PTSD and Suicidal Ideation: An Application of the Interpersonal-Psychological Theory of 
Suicide 
 
by 
 
Margaret Taylor Davis 
 
 
 
 
A thesis submitted to the Graduate Faculty of 
Auburn University 
in partial fulfillment of the 
requirements for the Degree of 
Master of Science 
 
Auburn, Alabama 
August 3, 2013 
 
 
 
 
Copyright 2013 by Margaret Taylor Davis 
 
 
Approved by 
 
Frank W. Weathers, Co-chair, Professor of Psychology 
Tracy K. Witte, Co-chair, Assistant Professor of Psychology 
Malissa Clark, Assistant Professor of Psychology 
  
   
 
ii 
 
 
 
Abstract 
 
The relationship between posttraumatic stress disorder (PTSD) and passive suicidal 
ideation (P-SI) is well established, but the mechanisms underlying it are not well understood 
(Panagioti, Gooding, & Tarrier, 2012).This study sought to enhance understanding of the 
PTSD/P-SI relationship using the interpersonal theory of suicide (IPTS; Joiner, 2005) as a 
conceptual framework. It was hypothesized that the strongest relationship between individual 
PTSD clusters and P-SI would be found for the emotional numbing symptoms, but that this 
relationship would be fully mediated by one or both IPTS constructs specified as proximal 
causes for SI: thwarted belongingness and perceived burdensomeness. Participants were trauma-
exposed college students (N = 334). Hypotheses were tested using structural equation modeling 
while controlling for gender and sadness. Although the numbing cluster had the strongest 
bivariate relationship with P-SI, full mediation through one or both IPTS constructs was not 
observed. Instead, numbing and reexperiencing were both directly related to P-SI, and all PTSD 
clusters had indirect relationships with P-SI, partially mediated through one or both IPTS 
constructs. Results, while not entirely consistent with hypotheses, support the utility of the IPTS 
for understanding P-SI among those with PTSD, and point to a unique relationship between P-SI 
and numbing. 
     
iii 
 
 
 
Table of Contents 
 
Abstract?????????????????????????????????.......ii  
List of Tables   ............................................................................................................................... iv  
List of Figures   ................................................................................................................................v  
Introduction   ....................................................................................................................................1 
Method   .........................................................................................................................................10 
 Participants  and Procedure ................................................................................................10 
 Measures   ..........................................................................................................................10 
Results  ...........................................................................................................................................14 
 Descriptive Statistics   ........................................................................................................14 
 Preliminary Analyses .........................................................................................................16 
 Hypothesis 1: Evaluation of PTSD Factor Structure .........................................................17 
 Hypotheses 2 and 3: Evaluation of the Full Model  ..........................................................17 
 Models with Individual Symptom Clusters   .....................................................................21 
Discussion  .....................................................................................................................................24 
References   ....................................................................................................................................31 
Appendix A: Tables   .....................................................................................................................40 
Appendix B: Figures   ....................................................................................................................55 
Appendix C: Item Content  ............................................................................................................58 
 
     
iv 
 
 
 
 
List of Tables 
 
Table A1: Descriptive Statistics for Items Used in Structural Analyses????????? ...41 
Table A2: PTSD Checklist Inter-item Correlations?????????????????. ..44 
Table A3: Bivariate Correlations Between PTSD Checklist Items and Other Study Variables...45 
 
Tables A4 Bivariate Correlations Between Sadness (BDI-II Item 1), and Items of the 
Interpersonal Needs Questionnaire, and Beck Suicide Scale?????????????... .46 
 
Table A5: Fit Statistics for Three Empirically Supported Models of PTSD Symptom 
Structure?????????????????????????????????... ..47 
 
Table A6: Standardized and Unstandardized Covariance Estimates for the Full Measurement 
Model??????????????????????????????????... ..48 
 
Table A7: Standardized and Unstandardized Parameter Estimates for the Full Structural 
Model??????????????????????????????????... ..50 
 
Table A8: Model Fit Statistics for Models Including Single PTSD Symptom Clusters???. ..52 
Table A9: Standardized and Unstandardized Parameter Estimates for Structural Models 
Containing Individual PTSD Symptom Clusters?????????????????? ...53 
 
Table A10: Indirect Effects of Each PTSD symptom cluster on Suicidal Ideation Through 
Perceived Burdensomeness and Thwarted Belongingness?????????????? ? 54 
 
Table C1: PCL Item Mapping for Numbing, Dysphoria, and Dysphoric Arousal CFA 
Models???????? ? ?????????????????????????? 59 
Table C2: Content Descriptors For All Items From the Interpersonal Need Questionnaire, Beck 
Suicide Scale, and Beck Depression Inventory (Second Edition) Used in 
Analyses???? ?????????????????????????????.. ...60 
     
 
v 
 
 
 
List of Figures 
 
Figure B1: Full Structural Model??????????????? ????????? .56 
Figure B2: Individual Symptom Cluster Models with all Specified Paths? ?? ?????.. 57 
 
 
 
 
 
 
1 
 
 
 
