .77), substance use (� � .93), emotional support (� � .83), instrumental support (� � .85), acceptance (� � .77), behavioral disengagement (� � .75), venting (� � .66), positive reframing (� � .81), planning (� � .79), humor (� � .84), religion (� � .89), and self-blame (� � .80).

Data Analytic Plan

Latent class analysis. An LCA was conducted to categorize participants into latent subgroups based on their past-month en- dorsement patterns of 20 PTSD symptoms (continuous indicators) and eight categories of substances (categorical indicators). Maxi- mum Likelihood estimation with robust standard errors as the estimator was used to address nonnormality and estimate missing data in Mplus. Missing data were minimal (i.e., �4% on any given item) and completely at random (�2 � 681.19, p � .090). To determine the optimal model, Akaike information criterion, Bayes- ian information criterion (BIC), and sample-size-adjusted BIC (SSABIC) values were evaluated; lower values indicate better model fit (Nylund, Bellmore, Nishina, & Graham, 2007). Further, adjusted Lo–Mendell–Rubin likelihood ratio test and bootstrapped likelihood ratio test values were examined; a statistically nonsig- nificant finding for the k – 1 class indicates a better fit for the k-class solution (Nylund et al., 2007). In terms of nonstatistical criteria, interpretability, parsimony, and size of latent classes as- sociated within each model were considered (Cloitre, Garvert, Weiss, Carlson, & Bryant, 2014; Lanza & Rhoades, 2013).

Three-step approach. Multinomial logistic regression (i.e., the three-step approach) was conducted in Mplus 8 to evaluate the con- struct validity of the optimal class solution. More specifically, the three-step approach, which accounts for misspecification bias (Asp- arouhov & Muthén, 2014; Vermunt, 2010), evaluated which coping strategies impacted likelihood of specific group classification (e.g., Classes 1 vs. Class 2). Minimal missing data were accounted for via listwise deletion in the secondary analyses (n � 1,173).


Latent Class Analysis (See Table 2)

Model selection. The five-class solution was selected as the optimal LCA model for two primary reasons. First, regarding the statistical criteria, the five-class solution produced the lowest SS- ABIC and BIC values of all the compared models with the de-

Table 2 Fit Indices From Analyses of One to Five Latent Classes

Model AIC BIC SSABIC Entropy LMR (p) BLRT (p)

1 Class 92,290.19 92,532.09 92,382.79 �.001 2 Class 79,485.34 79,871.35 79,633.12 .97 12,796.90 (�.001) �.001 3 Class 76,159.173 76,689.29 76,362.11 .96 3,365.35 (�.001) �.001 4 Class 74,672.60 75,346.83 74,930.71 .95 1,534.91 (.138) �.001 5 Class 73,862.03 74,680.36 74,175.30 .94 862.264 (.053) �.001

Note. AIC � Akaike information criterion; BIC � Bayesian information criterion; SSABIC � sample-size-adjusted BIC; LMR � Lo–Mendell–Rubin adjusted likelihood ratio test; BLRT � bootstrap likelihood ratio test.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl



crease in SSABIC and BIC values being minimal between the four- and five-class solutions. Further, the five-class solution con- tinued to have a significant bootstrapped likelihood ratio test value, indicating improved fit over preceding models. Additionally, the five-class solution had an entropy value similar to the more par- simonious models. Thus, aside from a nonsignificant Lo–Mendell– Rubin p value, which is not considered the most robust statistical indicator for LCA class selection (Nylund-Gibson & Masyn, 2016;), the majority of recommended statistical indicators sup- ported a five-class solution as the optimal model (DiStefano & Kamphaus, 2006). Second, regarding the nonstatistical criteria, the five-class solution evidenced adequate sample sizes (i.e., n � 90 in smallest class) and produced theoretically consistent and interpre- table classes (described below), consistent with the larger epide- miological literature on PTSD and substance use (Kilpatrick et al., 2013). Further, although limited, much of the existing literature on latent classes of PTSD comorbidities indicate more complex/ expansive solutions (Cloitre et al., 2014) than the standard low/ medium/high classes. Figure 1 provides a graphical depiction of the five-class model.

Classification. Generally, all classes evidenced similar prob- ability of endorsing alcohol, AmED, and marijuana/cannabis use. As such, for ease of interpretation, classes were defined as low, moderate, or high “illicit substance” use, indicating increased probability of endorsing past-month cocaine, prescription stimu- lant, and Schedule I/II and Schedule III hallucinogens use. Given these results, Class 1 (n � 546), which was relatively asymptom- atic with regard to PTSD severity and was generally characterized by lesser substance use, was labeled “asymptomatic PTSD–low illicit substance.” Comparatively, Class 2 (n � 139), which was also PTSD asymptomatic, but endorsed the highest levels of sub-

stance use, was labeled “asymptomatic PTSD– high illicit sub- stance.” Although more subtle, Classes 3 (n � 285) and 4 (n � 210) both had moderate PTSD intrusion and avoidance severity; however, Class 3 had lower PTSD NACM and AAR severity, relative to Class 4. Further, Class 3 evidenced less illicit substance use than Class 4. As such, Class 3 was labeled “low NACM/AAR– low illicit substance” and Class 4 was labeled “moderate PTSD– moderate illicit substance.” Lastly, Class 5 (n � 90) was charac- terized by the highest overall PTSD severity, as well as moderate substance use endorsement patterns, similar to Class 4. As such, Class 5 was labeled “high PTSD–moderate illicit substance.”

Three-Step Approach

Relative to the asymptomatic PTSD–low illicit substance use class, greater frequency of substance use coping (p � .001) and lesser behavioral disengagement (p � .031) and religious coping (p � .001) increased odds of being in the asymptomatic PTSD– high illicit substance class; greater frequency of self-distraction (p � .004), substance use (p � .016), venting (p � .017), accep- tance (p � .003), and self-blame coping (p � .011) increased odds of being in the low NACM/AAR–low illicit substance class; greater frequency of substance use (p � .001), behavioral disen- gagement (p � .020), venting (p � .040), and self-blame coping (p � .001) and lesser positive reframing coping (p � .004) increased odds of being in the moderate PTSD–moderate illicit substance class; and greater frequency of substance use (p � .001), behavioral disengagement (p � .001), and self-blame coping (p � .001) increased odds of being in the high PTSD–moderate illicit substance class. See Table 3 for full class comparison information (three-step approach).










In tr

u si

v e t

h o

u g

h ts

R e c u rr

e n t

N ig

h tm

a re


F la

sh b

a c k


E m

o ti

o n a l

re a c ti

v it


P h y si

o lo

g ic

a l

re a c ti

v it


A v o id

a n c e o

f th

o u g h ts

A v o id

a n c e o

f re

m in

d e rs

M e m

o ry

i m

p a ir

m e n t

N e g a ti

v e b

e li

e fs

B la

m e o

f se

lf o

r o th

e rs

N e g a ti

v e t

ra u

m a e

m o ti

o n s

L o

ss o

f in

te re


D e ta

c h m

e n t

R e st

ri c te

d r

a n g e o

f a ff

e c t

Ir ri

ta b il

it y /a

n g e r

R re

c k

le ss

b e h

a v

io r

H y p e rv

ig il

a n c e

E x

a g

g e ra

te d

s ta

rt le

r e sp

o n


D if

fi c u lt

y c

o n c e n tr

a ti

n g

D if

fi c u lt

y s

le e p in













A lc

o h

o l

A lc

o h o l

m ix

e d E

n e rg


D ri

n k


M a ri

ju a n a /C

a n n a b is

C o c a in


P re

sc ri

p ti

o n

S ti

m u

la n


S c h

e d

u le

I /I


a ll

u c in

o g

e n

S c h e d u le


H a ll

Attachment 4

Journal of Traumatic Stress February 2014, 27, 50–57

Associations Between Perceived Social Reactions to Trauma-Related Experiences With PTSD and Depression Among Veterans Seeking

PTSD Treatment

Jeremiah A. Schumm,1,2 Ellen M. Koucky,1,3 and Alisa Bartel1 1Trauma Recovery Center, Cincinnati Veterans Affairs Medical Center, Cincinnati, Ohio, USA

2Department of Psychiatry, University of Cincinnati, Cincinnati, Ohio, USA 3Department of Clinical Psychology, University of Missouri-St. Louis, St. Louis, Missouri, USA

The Social Acknowledgment Questionnaire (SAQ; Maercker & Mueller, 2004) is a measure of trauma survivors’ perceptions of social acknowledgment and disapproval from others, and these factors are shown to be associated with posttraumatic stress disorder (PTSD) among civilian trauma survivors. This study seeks to validate the structure of the SAQ among U.S. military veterans and test the hypothesis that family and general disapproval are associated with PTSD and depression among veterans. Participants were 198 U.S. veterans who experienced military trauma and completed an intake evaluation through a Veterans Affairs PTSD treatment program. Structural equation modeling (SEM) results supported a well-fitting 3-factor model for the SAQ that was similar to prior studies in capturing the constructs of social acknowledgment, general disapproval, and family disapproval. SEM results also showed that all 3 of the SAQ factors were associated with veterans’ depression (−.31, .22, and .39, respectively), whereas only general disapproval was related to veterans’ PTSD. This is the first study of which we are aware to investigate the factor structure of the SAQ in a veteran sample and to investigate the relationship between SAQ factors and trauma survivors’ depression. Results build upon prior findings by showing the importance of positive and negative social reactions to veterans’ traumatic experiences.

Negative social reactions toward trauma survivors are dam- aging to trauma survivors’ posttraumatic psychological adjust- ment. Examples of negative social reactions include perceptions that others are underestimating the severity of one’s traumatic experiences or that others do not understand the negative conse- quences of these traumas. Ullman and Filipas (2001) found that negative reactions of others toward trauma disclosure were pos- itively related to female sexual assault survivors’ posttraumatic stress disorder (PTSD) severity. Ullman (2007) and Ullman and Filipas (2005) demonstrated that after controlling for abuse severity, survivor self-blame, and survivor coping, negative so- cial reactions that others had toward adult sexual abuse sur- vivors were positively associated with their PTSD symptoms.

Preparation of this manuscript was supported by the Department of Veterans Affairs, including grant CDA-2-019-09S (Jeremiah A. Schumm). Content of this manuscript does not necessarily reflect those of the United States Govern- ment or Department of Veterans Affairs. We thank Misty Wolfe and Lindsey Davidson for their assistance in verifying and coding data.

Correspondence concerning this article should be addressed to Jeremiah A. Schumm, VA Medical Center, 1000 South Fort Thomas Avenue, Fort Thomas, KY 41075. E-mail: [email protected]

Copyright C© 2014 International Society for Traumatic Stress Studies. View this article online at wileyonlinelibrary.com DOI: 10.1002/jts.21879

In contrast, positive social reactions were not significantly re- lated to survivors’ PTSD symptoms.