Introduction 
A growing body of research has established a positive relationship between posttraumatic 
stress disorder (PTSD) and all forms of suicidal behavior, including suicidal ideation (SI; 
Panagioti, Gooding, & Tarrier, 2012). Several large epidemiological studies, including the 
European Study of Mental Disorders (ESMED; Bernal et al., 2007) and World Health 
Organization (WHO) International Mental Health survey (Nock et al., 2009) have identified 
PTSD diagnostic status as a significant risk factor for SI in both developed and developing 
countries. More recently, meta-analyses by Panagioti, Gooding, and Tarrier (2012) and 
Krysinska and Lester (2010) have confirmed the link between PTSD and SI, finding it to be 
consistent across age, gender, and trauma type. Krysinska and Lester further found that the 
relationship between PTSD and SI is non-spurious in that it is not fully accounted for by 
psychiatric comorbidity, including, most relevantly, major depressive disorder (MDD). This 
finding is particularly important because PTSD and MDD are highly comorbid (Breslau, Davis, 
Peterson, & Schultz, 2000), leading some researchers to argue that comorbid MDD accounts for 
the association between PTSD and SI (e.g., Oquendo et al., 2005).  
Although the relationship between PTSD and SI has been widely studied, the underlying 
mechanisms are not well understood (Panagioti, Gooding, Dunn, & Tarrier, 2011). Most research 
in this area has been descriptive, examining the presence, magnitude, and consistency of the 
PTSD-SI relationship but not addressing its causes. Two key limitations have hampered the 
investigation of causal factors. First, most studies have examined PTSD as a syndrome, despite 
consensus in the trauma field that PTSD is a heterogeneous disorder characterized by several 
 
 
2 
 
distinct but related clusters of symptoms (Wilson, 2004; Elhai & Palmieri, 2011). Failure to 
consider PTSD at the symptom cluster level may obscure a pattern of differential associations 
between SI and specific PTSD symptoms clusters (Guerra & Calhoun, 2010). Second, research 
on the PTSD-SI relationship has been conducted outside the framework of existing theory 
capable of explaining the development of SI (Panagioti et al., 2009).  
Unresolved questions regarding the core symptoms and latent structure of PTSD (Elhai & 
Palmieri, 2011) complicate study of both the disorder and its relationship to correlates such as SI. 
The Diagnostic and Statistical Manual, Fourth Edition, Text Revision (DSM-IV-TR; APA, 2002) 
implies a three-factor model of PTSD, grouping symptoms into three clusters: reexperiencing the 
trauma, avoidance of trauma-related cues and numbing of general responsiveness, and 
hyperarousal. However, confirmatory factor analyses (CFAs) support two competing four-factor 
models (Elhai & Palmieri, 2011). The first four-factor model, the numbing model (King, Leskin, 
King, & Weathers, 1998), separates avoidance and numbing into two clusters. Avoidance and 
numbing have been shown to be differentially related to treatment responsiveness and a variety 
of other types of psychopathology including SI (Asmundson, Stapleton, & Taylor, 2004), and are 
now separated into distinct clusters in DSM-5 PTSD criteria (APA, 2013). The second four-factor 
model, the dysphoria model, also separates avoidance from numbing, but combines numbing 
symptoms with three hyperarousal symptoms to form a dysphoria cluster (Simms et al, 2002). 
Recently, in an effort to resolve differences between the numbing and dysphoria models, Elhai 
and colleagues (2011) proposed a five-factor model, the dysphoric arousal model, which 
separates avoidance and numbing but also divides the hyperarousal cluster into anxious and 
dysphoric arousal clusters. A number of studies have shown that the dysphoric arousal model 
 
 
3 
 
provides superior fit compared to the numbing and dysphoria models (e.g., Carter, 2011; Wang, 
Long, Li, & Armour, 2011; Reddy, Anderson, Liebschultz, & Stein, in press).  
Several recent studies have examined the link between specific PTSD symptom clusters 
and SI with contradictory results. Tarrier and Gregg (2004) found that both hyperarousal and 
reexperiencing were positively associated with a composite variable used as an index of suicidal 
behavior (including SI, plans, and attempts). Others have found a positive association between SI 
or composite variables and either hyperarousal (Ben-Ya?acov & Amir, 2004; Steryn, Vawda, 
Wyatt, Williams, & Madu, 2013) or re-experiencing (Bell & Nye, 2007), but not both. Recently, 
Panagioti, Gooding, Dunn, and Tarrier (2011) re-analyzed the data from Tarrier and Gregg 
(2004), incorporating additional variables into their model (e.g., social and occupational 
impairment). They found that avoidance/numbing was indirectly positively associated with a 
composite variable of suicidal behavior through its association with impairment in social 
functioning. No other individual cluster was associated with suicidal behavior.  
In contrast, two studies utilizing multiple regression have found a negative relationship 
between avoidance/numbing and SI (Monteith, Mnefee, Pettit, Leopoulos, & Vincent, 2013), or a 
composite variable including SI (Ben-Ya?acov & Amir, 2004). However, their findings may be 
attributable to multicollinearity: In both studies, the regression coefficient corresponding to 
avoidance/numbing was negative in the prediction of SI, whereas the zero-order correlation 
between avoidance/numbing and SI was moderate and positive. Avoidance/numbing was also 
highly correlated with other variables included in the regression models in both studies, which 
may have caused the observed flipping of signs between the bivariate relationships and structural 
coefficients. This suggests that results may be biased by the presence of multicollinearity and 
thus should not be interpreted (Cheung & Lau, 2008).  
 
 
4 
 
Although the findings concerning the relationship between PTSD clusters and SI are mixed, 
none of the noted studies found support for a direct, positive relationship between 
avoidance/numbing and SI. This may due to use of the DSM-IV-TR three-factor model which 
conflates avoidance and numbing. Consistent with that possibility, two other studies that used the 
numbing model found that only numbing was associated with SI when all other PTSD symptom 
clusters and depressive symptoms were controlled for (Guerra & Calhoun, 2011; Hellmuth, 
Stappenbeck, Hoerster, & Jakupcak, 2012). Moreover, one of these studies (Guerra & Calhoun, 
2011) found that the numbing cluster had the largest bivariate relationship with SI, whereas the 
avoidance cluster had the smallest. Additionally, a currently unpublished study (Davis, Witte, & 
Weathers, 2013) examined the relationship between individual PTSD symptoms and SI. 
Numbing symptoms had the highest bivariate associations with SI.  Further, the numbing 
symptoms feelings of detachment or estrangement from others and sense of foreshortened future 
were the only PTSD symptoms with a significant positive association with SI after controlling 
for depressive symptoms, response bias, and all other PTSD symptoms. Thus, distinguishing 
between numbing and avoidance may be especially important when examining the relationship 
between PTSD symptoms and SI. However, deeper understanding of how and why high levels of 
numbing, or any other PTSD cluster might lead to the development of SI requires reference to 
theory capable of explaining how and why SI develops (Panagioti et al., 2009).  
One model with the potential to guide a theoretically informed investigation of the 
relationship between PTSD and suicidal behavior is the interpersonal-psychological theory of 
suicide (IPTS; Joiner, 2005; Van Orden et al., 2010). Unlike other models, the IPTS corresponds 
with the observed taxonomy of suicidal behavior and proposes distinct constructs to account for 
the prediction of different forms of suicidal behavior, including SI, suicide attempts (SA), and 
 
 
5 
 
death by suicide. These constructs are conceptualized as causal processes mediating the 
relationship between many specific risk factors (e.g., guilt and lack of social support in those 
with PTSD) and suicidal behavior. The theory postulates two constructs, perceived 
burdensomeness (PB) and thwarted belongingness (TB), as sufficient, proximal causes of SI. 
Perceived burdensomeness is defined as the perception or belief that one is a burden to others 
and that others would be better off without oneself. Likely contributors to the development of PB 
include loss of employment and functional impairment (Inoue et al., 2007; Kaplan, McFarland, 
Huguet, & Newsom, 2007). The second construct, TB, is hypothesized to develop when the 
fundamental human need to belong socially and to be meaningfully connected to others is not 
satisfied (Baumeister & Leary, 1995). Indicators of TB include isolation, loneliness, and lack of 
social support (Van Orden et al., 2010).  
The IPTS holds that PB and TB constitute distinct but related constructs, each of which 
serves as a proximal cause for the development of passive SI (i.e., thoughts of suicide without 
intent) when present in isolation. The simultaneous presence of PB, TB, and hopelessness serves 
as the proximal cause for active SI (Van Orden et al. 2010). In a review paper describing the 
IPTS, Van Orden and colleagues (2010) noted that the theory currently describes IPTS 
constructs, including PB and TB, as sufficient but not necessary for the development of various 
forms of suicidal behavior. They also describe an alternative testable hypothesis: IPTS constructs 
are necessary for the development of suicidal behavior (Van Orden et al., 2010). If this 
alternative were true, that would imply that the relationship between any suicide risk factor and 
suicidal behavior is fully accounted for by the IPTS constructs. 
Compared to other theoretical models attempting to account for suicidal behavior, the 
IPTS has advantages in terms of its ability to predict suicidal outcomes with maximal sensitivity 
 
 
6 
 
and specificity. First, the IPTS provides precise definitions of both its proposed constructs and 
the hypothesized causal pathways of those constructs to specific suicidal behavior. The level of 
detailed explanation provided enhances both the theory?s testability and clinical usefulness. 
Second, the IPTS differentially accounts for various forms of suicidal behavior and does so using 
hypothesized constructs whose rarity, or rarity in combination, is consistent with the observed 
frequency of the corresponding suicidal behaviors they predict (Van Orden et al., 2010). For 
example, the experience of passive SI is more common than active SI, and thus correspondingly 
in the IPTS experiencing either PB or TB is more common than experiencing PB, TB, and 
hopelessness together.  
In addition to its general advantages, the IPTS seems well suited for explaining SI among 
individuals with PTSD. In contrast to the inconsistent results found in studies examining 
relationships between the DSM-IV three-factor model and SI, the few studies that have not 
conflated avoidance with numbing have demonstrated a unique association between numbing 
symptoms and SI (Davis et al., 2013; Guerra & Calhoun, 2011; Hellmuth et al., 2012). This 
pattern of results fits within the IPTS framework in that numbing symptoms seem theoretically 
consistent as indicators of the presence of PB and TB. For example, detachment/estrangement, 
which is part of the numbing cluster, is the only PTSD symptom that directly refers to disrupted 
interpersonal connection, which both PB and TB reflect. Indeed, Davis et al. (2013) found that 
detachment/estrangement had a stronger association with SI than any other PTSD symptom.  
As noted earlier, most existing research on the relationship between PTSD and SI has been 
atheorethical, with a few recent exceptions. To date, six published studies have used the IPTS 
framework to examine the relationship between PTSD and suicidal behavior. The first two 
studies -- Brenner et al. (2008) and Anestis, Bryan, Cornette, and Joiner (2009) -- used 
 