Maercker and colleagues examined the impact of social reactions toward trauma survivors among various samples, including German interpersonal crime victims, former Ger- man political prisoners, and Chechnyan refugees (Maercker & Mueller, 2004; Mueller, Moergeli, & Maercker, 2008; Maer- cker, Povilonyte, Lianova, & Pöhlmann, 2009, respectively). These researchers developed a measure of social reactions toward the trauma survivor called the Social Acknowledg- ment Questionnaire (SAQ). The SAQ measures three distinct sociocognitive domains: Recognition (perceiving that others offered sympathy or help), General Disapproval (perceiving that the general society does not understand the survivor’s re- sponses), and Family Disapproval (perceiving that the family underestimates the survivor’s traumas or believes that the sur- vivor’s reactions are exaggerated). The SAQ total score had a stronger relationship to PTSD than a conventional measure of social support (Maercker & Mueller, 2004) and was longitu- dinally predictive of PTSD over and above a well-established measure of dysfunctional trauma-related cognitions (Mueller et al., 2008). General Disapproval was positively predictive of later PTSD, after accounting for other risk factors (Mueller et al., 2008). These findings show that social reactions to trauma survivors are distinct, important constructs and provide further


Social Reactions and Military Trauma 51

evidence that disapproving and unsupportive social reactions are predictive of survivors’ PTSD.

Two studies examined how social reactions impact PTSD among male Vietnam veterans. Using a large-scale community sample of Vietnam combat veterans, Fontana and Rosenheck (1994) showed that after accounting for risk factors such as combat severity and noncombat traumas, veterans’ perceptions of negative societal and familial homecoming reactions toward them were strongly and positively related to veterans’ PTSD. In another large-scale study of male Vietnam era veterans, Koenen, Stellman, Stellman, and Sommer (2003) found that after accounting for prior PTSD, combat exposure severity, and discomfort in disclosing their war experiences, veterans’ per- ceptions of negative community reactions were longitudinally and positively predictive of veterans’ PTSD; however, veter- ans’ perceptions of positive family support in readjusting to civilian life were unrelated to their PTSD. These findings are consistent with those from nonveteran samples (Mueller et al., 2008; Ullman, 2007; Ullman & Filipas, 2001, 2005), and may be explained by negative social reactions having a stronger relationship to PTSD than positive social reactions. It is also possible, however, that perceptions of societal versus famil- ial responses to trauma survivors may differentially relate to survivors’ PTSD.

The first goal of the current study was to test the factor structure of the SAQ (Maercker & Mueller, 2004) in a U.S. veteran treatment-seeking sample. We chose to examine the SAQ because this instrument provides a multilayered assess- ment of trauma survivors’ perceptions of societal, community, and familial reactions to their traumatic experiences. Another advantage is that the SAQ is applicable to veterans experiencing both combat and noncombat trauma. The SAQ was originally developed and tested in a German sample of crime victims and political prisoners; therefore, one goal was to test the hy- pothesized latent factor structure and convergent validity of the English language SAQ in a U.S. veteran sample. Maercker and Mueller (2004) used principal components factor analysis to test the structure of the SAQ. We built upon Maercker and Mueller’s (2004) approach by using structural equation mod- eling (SEM), which provides various tests of overall model fit (Kline, 2011).

This study had several other goals. First, studies by Fontana and Rosenheck (1994) and by Koenen et al. (2003) involved male Vietnam veterans who were surveyed 10–20 years ago. The current study sought to replicate the findings that social reactions are related to PTSD in a more recent sample of U.S. veterans. Second, the present study expanded upon prior re- search (Littleton, 2010) by examining whether social reactions are related to depression in addition to PTSD. Understand- ing the link between social reactions to trauma survivors and depression among veterans will be an important advance be- cause depression is a common co-occurring problem that fre- quently affects trauma survivors (e.g., Breslau, Davis, Peter- son, & Schultz, 2000). Third, the current study built upon prior community- and university-based samples of trauma survivors

to examine the relationship between social reactions and psy- chological outcomes among veterans who are seeking PTSD treatment.

Based upon prior research, we had the following hypotheses. 1. First, consistent with Maercker and Mueller’s (2004) factor analytic results, the SAQ would exhibit a well-fitting, intercor- related 3-factor model solution reflecting underlying constructs of Recognition, General Disapproval, and Family Disapproval. 2. Second, negative social reactions (i.e., General Disapproval and Family Disapproval) would be positively associated with veterans’ PTSD and depression. 3. Third, negative social reac- tions (i.e., General Disapproval and Family Disapproval) would have a stronger relationship with veterans’ PTSD and depres- sion than positive social reactions (i.e., Recognition).


Participants and Procedure

Participants were 198 U.S. military veterans who completed an initial PTSD diagnostic assessment for an outpatient Veterans Affairs PTSD treatment program and whose worst-reported in- dex trauma occurred in the military. Based upon the diagnostic assessment, three fourths of the sample met diagnostic criteria for PTSD and over half met criteria for major depressive dis- order (MDD). Participants were an average of 44.01 years old (SD = 14.01) and had an average of 13.53 years of education (SD = 1.99). The sample was predominantly male of either Caucasian or African American descent. Over three fourths of the sample served in combat, and most served during either Vietnam or Operation Iraqi Freedom/Operation Enduring Free- dom (see Table 1). Most veterans reported combat as their most severe index trauma.

Veterans completed a pretreatment diagnostic assessment that included gathering demographic data and completion of the study measures as part of a larger assessment battery. Be- cause assessments were delivered as part of routine clinical care, assessments were not recorded and interrater reliability data were unavailable. Item-level data were collected for the SAQ, but not for other measures. The University of Cincinnati institutional review board approved the use of veterans’ archival chart data for the purpose of this study.


The SAQ (Maercker & Mueller, 2004) is a 16-item scale that measures social acknowledgement as a victim or survivor. Re- spondents rate each item on a 4-point scale that ranges from 0 = I agree not at all to 3 = I agree completely. There are three subscales: Recognition, General Disapproval, and Fam- ily Disapproval. Recognition is characterized by perceptions that others offered help or sympathy in response to the trauma. General Disapproval is characterized by perceptions of being a societal outsider and others not understanding one’s strug- gles in response to the trauma. Family Disapproval involves

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

52 Schumm, Koucky, and Bartel

Table 1 Selected Demographic Characteristics of Sample

Variable n %

Male 171 86.4 Ethnicity

White 160 80.8 African American 34 17.2 Other 4 2.0

Marital status Married 104 52.5 Separated/divorced 66 33.4 Never married 24 12.1 Widowed 4 2.0

Service era Vietnam 50 25.3 Post-Vietnam 23 11.6 Persian Gulf 38 19.2 OIF/OEF 87 43.9

Had combat experience 165 83.3 Met diagnosis for PTSD 149 75.3 Met diagnosis for MDD 115 58.1 Index trauma

Sexual assault 24 12.1 Combat 140 70.7 Witness to death 9 4.5 Other 25 12.7

Note. N = 198. All index traumas occurred during military service. OIF = Opera- tion Iraqi Freedom; OEF = Operation Enduring Freedom; PTSD = posttraumatic stress disorder; MDD = major depressive disorder.

perceptions that family members underestimate one’s traumatic experiences and that the survivor’s reactions are exaggerated. Higher scores on the General Disapproval and Family Disap- proval subscales represent negative social responses, and higher scores on the Recognition subscale represent positive social responses to trauma. The SAQ subscales demonstrated high internal consistency, good test-retest reliability, and correlated moderately to strongly with other trauma-related constructs in- cluding PTSD severity and social support (Maercker & Mueller, 2004).

The Clinician-Administered PTSD Scale (CAPS; Blake et al., 1995, 2000) is a clinician-administered structured inter- view that assesses the frequency and intensity of PTSD symp- toms based on diagnostic criteria in the Diagnostic and Statisti- cal Manual of Mental Disorders (4th ed., text rev.; DSM-IV-TR; American Psychiatric Association, 2000).

PTSD symptoms were counted toward meeting diagnostic criteria if the frequency was rated at least 1 (symptom occurs monthly) and intensity at least 2 (symptom is associated with moderate distress). The sum of the frequency and intensity rat- ings of PTSD symptoms produce the PTSD severity score with higher scores indicating greater symptom severity. The CAPS was also used to determine whether participants met diagnos-

tic criteria for PTSD. The CAPS has established internal con- sistency (Blake et al., 1995) and has also demonstrated good- to-excellent interrater reliability and convergent and diagnostic criterion validity with veteran samples (Weathers, Keane, & Davidson, 2001).

The PTSD Checklist-Specific (PCL-S; Weathers, Litz, Herman, Huska, & Keane, 1993) is a 17-item self-report mea- sure of PTSD symptoms that corresponds with DSM-IV-TR diagnostic criteria. The items are summed to yield a total sever- ity score ranging from 17 to 85, where higher scores indicate greater symptom distress. The PCL-S has demonstrated ex- cellent internal and test-retest reliability in addition to strong convergent and discriminant validity within veteran samples (Weathers et al., 1993).

The Structured Clinical Interview for DSM-IV-TR Axis I Disorders (SCID; First, Spitzer, Gibbon, & Williams, 2002) is a semistructured clinical interview designed to determine whether a DSM-IV-TR Axis I diagnosis is present. For the current study, the results from the SCID were used to deter- mine whether or not participants had a current diagnosis of major depressive disorder. The SCID has demonstrated good- to-excellent interrater and test-retest reliability (Zanarini & Frankenburg, 2001).

The Beck Depression Inventory–II (BDI-II; Beck, Steer, & Brown, 1996) is a 21-item self-report scale that measures de- pression severity within the last 2 weeks. The item scores are summed to yield a total score range between 0–63. Higher scores indicate greater depression symptoms. The BDI-II has established psychometric properties including excellent inter- nal and test-retest reliability and strong convergent validity (Beck et al., 1996).

Data Analysis

Using Mplus (version 6.11; Muthén & Muthén, 2011), confir- matory factor analysis (CFA) was conducted to test the model fit of Maercker and Mueller’s (2004) 3-factor model solution for the SAQ. The maximum likelihood estimator with robust- ness correction for nonnormality was used. Statistical fit was evaluated by examining the model χ2 test, comparative fit in- dex (CFI; Bentler, 1990), Tucker-Lewis Index (TLI; Tucker & Lewis, 1973), and root mean square error of approximation (RMSEA; Browne & Cudeck, 1993). We also examined the structural coefficients of SAQ items to determine if items were adequate indicators of the hypothesized latent constructs.