 
7 
 
qualitative data to test applicability of the theory?s SI-relevant constructs in returning veterans. 
Both found support for the utility of the IPTS in understanding suicidal behavior among those 
with PTSD. Of the existing quantitative studies, two (Bryan, Cukrowicz, West, & Morrow, 2010; 
Bryan, Hernandez, Allison, & Clemans, 2012) are limited in that they focused on overall PTSD 
symptom severity, as opposed to PTSD symptom clusters. The two studies that have considered 
PTSD symptom clusters are also limited in that one (i.e., Monteith et al., 2013) utilized the DSM-
IV three-factor model, and the other (i.e., Bryan & Anestis, 2011) only reported results for the 
reexperiencing symptom cluster.  
Despite their limitations, the aforementioned studies provide mixed support for the notion 
that PTSD symptoms are associated with PB (Bryan et al., 2010; Monteith et al., 2013) and TB 
(Bryan et al., 2010; Bryan & Anestis, 2011; Bryan et al., 2012, Study 1; Monteith et al., 2013) . 
Further, there is some evidence that PB (Bryan et al., 2010; Bryan et al., 2012, Study 2; Monteith 
et al., 2013) and TB (Bryan et al., 2010; Bryan et al., 2012, Study 2) are associated with suicidal 
behavior even after accounting for depressive and PTSD symptoms. Regarding the question of 
whether PB and TB fully account for the relationship between PTSD symptoms and suicidal 
behavior, one study found that PTSD symptoms are still associated with suicidal behavior after 
accounting for PB and TB (Bryan et al., 2010), whereas both studies reported in Bryan et al. 
(2012) found no association between PTSD symptoms and suicidal behavior after accounting for 
PB and TB.  
In sum, a growing body of research has examined the utility of the IPTS in explaining the 
development of SI in a PTSD population, but with significant limitations. The aim of this study 
was to address gaps in the current literature by using the IPTS to identify mechanisms underlying 
the relationship between PTSD and SI. More specifically, the primary objective of the present 
 
 
8 
 
study was to examine the differential relationships between specific PTSD symptom clusters, 
PB, TB, and SI. Like the study by Monteith and colleagues (2013), the present study examined 
the relationship between IPTS constructs (PB and TB), SI, and specific PTSD symptom clusters 
in a trauma-exposed population. However, the study design differed in several important ways.  
First, Monteith and colleagues used a regression framework, whereas the present study used 
structural equation modeling (SEM) to test hypotheses. Use of SEM allowed for simultaneous 
testing a priori hypotheses regarding direct and indirect relationships between specific PTSD 
symptom clusters, interpersonal model constructs (PB and TB), and SI. Second, Monteith and 
colleagues combined passive SI (thoughts of suicide) and active SI (thoughts, desire, intent, and 
plans) in their outcome variable, whereas the present study focused exclusively on passive SI (P-
SI). Combining the two types of SI is inconsistent with the IPTS, which states that each type is 
predicted by different conditions. Given interest in the individual relationships between PB, TB 
and PTSD clusters, hypotheses and analyses in this study focused exclusively on P-SI, which 
according to the theory does not require the simultaneous presence of PB and TB. Third, 
Monteith and colleagues used the three-factor DSM-IV model of PTSD, whereas the current 
study used CFAs to evaluate fit of the three PTSD models with the most empirical support (i.e., 
the numbing, dysphoria, and dysphoric arousal models). Finally, as noted earlier, the results of 
Monteith and colleagues reflect the likely occurrence of multicollinearity, possibly due in part to 
the inclusion of depressive symptoms in their analyses. The present study sought to avoid 
multicollinearity by identifying and eliminating content overlap between symptoms of 
depression and PTSD. As a result, only sadness (Beck Depression Inventory- Second Edition 
Item 1) was controlled for in structural analyses (see Method section for a more detailed 
explanation and justification). This decision was made to minimize multicollinearity while still 
 
 
9 
 
accounting for an aspect of depression that was likely to be both relevant to P-SI and 
conceptually distinct from other constructs included in the model.  
The present study tested three hypotheses designed to address noted gaps in existing 
literature on the relationship between PTSD and SI. Consistent with existing factor analytic 
literature, the first hypothesis was that the dysphoric arousal model would show the best fit for 
PTSD symptom structure compared to the numbing and dysphoria models. The second 
hypothesis was that relative to other PTSD symptom clusters the numbing cluster would have the 
strongest relationship with PB, TB, and P-SI. The third hypothesis was that PB, TB, or both 
would fully mediate the relationship between numbing symptoms and P-SI when other relevant 
correlates (i.e., gender, sadness, and all other PTSD clusters) were controlled for. To test the 
second and third hypotheses, measurement and structural models were specified to include 
individual PTSD symptom clusters, gender, sadness, PB, TB, and P-SI. This was seen as 
providing a rigorous test of hypotheses concerning the relationships between numbing, PB, and 
TB in a trauma-exposed population. 
  