We next tested the hypothesized relationships involving the SAQ factors as correlates of PTSD and depression. This was accomplished through structural equation modeling (SEM) in Mplus (version 6.11; Muthén & Muthén, 2011). Following Kline’s (2011) procedures, SEM equations were computed step- wise. First, measurement models were computed to examine the fit of the observed variables to their corresponding latent factors. Second, structural regression models were computed to test the association between SAQ factors and PTSD and depression. PTSD was indicated by the following observed

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

Social Reactions and Military Trauma 53

variables: CAPS total severity score, PCL-S total score, and whether participants met DSM-IV-TR criteria for PTSD based upon the CAPS. Depression was indicated by the following observed variables: BDI-II total score and whether participants met major depressive disorder diagnosis based upon the SCID. Because PTSD and depression diagnoses were dichotomous variables, we used the weighted least squares with mean and variance adjustment estimator.

The proportion of missing data was low (1.4% of the dataset). To account for missing data, the full information maximum likelihood method was used in CFA and SEM analyses.



To test our first hypothesis regarding the SAQ factor structure, we conducted a CFA using Maercker and Mueller’s (2004) 3- factor structure. The model fit indices showed an unacceptable fit. The model χ2 test showed that the hypothesized model significantly deviated from the observed data, χ2 (101, N = 198) = 198.46, p < .001. In addition, the CFI (.80) and TLI (.76) were below a suggested minimum value of .95 (Hu & Bentler, 1999), and the RMSEA (.07) 90% confidence interval (CI) = [.06, .08] was above a suggested maximum value of .06 (Browne & Cudeck, 1993).

Given the unacceptable fit for the initial CFA, we explored reasons for model misspecification and ways that the model could be respecified to improve the model fit. First, we found that the structural coefficient for SAQ item 9 (“My family feels that they have to protect me”) did not significantly load (p = .855) onto the hypothesized latent factor, which Maercker and Mueller (2004) labeled as Family Disapproval. In addition, the modification indices did not provide support for Item 9 being related to either of the other two latent factors in the model, which Maercker and Mueller (2004) called Recognition and General Disapproval. Hence, Item 9 was not a good factor indicator in the present sample and was, therefore, removed from the respecified CFA model. Second, although Item 5 (“The only people who really understand me are those who have been through something similar”) significantly loaded onto General Disapproval (p = .002), the structural coefficient for this item (.31) was below a suggested minimum acceptable value of .32 (Comrey & Lee, 1992) and did not load onto other factors. Therefore, we removed this item from the respecified CFA.

Model modification indices suggested several other changes would improve model fit. First, Item 1 (“Most people can un- derstand what I went through”) was not significantly related to the latent factor General Disapproval. Modification indices suggested that the model fit would be improved by having Item 1 to load onto the Recognition factor. Therefore, the respecified model included Item 1 as a factor indicator of Recognition in- stead of General Disapproval. Second, modification measures showed that the model would be improved by allowing Item 11 (“My family showed a lot of understanding for me after the incident”) to load onto Recognition in addition to Family Dis-

approval. Rather than retain this item, we chose to remove it to avoid having this item load onto two separate factors. Third, modification indices showed that correlating the error terms between Item 14 (“Many people offered their help in the first few days after the incident”) and Item 16 (“My boss/superior showed full understanding for any absence from work after the incident”) would improve model fit. Following these changes, a respecified CFA was computed.

The respecified CFA achieved an adequate fit, χ2 (61, N = 198) = 78.06, p = .064; CFI = .96; TLI = .94; RMSEA = .04, 90% CI = [.00, .06]. In addition, item coefficients were in a consistent and expected direction with their hypothesized underlying factors, and two of the three factors were intercorre- lated (see Tables 2 and 3). In summary, although we did not fully replicate Maercker and Mueller’s (2004) exact factor structure for the SAQ, the respecified model provided partial support for our first hypothesis.

SEM Measurement Model

We next tested a measurement model to assess the model fit when PTSD and depression variables were included. The SAQ latent factors and observed indicators were identical to those described in the respecified CFA and shown in Table 2. Hence, there were 3 latent and 13 observed variables that were derived from the SAQ. PTSD and depression were modeled as latent variables. PTSD had three observed indicators (CAPS severity total, PCL-S total, CAPS PTSD diagnosis), and depression had two observed indicators (BDI-II total and SCID major depres- sive disorder diagnosis). Therefore, the model had a total of 5 latent and 18 observed variables.

The measurement model had reasonable fit. Specifically, the overall model fit indices were suggestive of a well-fitting model, χ2(124, N = 198) = 150.07, p = .0596, CFI = .96, TLI = .95, RMSEA = .03, 90% CI = [.00, .05]. All of the observed in- dicators loaded significantly onto their corresponding hypoth- esized latent factors (ps < .01). Item 4 loaded onto the General Disapproval latent factor at .28, however, which was below a suggested minimum value of .32 (Comrey & Lee, 1992). We removed Item 4 from the model and ran a respecified measure- ment model, which produced an adequate fit to the data, χ2(108, N = 198) = 116.53, p > .27, CFI = .99, TLI = .98, RMSEA = .02, 90% CI = [.00, .04]. Therefore, this respecified model was used as the basis for computing the structural model.

SEM Structural Model

To test our second and third hypotheses, we computed a struc- tural model that included the three SAQ latent variables as correlates of PTSD and depression. The overall model fit was acceptable, χ2(108, N = 198) = 116.34, p = .275, CFI = .98, TLI = .97, RMSEA = .02, [90% CI = .00, .04]. Recognition was significantly and positively related to depression, but not significantly related to PTSD. The model regression pathways were mostly supportive of Hypothesis 2. Specifically, General Disapproval and Family Disapproval were significantly and

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

54 Schumm, Koucky, and Bartel

Table 2 Standardized Structural Coefficients for Respecified CFA of Social Acknowledgment Questionnaire

Item Factor 1 Factor 2 Factor 3

1. Most people cannot understand what I went through. .39 2. Somehow I am no longer a normal member of society since the incident. .70 3. The people where I live respect me more since the incident. .56 4. There is not enough sympathy for what happened to me. .36 6. My family finds my reaction to the incident to be exaggerated. .59 7. Most people cannot imagine how difficult it is simply to continue with “normal” daily life. .64 8. My experiences are underestimated by my family. .85

10. My family feels uncomfortable talking about my experiences. .43 12. My friends show sympathy for what happened to me. .73 13. The reactions of my acquaintances were helpful. .71 14. Many people offered their help in the first few days after the incident. .56 15. Important figures of public life (e.g., mayor, priest) expressed their sympathy for me after

the incident. .52

16. My boss/superior showed full understanding for any absence from work after the incident. .56 Factor ρ coefficient .60 .79 .67

Note. N = 198. Results are based upon the respecified model solution that excludes Social Acknowledgment Questionnaire Items 5, 9, and 11. CFA = confirmatory factor analysis. p < .001 for all structural coefficients.

positively related to depression. General Disapproval, but not Family Disapproval, was significantly and positively related to PTSD (see Figure 1).

To test Hypothesis 3, χ2 difference test for models utilizing the WLSMV estimator (Muthén & Muthén, 1998–2010) were calculated in Mplus (Muthén & Muthén, 2011). These differ- ence tests compared the SEM equation that freely estimate the SAQ factors in accounting for PTSD and depression with models that constrain SAQ factors as being equal in absolute magnitude in accounting for PTSD or depression. Constraining the pathway from general disapproval to PTSD to be equal in absolute magnitude to the pathway from Recognition to PTSD did not significantly deteriorate the overall model fit, χ2(1, N = 198) = 2.89, p > .09. Constraining the General Disapproval and Recognition to be equal in absolute magnitude in accounting for depression also did not deteriorate the overall model fit, χ2(1,

Table 3 Standardized Factor Correlations for Respecified CFA of Social Acknowledgment Questionnaire

Factor Factor correlations

1. Factor 1 General Disapproval – 2. Factor 2 Recognition −.11 – 3. Factor 3 Family Disapproval .42*** −.24*

Note. N = 198. Results are based upon the respecified model solution that excludes Social Acknowledgment Questionnaire Items 5, 9, and 11. CFA = confirmatory factor analysis. Factor ρ coefficients were calculated according to Kline (2011, pp. 242). *p < .05. ***p < .001.

N = 198) = 0.22, p > .63. In comparing Family Disapproval to Recognition, constraining the absolute magnitude of these variables in accounting for PTSD did not deteriorate the model fit, χ2(1, N = 198) = 0.001, p > .97. Finally, constraining Family Disapproval and Recognition to be equal for depres- sion did not result in worse model fit, χ2(1, N = 198) = 0.32, p > .57. Hence, contrary to Hypothesis 4, General Disapproval and Family Disapproval were not significantly more strongly associated with PTSD or depression.


The first goal of this study was to replicate Maercker and Mueller’s (2004) 3-factor structure for the SAQ within a sample of U.S. veterans seeking treatment for PTSD. Our findings pro- vided support for a well-fitting, 3-factor model that was similar to the original findings for the SAQ among community samples of German trauma survivors. Hence, our results showed cross- cultural support for three distinct domains, which Maercker and Mueller (2004) labeled as Recognition, General Disapproval, and Family Disapproval. These findings add to the evidence that these sociocognitive constructs can be reliably measured among both civilian and military samples that have varying types of traumatic experiences.

Although our findings were generally supportive of the factor domains assessed by the SAQ, we found that several of the SAQ items contributed to model fit problems and needed to be respec- ified or removed from the model. For example, Items 5 (“My family feels that they have to protect me”) and 9 (“The only peo- ple who really understand me are those who have been through something similar”) had unacceptably weak coefficients in the

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

Social Reactions and Military Trauma 55

General Disapproval

Family Disapproval





















.85*** .72***


.81*** .68***




.83*** .50***

SAQ-1Q SAQ-3 SAQ-12 SAQ-13 SAQ-14 SAQ-15



.56*** .68*** .54*** .52***




R2 = .29***

R2 = .46***

Figure 1. Standardized structural equation model showing the association between social acknowledgment scale independent variables and PTSD and depression outcomes. SAQ = Social Acknowledgment Questionnaire—numbers following SAQ represent the scale item number; PCL = PTSD Checklist; CAPS Total = Clinician-Administered PTSD Scale total severity score; CAPS Dx = Clinician-Administered PTSD Scale PTSD diagnosis; BDI-II = Beck Depression Inventory-II; SCID Dx = Structured Clinical Interview for the DSM-IV major depressive disorder diagnosis. *p < .05. **p < .01. ***p < .001.

initial model, and Item 4 (“There is not enough sympathy for what happened to me”) was found to be a poor SAQ factor indicator in the subsequent model that included SAQ factors along with PTSD and depression. This suggests that although many of the items performed similarly in the current U.S. mil- itary sample in comparison to prior studies of German crime victims and political prisoners (Maercker & Mueller, 2004) and Chechnyan refugees (Maercker et al., 2009), some of these items did not effectively reflect the hypothesized constructs for the current study sample. Given the greater emphasis on indi- vidualism within the United States combined with expectations that military members are trained “protector[s]” who have the ability to independently “tough out” negative emotions related to their trauma (Nash, Silva, & Litz, 2009), U.S. veterans may not view family members’ attempts to “protect” them as be- ing a desirable way for families to show support or approval in response to veterans’ traumatic experiences. Also, among treatment-seeking U.S. military veterans, a lack of expressed sympathy may not be necessarily interpreted as behavior that reflects general disapproval toward their traumatic experiences. Rather, due to a tendency for individuals with PTSD to avoid discussing these experiences with others (Dekel & Monson, 2010), veterans who are entering but have yet to begin PTSD treatment may perceive that a lack of expressed sympathy from others in relation to the trauma is showing a degree of respect for …

Attachment 5

Treatment Choice Among Veterans With PTSD Symptoms and Substance- Related Problems: Examining the Role of Preparatory Treatments in

Trauma-Focused Therapy

Laura D. Wiedeman Edward Hines Jr. Veterans Affairs Hospital, Hines, Illinois, and

Veterans Affairs Northern California Health Care System, Martinez, California

Susan M. Hannan Edward Hines Jr. Veterans Affairs Hospital, Hines, Illinois, and

Lafayette College

Kelly P. Maieritsch Edward Hines Jr. Veterans Affairs Hospital, Hines, Illinois

Cendrine Robinson Edward Hines Jr. Veterans Affairs Hospital, Hines, Illinois, and

National Cancer Institute, Rockville, Maryland

Gregory Bartoszek University of Illinois at Chicago

Although common practice in Veterans Affairs (VA) PTSD clinics, it is unclear whether preparatory treatment improves trauma-focused treatment (TFT) completion and outcomes. Furthermore, little is known about whether treatment-seeking veterans in naturalistic settings would chose to prioritize preparatory treatment if given the option of a phase-based approach or direct access to TFT, and how substance-related problems (SRPs) influence this treatment choice. The first aim of this study was to explore how co-occurring SRPs (ranging from none to moderate/severe) influence PTSD treatment choices in a naturalistic setting where veterans were offered a choice between a phase-based approach (i.e., preparatory treatment) or direct access to TFT. The study also examined whether initial treatment choice and severity of co-occurring SRPs influenced TFT completion and outcomes. The second aim was to investigate whether preparatory treatment led to superior TFT completion or outcomes, irrespective of co-occurring SRPs. Analyses were conducted using archival data from 737 United States veterans referred for outpatient VA PTSD treatment. SRPs did not predict initial treatment choice or the length of preparatory group participation. Neither SRPs nor preparatory group participation predicted TFT completion or outcomes (measured as change in PTSD and depression symptoms from pre- to post-TFT). Preparatory group participation did not predict improved TFT completion or outcomes, irrespective of co-occurring SRPs. These findings suggest that veterans with PTSD symptoms and co-occurring SRPs may make similar treatment choices and benefit from either a phase-based approach or direct TFT initiation, and preparatory treatments may not increase patient readiness for veterans seeking TFT.

Keywords: posttraumatic stress, veterans, substance use, treatment selection, preparatory treatment

Supplemental materials: http://dx.doi.org/10.1037/ser0000313.supp

There are a number of barriers that can interfere with starting or completing trauma-focused therapies (such as prolonged exposure [PE] and cognitive processing therapy [CPT]). Patients (especially

military veterans) report distrust of mental health care providers and fear of stigma for seeking mental health treatment (Hoge et al., 2004). In addition, research suggests that individuals with post-

This article was published Online First November 26, 2018. Laura D. Wiedeman, Edward Hines Jr. Veterans Affairs

Hospital, Hines, Illinois, and Veterans Affairs Northern California Health Care System, Martinez, California; Susan M. Hannan, Edward Hines Jr. Veterans Affairs Hospital, and Department of Psychology, Lafayette College; Kelly P. Maieritsch, Edward Hines Jr. Veterans Affairs Hospital; Cendrine Robinson, Edward Hines Jr. Veterans Af- fairs Hospital, and National Cancer Institute, Rockville, Maryland; Gregory Bartoszek, Department of Psychology, University of Illinois at Chicago.

Kelly P. Maieritsch is now at the National Center for PTSD, Executive Division, White River Junction, Vermont. Gregory Bartoszek is now at the Department of Psychology, William Paterson University.

This project is the result of work supported with resources and the use of facilities at the Edward Hines Jr. Veterans Affairs Hospital, Hines, Illinois. The contents of this article do not represent the views of the U.S. Department of Veterans Affairs or the U.S. Government.

Correspondence concerning this article should be addressed to Laura D. Wiedeman, VA Northern California Health Care System, 150 Muir Road (116), Martinez, CA 94553. E-mail: [email protected]

Psychological Services In the public domain 2020, Vol. 17, No. 4, 405– 413



traumatic stress disorder (PTSD) who identify as low-income racial minorities are especially likely to face multiple barriers to accessing treatment (Davis, Ressler, Schwartz, Stephens, & Brad- ley, 2008). Furthermore, providers have emphasized the impor- tance of assessing patient readiness for trauma-focused treatment (TFT); however, a consistent definition of readiness is lacking (Cook, Simiola, Hamblen, Bernardy, & Schnurr, 2017; Hamblen et al., 2015; Osei-Bonsu et al., 2017).

Co-occurring substance use has frequently been cited as one reason patients may not be ready for TFT, and surveys of providers have revealed a perception that veterans with PTSD and co- occurring substance use disorders (SUD) are more difficult to treat (Najavits, Norman, Kivlahan, & Kosten, 2010; Osei-Bonsu et al., 2017). In fact, 75% of randomized controlled trials of PTSD treatment used substance-related exclusion criteria (Leeman et al., 2017), creating challenges for generalizing findings to the PTSD/ SUD population (Najavits & Hien, 2013). Research has been further limited by the lack of specificity in which SUD and related problems are described, as well as grouping those with SUD into one homogenous category (Kirisci et al., 2006).

Of the 5 million veterans seen within the Department of Veter- ans Affairs (VA) in 2012, 34% had a diagnosis of PTSD (Bowe & Rosenheck, 2015), and those with PTSD were three times more likely to have a SUD diagnosis compared with the general popu- lation (Petrakis, Rosenheck, & Desai, 2011). Furthermore, Pi- etrzak, Goldstein, Southwick, and Grant (2011) analyzed data from 34,653 U.S. adults and found that the prevalence of co-occurring PTSD and any alcohol or drug use disorder was 46.4%. Until recently, clinical practice guidelines for the treatment of PTSD/ SUD were lacking; clinicians were therefore left to determine how to utilize existing PTSD and SUD treatment options to address the needs of this population. Surveys of VA clinicians have high- lighted provider concerns about offering TFT to veterans with co-occurring SUD, including whether they can tolerate TFT safely (e.g., without relapse), whether existing TFTs need adaptation to be effective for those with PTSD/SUD, and whether there is a greater need for stabilization prior to TFT initiation (Najavits et al., 2010; Osei-Bonsu et al., 2017).

Historically, there has been a lack of consensus around when to offer phase-based treatment (prioritizing psychoeducation and present-focused stabilization before trauma processing) or TFT as a first-line approach (Cloitre et al., 2011; Hamblen et al., 2015; Raza & Holohan, 2015). A national survey of VA PTSD clinic directors revealed that the vast majority of VA PTSD treatment clinics include mandatory psychoeducational or coping skills groups as a method of preparation for TFT, with the length of preparatory groups ranging from 1 to 12 sessions (Hamblen et al., 2015; Raza & Holohan, 2015). According to Hamblen and col- leagues (2015), clinic directors perceive these preparatory groups as improving readiness for TFT. Proponents of a phase-based approach suggest that having a multicomponent therapy model provides greater benefits and a more nuanced, patient-centered approach to treatment, as patients have a range of treatment op- tions to choose from during the treatment planning process (Cloi- tre, 2015). In addition, given well documented clinician uncer- tainty and lack of consensus over when to offer TFT to individuals with co-occurring SUD (Najavits et al., 2010; Raza & Holohan, 2015), preparatory treatment options via a phase-based approach may provide a welcomed, readily accessible alternative.

In 2017, the VA and Department of Defense (DoD) released updated clinical practice guidelines for the treatment of PTSD (Department of Veterans Affairs, 2017), in which it is recom- mended that clinicians offer TFT even in the presence of co- occurring SUD. These updated clinical practice guidelines stem from recent research that has found little evidence to support a particular sequencing of services (i.e., sequential, integrated, con- current) or a particular treatment type (e.g., present/coping- focused, past-focused, addiction-focused, trauma-focused) for in- dividuals with PTSD/SUD (e.g., Najavits & Hien, 2013; Roberts, Roberts, Jones, & Bisson, 2015; van Dam, Vedel, Ehring, & Emmelkamp, 2012). Simpson, Lehavot, and Petrakis (2017) found that after matching the target PTSD/SUD interventions and com- parison condition on time and attention, the exposure-based (i.e., trauma-focused) approach was most effective for treating PTSD/ SUD. That said, Simpson and colleagues (2017) also noted that all participants on average showed symptom improvement in all study conditions (addiction-focused, coping-focused, and exposure- based).

Although the use of preparatory treatments to enhance readiness for TFT and increase the likelihood of treatment completion has been common practice, the effectiveness of this phase-based ap- proach is unknown (Hamblen et al., 2015). Research focused on assessing patients’ treatment preferences may help illuminate their own perception of readiness (Cook et al., 2017). This may be especially important for veterans presenting with PTSD symptoms and co-occurring substance-related problems (SRPs), given previ- ously mentioned provider concerns about offering TFT to veterans with these co-occurring conditions. Data from naturalistic settings are needed to assess what treatment path veterans with PTSD symptoms and SRPs choose (a phase-based approach by first engaging in preparatory treatments, or direct access to TFT), how the presence of SRPs may influence this decision, and whether preparatory treatment leads to enhanced TFT outcomes.

Therefore, the first aim of this study was to examine the treat- ment choices of veterans with PTSD symptoms in a naturalistic setting and to better understand how the presence of SRPs may influence PTSD treatment choices and outcomes. This study ex- pands on previous research with the PTSD/SUD population by considering SRPs on a continuum rather than as a dichotomous (yes/no) variable. This research occurred in a VA PTSD clinic prior the 2017 VA/DoD clinical practice guidelines, where veter- ans were offered the option of selecting either a phase-based approach (by opting for preparatory treatment) or direct access to TFT. On basis of provider concerns about offering TFT to veter- an’s with PTSD and co-occurring SRPs (Najavits et al., 2010; Osei-Bonsu et al., 2017; Raza & Holohan, 2015) and the common use of preparatory treatments to improve coping skills and symp- tom management prior to TFT (Hamblen et al., 2015), we hypoth- esized the following: veterans with PTSD symptoms and greater SRPs would be (1) more likely to choose preparatory treatment and (2) more likely to participate in a longer course of preparatory treatment prior to TFT than veterans with fewer SRPs. In addition, we hypothesized that veterans with PTSD symptoms and greater SRPs would be (3) more likely to complete TFT and (4) more likely to have superior TFT outcomes (i.e., greater PTSD and depression symptom reduction from pre- to post-TFT) if they first participated in preparatory treatment than veterans with fewer SRPs.