 
 
10 
 
 
 
Method 
Participants and Procedure 
Participants were undergraduates enrolled in psychology courses at a large public 
university in the southeastern United States. With the approval of the university?s internal review 
board, 512 participants who self-identified as experiencing a ?very stressful life event? were 
recruited between Spring 2011 and Spring 2012. Study participation involved completion of a 
battery of self-report questionnaires during proctored research sessions conducted in a university 
computer lab. Each session accommodated up to 20 participants with one clinical psychology 
graduate student and one undergraduate research assistant present to supervise. Participants were 
included in analyses if their index event met DSM-IV Criterion A1 as classified by a trained team 
of graduate students; this process is described in further detail below. The final sample included 
334 individuals ranging in age from 17 to 54 (M = 19.65 years, SD = 3.30). The sample was 
predominantly female (72.20 %, n = 241), and the ethnic breakdown was 79.30 % Caucasian, 
12.60 % African-American, 3.60 % Latino/Hispanic, 3.00 % Asian, and 1.50 % Other.  
Measures  
Trauma exposure and PTSD. The Life Events Checklist (LEC) was used to assess 
trauma type and exposure level. The LEC is the self-report trauma assessment portion of the 
Clinician-Administered PTSD Scale (CAPS; Blake et al., 1995; Weathers, Keane, & Davidson, 
2001) and assesses exposure to 17 types of traumatic events (e.g., natural disaster, sexual 
assault). The LEC has been demonstrated to have good psychometric properties (Gray, Litz, Hsu, 
& Lombardo, 2004). Based on responses to the LEC the three most commonly reported 
 
 
11 
 
traumatic event types were car accidents (23.70 %, n = 79), sudden and unexpected death of a 
loved one (18.00 %, n = 60), and life threatening illness or injury (12.9%, n = 43).  
After completing the LEC, participants wrote a brief narrative description of the worst 
traumatic event they had experienced, and this served as the index event for PTSD symptom 
assessment. Narrative event descriptions were coded by a team of two graduate students, 
supervised by Dr. Weathers, to determine whether participants met DSM-IV Criterion A1 for 
PTSD. Only participants whose index event met Criterion A1, which specifies exposure to an 
event that ?involved actual or threatened death or serious injury or a threat to the physical 
integrity of self or others,? were included in analyses (APA, 1994, p. 427). Interrater reliability 
for Criterion A1 coding was in the acceptable range (? = .77; Landis and Koch, 1977). 
Discrepancies were resolved by discussion among the two raters and Dr. Weathers. 
The PTSD Checklist ? Specific Version (PCL-S; Weathers, Litz, Herman, Huska, & 
Keane, 1993) was used to measure PTSD symptom severity. The PCL-S is a widely used, well-
validated self-report measure that assesses each of the 17 DSM-IV symptoms of PTSD. 
Respondents first identify an index stressful life event, then indicated how much they have been 
bothered by each PTSD symptom in the past month using a 5-point scale (1 = not at all to 5 = 
extremely). The PCL has excellent psychometric properties (McDonald & Calhoun, 2010; 
Wilkins, Lang, & Norman, 2011). In the current sample, internal consistency for the PCL-S was 
high (? = .93), mean total score was 29.86 (SD = 12.45), and 13.80 % of the sample (n = 46) met 
criteria for likely PTSD diagnosis based on a cutoff score of 44 established for diagnosing PTSD 
among treatment-seeking trauma survivors (Ruggiero, Del Ben, Scotti, & Rabalais, 2003).  
Suicidal ideation. The Beck Suicide Scale (BSS; Beck & Steer, 1991) is a 21-item self-
report measure designed to assess the presence and severity of SI. The first 19 items assess 
 
 
12 
 
suicidal ideation, planning, and preparation and are presented with variable response options 
which range in severity on a scale from 0 to 2. Items 20 and 21 assess history of suicide attempts 
and were therefore not included in study analyses. The BSS has been widely used and has strong 
psychometric properties (Beck, Brown, & Steer, 1997). The decision to include only select BSS 
items in structural analyses is discussed below in the Results section 
IPTS constructs, The Interpersonal Needs Questionnaire (INQ) is a 15-item self-report 
measure designed to assess PB (6 items) and TB (9 items). Items are presented on a 7-point 
Likert scale with response options ranging from 0 (not at all true for me), to 6 (very true for me). 
The INQ has good psychometric properties (Van Orden et al., 2012). In the current sample, 
internal consistency for the INQ was excellent for both subscales (? = .90 for PB; ? = .89 for 
TB), mean score for PB was 8.30 (SD = 4.48), and mean score for TB 19.91 (SD = 11.07). 
Depressive symptomatology. Symptoms of depression were assessed using the Beck 
Depression Inventory-II (BDI-II), a widely used measure which has good psychometric 
properties (Beck, Steer, & Brown, 1996; Osmond, Downs, Barrios, Kopper, Gutierrez, & Chiros, 
2008). On the BDI-II, respondents are presented with 21 groups of statements describing 
depressive symptoms and are asked to select the one statement from each group that most closely 
matches how they have been feeling in the past two weeks. The statements range in severity on a 
scale from 0 to 3. In the current sample, internal consistency for BDI-II was excellent (? = .93), 
and mean total score on the BDI-II was 10.56 (SD = 9.40).  
 Of concern, Monteith and colleagues (2013) reported a strong positive zero-order 
correlation between the avoidance/numbing PTSD symptom cluster and depression as measured 
by the BDI-II (r = .53). This information suggests that the inclusion of depressive symptoms may 
have contributed to the occurrence of multicollinearity in their study. To further investigate this 
 