As preparatory groups are commonly used within VA PTSD clinics to enhance readiness for TFT (Hamblen et al., 2015), the second aim of this study was to assess whether participating in preparatory treatment led to superior TFT outcomes in all treat- ment seeking veterans with PTSD symptoms (with or without co-occurring SRPs). We predicted that participation in any prepa- ratory treatment would lead to (1) higher rates of TFT completion and (2) superior TFT outcomes compared with veterans who chose direct access to TFT.



The current study is based on archival data of 737 (92.0% male) U.S. military veterans referred for treatment to a Midwestern United States VA outpatient PTSD specialty clinic. The average age was 50.45 (SD � 15.89), and slightly over half self-identified as non-Hispanic White (55.5%). Additionally, 326 (44.2%) veter- ans reported service during the Vietnam era. Traumatic events were defined in accordance with Diagnostic and Statistical Man- ual of Mental Disorders criteria (4th ed., text rev. [DSM-IV-TR]; American Psychiatric Association, 2000). Index trauma was col- lected via the electronic consult to the PTSD specialty clinic, which requires identification of the primary trauma for focus of treatment. Of the veterans referred to the PTSD clinic, 78.6% reported combat trauma as their index trauma.

At the time of PTSD specialty clinic referral, 78.0% of veterans met criteria for probable PTSD (i.e., a total score of 50 or higher on the PTSD Checklist, see below for additional details). A com- prehensive diagnostic assessment for PTSD was not conducted as part of routine clinical practice. As such, reference to PTSD in this study refers to a spectrum of PTSD symptomology rather than a dichotomous (yes/no) diagnostic category.

Upon completion of orientation in the PTSD specialty clinic, the local SUD/PTSD specialist conducted a comprehensive review of the medical chart and assigned each veteran an SRP level, which provided clinicians with a concise summary of the severity of concurrent SRPs. Medical charts were reviewed for all veterans referred to the PTSD clinic from May 2012 through February 2014. The SRP level was determined based on evidence of the current pattern of substance use (gathered from veteran self-report and provider documentation in the medical record), associated consequences or impairments, presence of coping skills and pro- tective factors (e.g., social support, stable housing), and risk status (i.e., suicidality, homicidality, physical withdrawal symptoms). Although not a psychometrically validated scale, the SRP level was developed to improve upon the limitations of SUD diagnoses via the electronic medical chart (e.g., not updated/current) and to highlight additional clinical but nondiagnostic issues (e.g., social support, coping skills). SRP levels ranged in severity from 0 (no current or historical substance-related problems) to 4 (moderate/ severe substance-related problems with acute risk factors). SRP level descriptions are included as supplementary material.

Of the 737 veterans in this study, 325 had no history of SRPs (SRP Level 0); 213 had a lifetime history of SRPs, but none within the past year (SRP Level 1); 138 reported SRPs within the past year, but minimal functional impairment or risk concerns (SRP Level 2); 53 reported current SRPs with moderate/severe impair-

ments and chronic risk factors, but no acute risk concerns (SRP Level 3); and eight reported current SRPs with moderate/severe impairments, chronic risk factors, and current acute risk concerns (SRP Level 4). For data analysis purposes, we combined SRP Levels 3 and 4 (subsequently referred to as SRP Level 3 within this study), resulting in 61 veterans with current moderate/severe SRPs with chronic and/or acute risk factors. Descriptive statistics for demographic variables for the total sample and grouped by SRP level are presented in Table 1.


Veterans with PTSD or subclinical posttraumatic stress symp- toms were referred solely by mental health providers via electronic consult to the VA PTSD specialty clinic. Upon referral, veterans attended an orientation class where they were provided psychoe- ducation about PTSD etiology, symptoms, and recovery-oriented treatment. The orientation class was primarily conducted in a group setting, though individual orientation classes were provided to veterans with a strong preference for an individual format (e.g., a veteran with military sexual trauma requesting an individual format to avoid being around veterans of a specific sex). The clinic was designed with two treatment tracks; trauma-preparatory track (coping-focused groups) or trauma-focused track (individual CPT or PE). At the completion of orientation, veterans self-selected their treatment track and specific treatment choice within that track (e.g., which preparatory group). In the trauma-preparatory track veterans could select from the following coping-focused groups: general coping skills (10 sessions/60 min), anger management (12 sessions/90 min), or emotion management (12 sessions/90 min). These skills-based groups are not trauma-focused, and though not an evidence-based practice, they are typical of treatments offered in this naturalistic setting (Hamblen et al., 2015).

Veterans attended one group at a time, with the option to sequentially attend each available preparatory group. Veterans did not receive concurrent individual therapy within the PTSD clinic while attending preparatory groups. Veterans’ specific group choice could be influenced by desire to attend next available cohort versus wait for their first choice, or a modality preference (e.g., group vs. individual). For the purpose of this study, the length of preparatory group participation is defined as the total number of preparatory group sessions attended. After completion of each group veterans could elect to transition to the trauma- focused track. The veteran’s interest in CPT (12 sessions/60 min) or PE (10 to 12 sessions/90 min) was discussed with his or her individual therapist, with the final choice directed by the veteran. All therapies were provided weekly. As this was a naturalistic design, there were no specified exclusionary criteria; however, veterans who were actively suicidal, homicidal, or demonstrating acute psychotic symptoms would not have been referred to this level of care and were therefore not included in the sample. Therapists followed manualized protocols for all TFT (e.g., CPT; Resick, Monson, & Chard, 2007; PE: Foa, Hembree, & Rothbaum, 2007); however, fidelity to the models was not directly assessed.

Out of the 737 veterans referred for PTSD specialty treatment, 614 (83.3%) chose to initiate services at the completion of orien- tation. Of those 614 veterans, 342 (55.7%) chose to initiate TFT at the time of referral, whereas 272 preferred to initiate a coping- focused group via the trauma-preparatory track. Of the 272 who


initiated trauma-preparatory treatment, 86 (31.6%) eventually chose to transition to the trauma-focused track. In total, 428 (69.7%) veterans referred for PTSD specialty treatment requested to begin TFT (either at the time of referral or after preparatory groups); 247 (57.7%) of those who requested TFT attended at least one session.

Patient information (e.g., outcome measures) was collected at multiple time points and data were maintained in a clinical data repository. Collected patient information was primarily used for clinical and administrative purposes, with a secondary purpose of retrospective analysis for research. Approval for this study was provided by the local institutional review board and a waiver of informed consent for access to protected health information was granted.


All veterans completed a general information form at the PTSD clinic orientation class. This form assessed demographic informa- tion such as age, race, gender, and education level. In addition, the general information form assessed information related to military service history (e.g., era of service) and the traumatic experience

for which the veteran was referred to address in treatment. Indi- vidual therapy clinicians submitted paperwork at the conclusion of TFT and determined TFT completion status (e.g., completed, dropped out). TFT was completed at session numbers ranging from six to 16 within this study, as determined by the treating clinician. For the purpose of the current study, the authors also reviewed each veteran’s medical record to determine participation in a VA SUD treatment program (defined as attendance of at least one session) during the 6 months prior to the PTSD clinic referral date (yes/no) or concurrent treatment in SUD and PTSD clinics (yes/ no).

PTSD and depression symptoms were measured at three sepa- rate time points (orientation, pre- and post-TFT). Symptoms mea- sured at orientation are considered baseline symptoms for all analyses. PTSD symptoms were measured using the PTSD Checklist-Specific Stressor Version for DSM-IV-TR (PCL-S; Weathers, Litz, Huska, & Keane, 1994). This self-report measure has demonstrated good psychometric properties in trauma popu- lations (e.g., Wilkins, Lang, & Norman, 2011) and consists of 17 items answered on a five-point Likert scale. Scores range from 17 to 85, with higher scores indicating greater PTSD severity and a

Table 1 Characteristics of Total Sample and Grouped by SRP Level

SRP Level

Total 0 1 2 3 (N � 737) (n � 325) (n � 213) (n � 138) (n � 61)

Characteristic n or M % or SD n or M % or SD n or M % or SD n or M % or SD n or M % or SD

Age 50.45 15.89 50.99 16.21 54.15 14.64 46.92 15.46 42.61 15.30 Male gender 678 92.0 287 88.3 202 94.8 133 96.4 56 91.8 Racea

Non-Hispanic white 409 55.5 200 61.5 117 54.9 60 43.5 32 52.5 Black 203 27.5 72 22.2 67 31.5 46 33.3 18 29.5 Hispanic/White 88 11.9 33 10.2 24 11.3 23 16.7 7 11.5 Hispanic/Black 11 1.5 6 1.8 0 .0 3 2.2 2 3.3 American Indian/Alaskan native 3 .4 2 .6 0 .0 1 .7 0 .0 Asian 9 1.2 5 1.5 1 .5 1 .7 2 3.3 Other/unknown 13 1.8 6 1.8 3 1.4 4 2.9 0 .0

Education levelb 13.27 2.07 13.51 2.24 13.04 2.09 13.16 1.77 13.00 1.56 Service erac

Pre-Vietnam 10 .1 6 1.8 0 .0 2 1.4 2 3.3 Vietnam 326 44.2 135 41.5 95 44.6 66 47.8 30 49.2 Post-Vietnam 66 9.0 34 10.5 16 7.5 11 8.0 5 8.2 Persian gulf 64 8.7 31 9.5 22 10.3 7 5.1 4 6.6 OEF/OIF/OND 269 36.5 118 36.3 80 37.6 51 37.0 20 32.8

Index traumad

Combat 579 78.6 257 79.1 172 80.8 105 76.1 45 73.8 MST 55 7.5 28 8.6 10 4.7 9 6.5 8 13.1 Adult physical assault 16 2.2 7 2.2 6 2.8 2 1.4 1 1.6 Childhood physical assault 8 1.1 4 1.2 1 .5 2 1.4 1 1.6 Childhood sexual abuse 10 1.4 4 1.2 2 .9 3 2.2 1 1.6 Adult sexual assault 1 .1 0 .0 1 .5 0 .0 0 .0 Motor vehicle accident 9 1.2 2 .6 3 1.4 2 1.4 2 3.3 Other 57 7.7 23 7.1 17 8.0 15 7.0 2 3.3

Recent SUD treatment 63 8.5 0 .0 8 3.8 30 21.7 25 41.0 Concurrent SUD treatment 54 7.3 0 .0 4 1.9 26 18.8 24 39.3

Note. SRP � substance-related problems; MST � military sexual trauma; OEF/OIF/OND � Operation Enduring Freedom/Operation Iraqi Freedom/ Operation New Dawn; SRP Level 0 � no history of SRP; SRP Level 1 � lifetime history of SRP, none within the past year; SRP Level 2 � SRP within the last year with minimal functional impairment or risk concerns; SRP Level 3 � current moderate/severe SRP with chronic and/or acute risk factors. a One veteran did not provide data on race. b 10 veterans did not provide data on education level. c Two veterans did not provide data on service era. d Two electronic consults did not provide data on index trauma.


score of 50 or higher indicating “probable” PTSD (Weathers et al., 1994). The PCL-S had good internal consistencies measured across all time points (� range � .89 –.97).

The Beck Depression Inventory II (BDI-II) measures emotional (e.g., loss of pleasure), cognitive (e.g., concentration), and physical (e.g., tiredness) symptoms of depression, including suicidal thoughts. The measure has good psychometric properties (Beck, Steer, & Brown, 1996) and consists of 21 items each answered on a 4-point Likert scale. Scores range from 0 to 63, with higher scores indicating more severe levels of depression. The BDI-II had good internal consistencies measured across all time points (� range � .93–.95).

Data Analysis

A series of preliminary analyses of variances (ANOVAs) and chi-square tests were run to assess baseline PTSD and depression symptom differences among the different SRP levels, as well as to assess demographic differences in initial treatment choice. We ran a logistic regression analysis to test the first hypothesis (i.e., veterans with PTSD symptoms and greater SRPs would be more likely to choose preparatory treatment). Treatment choice was a dichotomous variable coded as 1 � TFT and 2 � trauma-prep. A one-way analysis of covariance (ANCOVA) tested the second hypothesis (i.e., veterans with PTSD symptoms and greater SRPs would be more likely to participate in a longer course of prepara- tory treatment prior to TFT). We again utilized a logistic regres- sion analysis to test the third hypothesis (i.e., veterans with PTSD symptoms and greater SRPs would be more likely to complete TFT). TFT completion was coded as 1 � completed and 0 � not completed. Finally, we conducted two separate factorial ANOVAs (SRP Level � Preparatory Group Participation) to test the final hypotheses concerning TFT outcomes. We collapsed race into two groups (non-Hispanic White [coded as 1] and racial minorities [coded as 0]) for all analyses.

Regarding the current study’s second aim, we ran a one-sample t test to assess veterans’ initial choice for preparatory treatment or TFT following the orientation class. In addition, we analyzed results from the logistic regression (used to Test Hypothesis 3) and the two factorial ANOVAs (used to Test Hypothesis 4) to assess whether engagement in preparatory treatment increased engage- ment in and completion of TFT in all treatment seeking veterans.

Furthermore, we assessed whether VA SUD treatment (outpa- tient or residential) 6 months prior to the PTSD clinic referral or concurrent with PTSD clinic services were related to the current study’s outcome variables. Neither recent nor concurrent SUD treatment were correlated with any outcome variables of interest and were therefore not included as covariates in subsequent anal- yses. Missing data constituted less than five percent of cases for each study variable and were therefore assumed to be missing at random. All analyses were conducted using SPSS Version 21.0 (IBM Corp, 2012).


All data were normally distributed. At orientation, the average PCL-S score was 59.87 (SD � 12.55; n � 727) and the average BDI-II score was 28.02 (SD � 11.60; n � 727). The SRP groups differed in PCL-S score at orientation, F(3, 726) � 7.45, p � .001,

�p 2 � .03. A Tukey post hoc test revealed that, compared with

veterans at SRP Level 0 (M � 57.71, SD � 13.14), veterans at SRP Level 3 (M � 64.53, SD � 11.89; p � .001) and a SRP Level 1 (M � 61.52, SD � 11.64; p � .003) had higher PCL-S total scores. There were no significant differences in PCL-S score between veterans at SRP Level 0 and SRP Level 2, SRP Level 1 and SRP Level 2, SRP Level 1 and SRP Level 3, or SRP Level 2 and SRP Level 3. The SRP level groups also differed in BDI-II total score at orientation, F(3, 727) � 7.80, p � .001, �p2 � .031. A Tukey post hoc test revealed that BDI-II total scores were significantly higher in veterans at SRP Level 3 (M � 32.10, SD � 12.16; p � .001) and SRP Level 1 (M � 29.74, SD � 11.44; p � .001) compared with SRP Level 0 (M � 25.93; SD � 11.09). There were no significant differences between veterans at SRP Level 0 and SRP Level 2, SRP Level 1 and SRP Level 2, SRP Level 1 and SRP Level 3, or SRP Level 2 and SRP Level 3. Because of significant differences in baseline PTSD and depres- sion symptoms as a function of SRP level, these variables were entered as covariates in all subsequent analyses.

Initial Treatment Choice

Veterans who initiated PTSD treatment after orientation were compared with those who declined. Initial treatment choice did not differ as a function of SRP level, �2(3, N � 736) � 7.76, ns. Veterans who reported greater baseline PTSD and depression symptoms were more likely to initiate treatment than veterans who reported fewer baseline PTSD and depression symptoms (PTSD symptoms: F[1, 726] � 5.38, p � .021; depression symptoms: F[1, 727] � 6.72, p � .010). When compared on demographic vari- ables, non-Hispanic White veterans were more likely than racial minorities to initiate treatment, �2(1, N � 735) � 9.29, p � .002. Additionally, veterans who reported greater years of education were more likely to initiate treatment than veterans with fewer years of education, F(1, 725) � 17.67, p � .042.

For the veterans who chose to initiate PTSD treatment after orientation, a one-sample t test demonstrated that veterans were more likely to initially choose individual TFT (n � 342) than preparatory groups (n � 272), t(613) � 71.92, p � .001. A logistic regression was used to predict the probability of veterans choosing preparatory groups (vs. TFT) from their SRP level (see Table 2). Race was correlated with choice of treatment (r � �.09, p � .026)

Table 2 Logistic Regression of Predictors of Choosing Preparatory Groups vs. Individual Trauma-Focused Therapy

Predictor B SE B Wald Odds ratio 95% CI

Race Non-Hispanic white .36 .17 4.61 1.44� [1.03, 2.00]

Baseline PCL-S �.01 .01 .21 .99 [.98, 1.01] Baseline BDI-II �.01 .01 .51 .99 [.97, 1.01] SRP Level 1 .36 .20 3.12 1.43 [.96, 2.13] SRP Level 2 .51 .23 5.00 1.67 [1.07, 2.61] SRP Level 3 .07 .43 .10 .87 [.58, 1.98]

Note. n � 604. CI � confidence interval; PCL-S � Posttraumatic Stress Disorder Checklist–Specific Stressor; BDI-II � Beck Depression Inventory– II; SRP � substance-related problem; reference category was SRP level (0). � p � .008 (adjusted for Bonferroni).

Attachment 6

Journal of Traumatic Stress October 2012, 25, 527–534

CE Article

Modeling PTSD Symptom Clusters, Alcohol Misuse, Anger, and Depression as They Relate to Aggression and Suicidality

in Returning U.S. Veterans

Julianne C. Hellmuth,1 Cynthia A. Stappenbeck,1 Katherine D. Hoerster,2 and Matthew Jakupcak2 1VA Puget Sound Health Care System, Seattle, Washington, USA

2Mental Illness Research, Education, and Clinical Center (MIRECC), VA Puget Sound Health Care System, Seattle, Washington, USA

Suicidal ideation and aggression are common correlates of posttraumatic stress disorder (PTSD) among U.S. Iraq and Afghanistan war veterans. The existing literature has established a strong link between these factors, but a more nuanced understanding of how PTSD influences them is needed. The current study examined the direct and indirect relationships between PTSD symptom clusters and suicidal ideation in general aggression (without a specified target) regarding depression, alcohol misuse, and trait anger. Participants were 359 (92% male) U.S. Iraq/Afghanistan war veterans. Path analysis results suggested that the PTSD numbing cluster was directly (β = .28, p < .01) and indirectly (β = .17, p = .001) related through depression. The PTSD hyperarousal cluster was indirectly related to suicidal ideation through depression (β = .13, p < .001). The PTSD reexperiencing cluster was directly related to aggression (β = .17, p < .05), whereas the PTSD numbing and hyperarousal clusters were indirectly related to aggression through trait anger (β = .05, p < .05; β = .20, p < .001). These findings indicate that adjunct treatments aimed at stabilizing anger, depression, and alcohol misuse may help clinicians ameliorate the maladaptive patterns often observed in veterans. These results also point to specific manifestations of PTSD and co-occurring conditions that may inform clinicians in their attempts to identify at risk veterans and facilitate preventative interventions.

Posttraumatic stress disorder (PTSD) is the most common mental health disorder diagnosed in veterans returning from the Iraq and Afghanistan wars (Hoge, Auchterlonie, & Mil- liken, 2006; Sayers, Farrow, Ross, & Oslin, 2009). Two of the most salient and maladaptive correlates of PTSD are suicidal ideation and aggression (Jakupcak et al., 2009; Kang & Bull- man, 2008; Panagioti, Gooding, & Tarrier, 2009; Pietrzak et al., 2010; Taft, Street, Marshall, Dowdall, & Riggs, 2007). Iraq and Afghanistan war veterans reporting symptoms of PTSD are more likely to report suicidal ideation (Jakupcak et al., 2009)

Julianne C. Hellmuth is now at Department of Psychiatry, Yale University. Cynthia A. Stappenbeck is now at Department of Psychiatry and Behavioral Sciences, University of Washington.

This material is the result of work supported by resources from the VA Puget Sound Health Care System, Seattle, Washington, and the National Institute on Drug Abuse (T32DA019426).

Correspondence concerning this article should be addressed to Julianne C. Hellmuth, Department of Psychiatry, Division of Prevention and Commu- nity Research, 389 Whitney Ave., New Haven, CT 06511. E-mail: Julianne. [email protected]

Published 2012. This article is a US Government work and is in the public domain in the USA. View this article online at wileyonlinelibrary.com DOI: 10.1002/jts.21732

and to perpetrate aggressive acts compared to veterans with- out PTSD (Fontana & Rosenheck, 1995; Teten et al., 2010). Some research has asserted that suicidality and interpersonal aggression can also be characterized as inwardly or outwardly directed aggression (Dyer et al., 2009; Novacco & Chemtob, 2002). Indeed, suicidality and aggression share many common precipitating factors including anger, depression, anxiety, sub- stance use, and PTSD (Jakupcak et al., 2007, 2009; Lemaire & Graham, 2011; Taft et al., 2007). Currently, our understanding of how suicidal ideation and aggression can be differentiated, and therefore, more effectively managed, in clinical contexts is limited because prior studies have not investigated these fac- tors concurrently within Iraq and Afghanistan war veterans. Doing so is an integral step toward better understanding PTSD components that may be common to both suicidal ideation and aggression. It is also essential to improving clinicians’ capac- ity to effectively assess veterans’ treatment needs, as clinicians are employed with the difficult task of preventing these prob- lems while also addressing trauma-related symptoms (Elbogen et al., 2010; Martin, Ghahramanlou-Holloway, Lou, & Tuccia- rone, 2009).