 
13 
 
possibility, BDI-II item content was compared with that of other measures in the present study to 
identify areas of potential conceptual overlap. Multiple instances of content overlap were 
identified between the BDI-II and items on the PTSD Checklist (PCL). For example, BDI-II 
items 4 and 12 (loss of interest and loss of pleasure respectively) closely resemble PCL item 9 
which refers to ?loss of interest in activities that you used to enjoy.? Similarly, BDI-II items 17 
and 19 (reflecting irritability and difficulty concentrating) are virtually identical to PCL items 14 
feeling irritable or having angry outbursts and 15 having difficulty concentrating. Each pair of 
BDI-II and PCL items mentioned was correlated between .57 and .60. Tables C1 and C2 contain 
content descriptors for all study items. Thus, as noted, to avoid potential conceptual overlap and 
related concerns regarding multicollinearity, only BDI-II item 1 (sadness) was controlled for in 
structural analyses. 
  
 
 
14 
 
 
 
Results 
Descriptive Statistics 
All analyses were performed in Mplus version 7.0 (Muth?n & Muth?n, 1998-2012) and 
IBM SPSS version 20.0. Descriptive statistics including means, standard deviations, and skew 
can be found in Table A1. Item-level zero-order correlations between study variables are 
displayed in Tables A2 - A4. In keeping with recommendations from Bentler and Chou (1987) 
the distributions of responses to study variables were examined at the item level to determine if 
items should be treated as continuous or ordered categorical. Most of the items that constitute the 
PCL, INQ, BSS, and item 1 of the BDI-II were found to be positively skewed (skewness values 
between 0.84 and 8.60), and visual inspection of the item distributions revealed that none closely 
approximated a normal distribution (Table A1 contains skewness and kurtosis values for all 
items used in analyses). Therefore, all variables were treated as ordered-categorical in structural 
analyses. Given this decision, a mean- and variance-adjusted weighted least squares (WLSMV) 
estimator was implemented for all analyses, which has been described as the preferred estimation 
method for analyses involving ordered-categorical variables (Muth?n, du Toit, & Spisic, 1997; 
Flora & Curran, 2004). In Mplus, the WLSMV estimator produces probit coefficients for both 
factor loadings and structural coefficients, which cannot be interpreted as linear regression 
coefficients would be. However, correlations between latent variables in such models are 
equivalent to standard linear regression coefficients (Muth?n & Muth?n, 1998-2012).  
Missing data were handled using pairwise present deletion, the default strategy used in 
Mplus when the WLSMV estimator is employed (Asparouhov & Muth?n, 2010; Muth?n & 
 
 
15 
 
Muth?n, 1998-2012). To evaluate the appropriateness of this strategy, the pattern and degree of 
missing data were examined. Little?s Test was significant (?2  [855] = 1249.18, p < .01) 
suggesting that data were not missing completely at random (MCAR), and therefore a robust 
strategy for handling missing data would be more appropriate. However, use of full information 
maximum likelihood is not possible when a WLSMV estimator is used, and multiple imputation 
(MI) was attempted but could not be implemented due to difficulties with convergence (Kline, 
2011). Given that only a small proportion of the data was missing (covariance coverage values 
were all between .96 and .99 for all pairs of items included in analyses) use of pairwise deletion 
was deemed acceptable, despite recognition that this strategy may lead to biased results when 
data are not MCAR (Brown, 2006).  
Across analyses, evaluation of model fit was accomplished by applying cutoff guidelines 
recommended by Kline (2011) to various model fit indices. First, non-significant ?2 values were 
considered an indication of good model fit (Hayduk, 1996). Second, Root Mean Square Error of 
Approximation (RMSEA) estimates were evaluated according to two hypotheses. Estimates 
lower than or equal to .05 (or the presence of that value within the RMSEA 90 % confidence 
interval) resulted in a failure to reject the hypothesis that the model was a ?close fit,? while 
estimates greater than or equal to .10 (or the presence of that value within the RMSEA 90 % 
confidence interval) resulted in failure to reject the hypothesis that the model was a ?poor fit? 
(Browne & Cudeck, 1993). Finally, values greater than or equal to .95 on Bentler?s Comparative 
Fit Index (CFI) and Tucker-Lewis Index were considered indicators of acceptable model fit (Hu 
& Bentler, 1999). 
 