A 4-factor structure representing trauma-related reexperi- encing, avoidance, numbing, and hyperarousal symptoms has garnered strong empirical support (King, Leskin, King, &


528 Hellmuth et al.

Weathers, 1998; Yufik & Simms, 2010). Because this frame- work captures specific components of trauma-related seque- lae, these symptom clusters may differentially inform suicidal ideation and aggression. In an effort to better identify the cir- cumstances under which veterans may be at risk for these dan- gerous problems, the current study examined PTSD symptom clusters and their direct and indirect relationships with suici- dal ideation and aggression via depression, alcohol misuse, and trait anger.

Based on past research (Jakupcak et al., 2007; McFall et al., 1999; Norstrom & Pape, 2010; Orcutt, King, & King, 2003; Savarese, Suvak, King, & King, 2001), we hypothesized that the numbing and hyperarousal symptom clusters, alcohol mis- use, and trait anger would be associated with aggression. Alco- hol misuse, trait anger, and depression were chosen as potential influencing variables in this investigation based on literature indicating that they are highly prevalent in this population and particularly clinically relevant. The presence of these prob- lems often necessitates heightened attention during the course of treatment to facilitate treatment engagement and comple- tion and minimizing the risk for dangerous distress behaviors (Garcia, Kelley, Rentz, & Lee, 2011; Lu, Duckart, O’Mally, & Dobscha, 2011; Trusz, Wagner, Russo, Love, & Zatnick, 2011). Although depression has consistently been shown to be a predictor of suicidal ideation in Iraq and Afghanistan war veterans (Jakupcak et al., 2009; Pietrzak, Goldstein, Malley, Rivers, Johnson, & Southwick, 2010), the remaining literature regarding suicidal ideation is less clear. Some studies have indi- cated that the PTSD numbing cluster is most relevant to suicidal ideation (Guerra & Calhoun, 2010), whereas other research has found that reexperiencing symptoms play a more significant role (Nye & Bell, 2007). Similarly, alcohol misuse is often considered to put some veterans at increased risk for suicidal ideation (Panagioti et al., 2009; Pietrzak, Goldstein, Malley, Rivers, Johnson, & Southwick, 2010), although recent studies have not supported this relationship (Ilgen et al., 2010; Kang & Bullman, 2008). As a result of these conflicting results, we expected both the reexperiencing and numbing symptom clus- ters, along with depression and alcohol misuse, to be associated with suicidal ideation in the proposed model.



The pool for participants comprised U.S. Iraq and Afghanistan war veterans (N = 653; 91% male) presenting to the Deploy- ment Health Clinic of the VA Puget Sound Health Care System between 2004 and 2009. All participants completed assess- ments during their intake appointments. The reason for seeking treatment was not assessed in the intake packets, and in an effort to reduce barriers to mental health services, all veterans pre- senting to this service were screened for physical and mental health problems following their return from military deploy-

ment. Veterans were then offered brief medical and/or mental health interventions to stabilize symptoms, or were referred to appropriate clinics to address long-term treatment needs. The use of deidentified clinic data from the assessment packets for research purposes was approved by the Institutional Review Board and Research & Development Committee of VA Puget Sound Health Care System.

To focus on clinical levels of PTSD most germane to treat- ment settings, the sample was limited to those who reported subthreshold (operationalized in the measures section; n = 135) or threshold PTSD (n = 224) for a total sample of 359. Dif- ferences in primary demographic characteristics between those included versus excluded (n = 294) were examined using chi- square analyses for comparisons of gender and race/ethnicity and analyses of variance (ANOVAs) for comparisons of age. Those excluded from analyses did not differ from those included on gender, χ2(1, N = 645) = .704, p = .40; race/ethnicity, χ2(5, N = 594) = 9.24, p = .10; or age, F(1, 617) = 4.02, p = .81.

The sample was predominantly male (92%), White (65%), married (42%), had a mean age of 30.6 (SD = 8.0) years, and served in the U.S. Army (69%) while deployed to Iraq or Afghanistan. The majority graduated high school (96%) and were employed fulltime (51%). Twenty percent reported an annual combined household income between $0 and $14,999, 25% reported between $15,000 and $24,999, 20% between $25,000 and $34,999, 14% between $35,000 and $49,999, and 18% $50,000 and above.


PTSD symptoms were measured using the PTSD Checklist Military Version (PCL-M; Weathers, Litz, Herman, Huska, & Keane, 1993). Using a 5-point scale, this 17-item measure as- sesses the risk for PTSD diagnosis by determining the frequency and severity of PTSD symptoms defined by the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association, 1994) experienced over the course of the past month, with global scores ranging from 17 to 85. The PCL-M cannot be used to diagnose PTSD in the absence of a clinical interview, and is only suggestive of risk of PTSD diagnosis. Previous research has established a conserva- tive cutoff point of 50 for risk of meeting full diagnostic criteria for combat veterans (Forbes, Creamer, & Biddle, 2001). The present study also included individuals with scores between 35 and 49, classified as at risk for meeting subthreshold PTSD diagnostic criteria. Subthreshold PTSD symptomatology has been shown to elevate risk for suicidality and aggression rela- tive to veterans with scores reflective of no or only limited levels of PTSD (Jakupcak et al., 2007, 2011; Marshall et al., 2001; Yarvis & Scheiss, 2008). Four PTSD symptom clusters were created in accordance with recommendations by King et al. (1998); averages of the respective items were created for re- experiencing, avoidance, numbing, and hyperarousal symptom clusters, ranging from 1 to 5. Cronbach’s α = .90 for the total

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

Modeling Correlates of Aggression and Suicidal Ideation 529

score; for the symptom clusters of reexperiencing, avoidance, numbing, and hyperarousal, the values were .86, .77, .82, and .75, respectively.

Trait anger was assessed using the 10-item Trait Anger Scale (Spielberger, Jacobs, Russel, & Crane, 1983), a subscale of the State-Trait Anger Expression Inventory (Spielberger, 1988) with strong psychometric properties (Spielberger et al., 1983). Participants were asked to use a 4-point scale ranging from 1 = almost never to 4 = almost always to rate the degree to which they react to situations and people in an angry fashion. Responses were summed for a global score. Cronbach’s α = .88 in these data.

Aggression was measured using three items adapted from the National Vietnam Veterans Readjustment Study (see McFall et al., 1999). Participants were asked to indicate whether or not (i.e., yes or no) they threatened the use of violence with or without a weapon, or engaged in a physical fight within the past 4 months. Participants who endorsed at least one of these behaviors were categorized as aggressive. Cronbach’s α = .54 in these data was comparable to the value in prior studies (Jakupcak et al., 2007; McFall et al., 1999).

Depression and suicidal ideation were assessed using the de- pression subscale of the Patient Health Questionnaire (PHQ; Spitzer, Kroenke, & Williams, 1999). The PHQ-9 depression subscale measures the endorsement of nine symptoms of de- pression defined by the DSM-IV diagnostic criteria. Participants responded to a 4-point scale (i.e., 0 = not at all, 1 = several days, 2 = more than half the days, 3 = nearly every day) to indicate the frequency with which each symptom was expe- rienced during the prior 2 weeks. The eight nonsuicide items were summed to form a total depression severity score; internal consistency was α = .88. The following single item was used to assess suicidal ideation: “Over the last 2 weeks, how often have you been bothered by thoughts that you would be better off dead or of hurting yourself in some way?” Single-item versions of the PHQ to screen for suicidality have received empirical support within VA populations (Corson, Gerrity, & Dobscha, 2004; Williams et al., 2004). This item was dichotomized to reflect the presence or absence of suicidal ideation (no = 0, yes = 1).

Alcohol misuse also was assessed with the PHQ-5 (Spitzer et al., 1999). The five items assessing alcohol misuse on the PHQ-5 have demonstrated strong psychometric properties in past studies (Jakupcak et al., 2007; Nunnaly, 1967). Partici- pants who reported any alcohol consumption were asked to indicate whether or not they experienced any of the five symp- toms of alcohol abuse as defined by DSM-IV diagnostic criteria. Cronbach’s α = .75 for this subscale.

Data Analysis

Path analyses were conducted with MPlus 5 (Muthén & Muthén, 1998–2010) to examine the direct and indirect rela- tionships between PTSD symptom clusters and suicidality and aggression. All direct paths between PTSD symptoms clusters

Table 1 Mean, Standard Deviation, and Range of the Symptom Variables

Variable n Observed Range M SD

Reexperiencing 359 1–5 3.14 0.99 Avoidance 359 1–5 3.34 1.12 Numbing 359 1–5 3.01 1.00 Hyperarousal 359 1–5 3.66 0.84 Alcohol misuse 351 0–5 0.63 1.17 Trait anger 358 1–4 2.22 0.67 Depression 356 0–24 13.55 6.10

and suicidal ideation and aggression and indirect paths through alcohol misuse, trait anger, and depression were included in the model. A test of the direct and indirect effects was examined. Additionally, the disturbances among the PTSD symptom clus- ters and between suicidal ideation and aggression were corre- lated given the strong interrelations observed. Because the risk variables were dichotomous, probit regression coefficients were calculated for all variables regressed on suicidal ideation and aggression, and can be interpreted as the amount of change in the dependent variable (i.e., suicidal ideation and aggression) for a one-unit change in the independent variable.

Among the advantages of MPlus is the use of maximum likelihood estimation procedures, which can effectively handle the nonsystematic missing data in the current dataset (Kline, 2005). Only 2.2% of the sample had any missing data. We used weighted least squares with robust standard errors and mean- and variance-adjusted chi-square as an estimator, and assessed overall model fit by examining the model chi-square statistic, which can be erroneously significant with large sample sizes such as the one included in the current study (Kline, 2005). Other fit indices were also examined, including the root mean square error of approximation (RMSEA), which indicates a reasonable fit with values around .08 (Browne & Cudeck, 1993), and the comparative fit index (CFI) and Tucker-Lewis Index (TLI), which indicate reasonably good fit with values greater than .90 (Hu & Bentler, 1999).


Approximately one third of our sample reported suicidal ideation (32.3%), at least one act of physically aggressive behavior (31.8%), and at least one alcohol-related problem (34.3%). Of the sample, 16.4% reported suicidal ideation only, 15.9% reported aggressive behavior only, and 15.9% reported both suicidal ideation and aggressive behavior. Ninety-eight (27.7%) participants endorsed less-severe physical aggression (i.e., threatening without a weapon), and 68 (18.9%) partici- pants endorsed more-severe physical aggression (i.e., physical fight and threatening with a weapon). Average scores for PTSD symptom clusters, alcohol misuse, trait anger, and depression are shown in Table 1. Table 2 presents the bivariate correlations among all study variables.

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

530 Hellmuth et al.

Table 2 Correlations Among all Study Variables

Variable 1 2 3 4 5 6 7 8

1. Suicidal ideation - - - - - - - - 2. Aggression .26 - - - - - - - 3. PTSD reexperiencing .21 .24 - - - - - - 4. PTSD avoidance .19 .19 .64 - - - - - 5. PTSD numbing .42 .22 .43 .49 - - - - 6. PTSD hyperarousal .32 .24 .53 .49 .60 - - - 7. Alcohol misuse .14 .25 .13 .14 .17 .17 - - 8. Trait anger .25 .36 .29 .23 .42 .53 .23 - 9. Depression .47 .19 .40 .39 .63 .62 .15 .32

Note. Sample size range = 351–359. Significant of all coefficients ranges from p < .05 to p < .001. PTSD = posttraumatic stress disorder.

The overall path analysis model fit the data well, χ2(2) = 7.97, p = .02, CFI = .98; TLI = .90; RMSEA = .09. The significant paths from the full model are presented in Figure 1. The standardized coefficients of the direct and indirect relations between PTSD symptom clusters and suicidal ideation and ag- gression are shown in Table 3. The PTSD numbing cluster was directly and indirectly related to suicidal ideation through de- pression, indicating that PTSD numbing was associated with suicidal ideation whether or not other symptoms of depression were present. The PTSD hyperarousal cluster was indirectly related to suicidal ideation through depression, suggesting that hyperarousal was only related to suicidal ideation when other symptoms of depression were endorsed. There were direct asso- ciations between PTSD reexperiencing symptoms and alcohol misuse and veterans’ aggressive behavior. Those with more re- experiencing symptoms and those with alcohol problems were more likely to report aggressive behavior. Numbing and hyper- arousal symptoms were only associated with aggression when veterans endorsed higher levels of trait anger. Although the PTSD avoidance cluster was significantly and negatively asso- ciated with trait anger, which was significantly associated with

aggression in the path analysis, this indirect path was nonsignif- icant overall.


This investigation is the first of which we are aware to link the perspectives of two emerging literatures specific to return- ing Iraq and Afghanistan war veterans: one that focuses on the relationship between PTSD and suicidal ideation, and another that focuses on the relationship between PTSD and aggressive behavior. The purpose of this report was to illuminate how the specific features of PTSD, in combination with other common clinical characteristics, are differentially related to both suici- dal ideation and aggression. Results indicate that aggression was directly related to PTSD reexperiencing symptoms and alcohol misuse, and suicidal ideation was directly related to PTSD numbing symptoms. Additionally, the PTSD numbing and hyperarousal clusters were related to suicidal ideation and aggression through distinct pathways involving depression and trait anger, respectively. Although reexperiencing and avoid- ance symptoms are often considered hallmark characteristics of PTSD, these results underscore the role that numbing and hyper- arousal symptoms play in PTSD-related impairments (Lunney & Schnurr, 2007; Pietrzak, Goldstein, Malley, Rivers, & South- wick, 2010). In light of the likely overlap between PTSD hyper- arousal symptoms and aggression, the PCL-M item within the hyperarousal subscale assessing angry outbursts was removed and the original model was rerun. Although the PTSD hyper- arousal cluster no longer predicted alcohol misuse, all other significant paths remained significant, including the indirect ef- fect of hyperarousal on aggression through trait anger, lending strength to the findings presented here.

Although alcohol misuse and suicidal ideation were corre- lated in our preliminary analyses, they were not related in the context of other variables in this study. Contrary to results from other studies (Ilgen et al., 2010; Kang & Bullman, 2008), our finding may be an effect of the relatively small number of alco- hol problems reported by this sample. The associations between PTSD, alcohol misuse, and suicidal ideation needs further in- vestigation. Although avoidance was associated with trait anger,

Table 3 Standardized Coefficients of the Direct and Indirect Relations Between PTSD Symptom Clusters and Suicidal Ideation and Aggression

Suicidal ideation Aggression

Variable Alcohol Anger Depression Direct effect Alcohol Anger Depression Direct effect

Reexperiencing .00 .00 .02 .02 .00 .02 .00 .17*

Avoidance .00 −.01 .01 −.07 .00 −.05 .00 .03 Numbing .00 .01 .17*** .28** .01 .05* .00 .04 Hyperarousal .01 .03 .13*** −.03 .04 .20*** .00 −.10 Note. N = 359. Due to the small number of missing values on four of our measures in this path analyses, weighted least squares with robust standard errors were used. PTSD = Posttraumatic stress disorder. *p < .05. **p < .01. ***p < .001.

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

Modeling Correlates of Aggression and Suicidal Ideation 531

PTSD Symptom Clusters











Alcohol misuse

Trait anger


Suicidal ideation














Figure 1. All significant paths shown with unstandardized probit coefficients for paths to the dichotomous outcome variables of suicidal ideation and aggression, and unstandardized regression coefficients for all other paths. N = 359. *p < .05. **p < .01. ***p < .001.

it did not have a direct or indirect association with aggressive behavior in the current study. Further exploration of the possible indirect association between avoidance and aggression through trait anger is warranted.

The emergence of distinct pathways between PTSD symp- tom clusters and suicidal ideation and aggression emphasizes the value of raising clinicians’ awareness of the differing etiolo- gies that contribute to each of these problems. Some literature has identified internalizing and externalizing types of PTSD (e.g., Miller, 2003) and others have acknowledged the common origins shared by suicidal ideation and aggression (Dyer et al., 2009; Novacco & Chemtob, 2002). Our data, however, indicate that some veterans experience both suicidal ideation and ag- gression. These data cannot distinguish whether veterans might transition from experiencing suicidal ideation only to aggres- sion only to experiencing them concurrently, or what factors might precipitate those transitions, due to the cross-sectional nature of these data as well as the different time frame of the measures administered. The fact that some of the veterans in our sample experienced both suicidal ideation and aggression high- lights the need for future studies to explore these relationships further.

Disentangling these complicated relationships and facilitat- ing the knowledge base through which clinicians conduct risk assessments and treatment planning may help them evaluate and stabilize targeted symptoms and prevent suicidal ideation and

aggression associated with PTSD symptoms (Elbogen et al., 2010; Fontana & Rosenheck, 1995; Jakupcak & Varra, 2010; Pietrzak, Goldstein, Malley, Rivers, & Southwick, 2010). Tak- ing the pathways demonstrated by this study into considera- tion may aid some clinicians in creating or adjusting treatment plans, particularly among more complicated cases. For exam- ple, Taylor et al. (2001) found that prominent PTSD numbing and anger symptoms at pre treatment were associated with par- tial response to cognitive behavioral therapy for PTSD. Under the circumstances where the combinations of features associ- ated with suicidal ideation or aggression demonstrated here are present, integrating techniques from modalities such as Dialec- tical Behavior Therapy or Cognitive Behavioral Therapy may be of use. Perhaps utilizing adjunct interventions to manage suicidal ideation, depression, and anger may help to stabilize suicidality and aggression prior to initiating trauma-focused therapies, which could possibly improve treatment outcome for some veterans (Britton, Patrick, Wenzel, & Williams, 2011; Chemtob, Novaco, Hamada, Gross, & Smith, 1997; Harned, Jackson, Comtois, & Linehan, 2010; Jakupcak & Varra, 2010).


Several factors limit the extent to which these results may be generalized. The sample comprised veterans who were seek- ing postdeployment VA care within an integrated clinic and the

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

532 Hellmuth et al.

reason for treatment-seeking was not captured within this data set; as such, findings may not generalize to veterans not en- rolled in VA care, those not reporting military-related traumas, or those who are seeking specialized VA treatment. Although we attempted to run the model with the full sample of vet- erans for whom we had data (i.e., including those reporting PCL-M scores that fell below subthreshold for PTSD diagnosis risk), the model fit was poor and therefore could not be inter- preted. Therefore, these results may only generalize to those representing threshold and subthreshold levels of PTSD symp- tomatology.

All data were collected via self-report measures as opposed to structured diagnostic interviews, and therefore may be sub- ject to self-report bias. Further, PCL-M scores are suggestive, but not necessarily indicative of PTSD diagnosis in the ab- sence of a clinical interview. The cross-sectional nature of this study and the fact that veterans reported on suicidal ideation and aggression in slightly different time frames prohibits us from determining causal relationships between variables. Fu- ture studies would benefit from modeling these relationships over time. Other influential factors (e.g., trauma history and severity, social isolation, stigma, personal coping styles) may contribute to explaining these problems (Ilgen et al., 2010; Pietrzak, Russo, et al., 2010; Pietrzak et al., 2010) and were not measured in the present study.

The low internal consistency demonstrated by the measure of aggression used in this study highlights the need for future research to include more sophisticated measures of aggression in veteran populations. We, however, examined the individ- ual aggression items that had enough endorsement (threatening without a weapon and getting into a physical fight) as outcomes in the model. With one exception (i.e., that the PTSD reexpe- riencing cluster did not directly predict getting into a physical fight), all other direct and indirect paths were identical for both items compared to the original model. Different PTSD symp- toms may be related to different types or severity of aggression, and the present study’s capacity to address that question is lim- ited by a restrictive measure of aggression. Future studies would benefit from using a more comprehensive measure of aggres- sion to capture a wider range of physically and psychologically aggressive behaviors.

The association between depression and suicidal ideation was expected and may be inflated due to shared method variance as both were measured using items of the PHQ-9 depression subscale. Future studies should use a more comprehensive and distinct measure to assess suicidality to improve specificity of findings. Findings also need to be replicated in larger samples, with female veterans, and with veterans from other eras of service.


The present study’s findings indicate that numbing, reexpe- riencing, and hyperarousal PTSD symptom clusters are most strongly associated with suicidal ideation and aggression.

Numbing and hyperarousal symptom clusters distinguished these high-risk problems through pathways involving trait anger and depression whereas reexperiencing symptoms had a direct effect on aggression. These findings target a need identified by Elbogen and colleagues (2010) for more empirically supported frameworks that may help clinicians more effectively conduct risk assessment and decision making in their clinical work with veterans. These findings may help clinicians identify veterans who are at elevated risk for problems related to PTSD and to tailor interventions to more effectively meet their treatment needs. Adjunct interventions aimed at reducing alcohol misuse, depression, and anger in veterans at elevated risk for suicide and aggression may be useful in augmenting care for more positive treatment outcomes. Such interventions may facilitate stabilization in coordination with trauma-focused therapies for some veterans.

References American Psychiatric Association. (1994). Diagnostic and statistical manual

of mental disorders (4th ed.). Washington, DC: Author.

Britton, P. C., Patrick, H., Wenzel, A., & Williams, G. C. (2011). Integrating motivational interviewing …