 
16 
 
Preliminary Analyses 
As a preliminary step, a CFA was performed to confirm the latent structure of the BSS 
and to inform decisions on which items to include in structural analyses. Review of existing 
literature on BSS factor structure supported a two-factor solution, dividing items reflecting 
suicidal ideation, desire, and intent (SI), and resolved plans and preparation (RPP; Steer, 1993: 
Holden & DeLisle, 2005; Joiner, Rudd, & Rajab, 1997; Witte et al., 2006). Both because study 
hypotheses focused on P-SI only, and given the extreme restriction of range in responses to RPP 
items, all RPP items (BSS item 7-BSS item 19) were removed and, a one-factor model including 
BSS items 1-6 was specified and tested using a WLSMV estimator. BSS items 4 (suicidal desire) 
and 6 (length/ duration of SI) had a tetrachoric correlation of 1.00 which prevented the model 
from running when both were included. That is, although not all respondents answered the two 
questions identically (their bivariate correlation was r = .77), 100% of individuals who 
responded with a ?1? or a ?2? on BSS item 6 responded with a ?1? on BSS item 4. Based on 
recommendations by Mplus software developers (L. Muth?n, personal communication, 2008), 
the items were combined into one variable such that endorsing a 0 on BSS item 4 and a 0 on BSS 
item 6 resulted in a 0 on the combined variable, endorsing a 0 on BSS item 4 and a 1 on BSS 
item 6 resulted in a 1 on the combined variable, endorsing a 1 on both variables resulted in a 2 on 
the combined variable, and endorsing a 1 on the BSS item 4 and a 2 on the BSS item 6 resulted 
in a 3 on the combined variable  The modified model, specified to load BSS items 1, 2, 3, 5, and 
the combined BSS 4 and 6 item on one latent factor (SI) showed good fit (?2 [5]= 6.58, p = .25, 
TLI = .99, CFI = .99, RMSEA [90 % CI] = .03 [.00 - .08]) and was therefore incorporated into 
structural analyses. All structural results made use of the resultant P-SI variable. Based on 
 
 
17 
 
summed responses to the remaining 5 items comprising P-SI, 14.97 % (n = 50) of participants 
reported experiencing some level of P-SI as indicated by a summed score of 1 or greater.  
Hypothesis 1: Evaluation of PTSD Factor Structure 
To test Hypothesis 1, three CFAs were conducted to evaluate the fit of competing models 
of the latent factor structure of PTSD symptoms in our sample. CFAs were specified to evaluate 
the three models of PTSD symptom structure with the strongest empirical support in the 
literature: the four-factor numbing and dysphoria model, and five-factor dysphoric arousal model 
(Elhai & Palmieri, 2011). All three models were determined to have acceptable fit based on 
evaluation of standard fit indices according to the referenced cutoff guidelines. Fit information 
for all three models is presented in Table A5. Because the numbing and dysphoria models are not 
nested, a formal comparison of model fit using the ?2 difference test was not possible. However, 
?2 difference tests were conducted to determine whether the dysphoric arousal model provided a 
superior fit to each of the four-factor models. A two-step alternative procedure for ?2 difference 
testing was employed using the ?DIFFTEST? command in Mplus to account for the fact that the 
WLSMV estimator was used (Brown, 2006; Muth?n & Muth?n, 1998-2012). Consistent with 
Hypothesis 1, results indicated that the five-factor model provided significantly better fit than 
either the numbing (?2 [4] = 29.85, p < .01) or dysphoria (?2  [4] = 72.86, p < .01) four-factor 
models. As such, the five-factor model, which divides PTSD symptoms into clusters reflecting 
reexperiencing, avoidance, numbing, dysphoric arousal, and anxious arousal respectively, was 
used in structural analyses.  
Hypotheses 2 and 3: Evaluation of the Full Model 
 Measurement and structural models were specified to assess the hypothesized 
relationships between all individual PTSD symptom clusters, PB, TB, and P-SI (Hypotheses 2 
 
 
18 
 
and 3). The measurement model was specified first with covariances estimated between all latent 
variables and the two observed variables-gender and sadness. The covariance between the 
residual terms of INQ items 11 and 12, both of which were specified to load on the TB latent 
factor, was also estimated based on guidance from the literature (Van Orden et al, 2012). 
Examination of model-fit statistics revealed evidence of acceptable overall model fit (?2 [696] = 
1009.48, p < .01, TLI = .98, CFI = .98, RMSEA [90 % CI] = .04 [.03 - .04]), and all standardized 
factor loadings were above the recommended guideline of .40 (Brown, 2006). It is important to 
note that the chi-square test of model fit in this and other models presented here was significant 
and therefore not suggestive of adequate model fit. However, this index has been shown to be 
sensitive to large sample sizes, and as such may not provide an accurate estimate of model fit in 
this case (Brown, 2006). Recently some have called the practice of discounting significant chi-
square values into question. Kline (2011), for example, argued that the presence of a significant 
chi-square value should be taken as an indication of a problem with the specification of the 
model, and recommends further diagnostics be performed to determine how the model should be 
respecified. Examination of the pattern of residuals to identify potential sources of strain 
affecting model fit did not point to any theoretically justifiable modifications to model 
specification for any of the models presented. Modification indices were unavailable given use of 
the WLSMV estimator. 
 Standardized and unstandardized parameter estimates from the full measurement model 
are available in Table A6. These values were examined to evaluate Hypothesis 2 concerning the 
differential relationships between numbing and other PTSD clusters with PB, TB, and P-SI. 
Consistent with Hypothesis 2, numbing had stronger bivariate relationships (i.e., standardized 
covariance estimates) with PB, TB, and P-SI than any other PTSD symptom cluster in the 
 
 
19 
 
measurement model. Further, examination of the 95% confidence intervals surrounding each 
estimate confirmed that the bivariate relationship between numbing and PB was significantly 
larger than that of any other cluster except dysphoric arousal. Similarly, the relationship between 
numbing and TB was significantly larger than that of any other cluster.  Finally, the relationship 
between numbing and P-SI was significantly larger than the relationships between other PTSD 
clusters and P-SI with two exceptions: anxious arousal and dysphoric arousal. Interestingly, the 
bivariate relationship between numbing and P-SI was also larger than the relationship between 
TB and P-SI, although not significantly so. Taken together these findings support the hypothesis 
that numbing would have a stronger relationship with PB, TB, and P-SI than any other cluster. 
Notably, gender had significant positive bivariate relationships with and both avoidance and 
sadness, suggesting higher values of both constructs among females.  
Next, to evaluate Hypothesis 3, a corresponding structural model was specified with 
covariances estimated between the latent variables for all PTSD symptom cluster, sadness, and 
gender, and between PB and TB. As in the measurement model, INQ items 11 and 12 were also 
allowed to covary. Direct paths were specified from all PTSD symptom clusters, sadness, and 
gender to PB, TB, and SI, and from PB and TB to SI. See Figure B1 for depiction of the full 
structural model and specified direct paths. Examination of fit indices for the structural model 
also suggested acceptable model fit (?2 [697] = 1273.73, p < .01, TLI = .96, CFI = .99, RMSEA 
[90 % CI] = .05 [.04 ? .05], with all standardized factor loadings above .40. However, two 
indications of multicollinearity were also observed. First, the standardized coefficients 
corresponding to the paths from numbing to P-SI and numbing to TB were both larger than 1.0 
(? = 1.24 in both cases), which may indicate the presence of severe multicollinearity (e.g., 
J?reskog, 1999; Kline, 2011). Second, a number of structural coefficients (specifically the paths 
 
 
20 
 
from TB, avoidance, and dysphoric arousal to SI, and from various PTSD symptom clusters to 
both PB and TB) were negative in the structural model, despite moderate, positive associations 
between the same pairs of variables in the corresponding measurement model. For example, 
dysphoric arousal and TB had a negative association in the structural model (? = -.53, p = .01), 
despite their positive association in the corresponding measurement model (r = .49, p < .01). All 
standardized and unstandardized parameter estimates from the full measurement (Table A6) and 
structural (Table A7) models are presented to illustrate the referenced discrepancies. As noted 
earlier, this flipping of signs between the measurement and structural models is a key marker of 
multicollinearity when predictors are highly intercorrelated (Cheung & Lau, 2008). Of particular 
concern, dysphoric arousal and numbing were correlated .89, and reexperiencing and avoidance 
were correlated .91 in the measurement model (inter-item correlations between all PCL items are 
also available for review in Table C2). Because the presence of multicollinearity can bias 
parameter estimates, the results of this model were not interpreted. 
Based on these findings and recommendations by Kline (2011) and others concerning the 
appropriate handling of multicollinearity in structural equation modeling, a decision was made to 
re-specify the model (Grewal, Cote, & Baumgartner, 2004). To determine what changes were 
appropriate, the pattern of correlations in the measurement model was examined. It was 
determined that the observed presence of multicollinearity was at least partially accounted for by 
the fact that many of the PTSD symptoms clusters were strongly positively associated (as noted 
above). To avoid problems associated with multicollinearity, five separate models were 
specified, each containing only one PTSD symptom cluster along with PB, TB, gender, sadness 
and SI. 
 
 
21 
 
Models with Individual Symptom Clusters  
Figure B2 depicts the structural model specified for each individual symptom cluster 
including all specified direct paths. Both measurement and structural models for each of the five 
individual symptom clusters showed acceptable fit (See Table A8 for fit statistics associated with 
each model). None of the individual model results were suggestive of multicollinearity. Table A9 
lists both standardized and unstandardized parameter estimates for each model. Both control 
variables (gender and sadness) performed consistently across models. Specifically, in all five 
structural models sadness had positive direct paths to TB, PB, and SI. None of the direct paths 
from gender to TB, PB, or P-SI were significant, and only one variable (avoidance) was 
significantly correlated with gender (r = .14, p = .03, suggesting that females reported slightly 
higher levels of avoidance). 
Given that study hypotheses involve multiple mediators, a bias-corrected (BC) 
bootstrapping procedure recommended for use with multiple mediators by Preacher and Hayes 
(2008) was used to estimate indirect effects in each model. BC bootstrapping involves 
resampling repeatedly from the original data set with replacement and calculating an estimate of 
the population indirect effects in each sample. Five-thousand resamplings were performed for 
each model based on recommendations from the literature (Preacher & Hayes, 2008). Specific 
indirect effects and the associated 95 % confidence intervals for each symptom cluster through 
both PB and TB are presented in Table A10.  
Given that the full structural model containing all PTSD symptom clusters could not be 
interpreted, Hypothesis 3 could not be evaluated as intended (with all other PTSD symptoms 
controlled for). However, contrary to Hypothesis 3, results of the model containing numbing 
alone did not support full mediation of the relationship between numbing and P-SI through either 
 
 
22 
 
PB or TB. Further, also contrary to Hypothesis 3, results for the models containing numbing and 
reexperiencing were quite similar. Both numbing and reexperiencing had direct paths to both PB 
and TB. PB, in turn, had a significant direct path to P-SI in both models, whereas TB had a 
significant direct path to P-SI in the reexperiencing model but not the numbing model. Finally, 
contrary to Hypothesis 3, numbing and reexperiencing both had significant direct paths to P-SI, 
as well as significant indirect effects on P-SI through PB, but not TB. Thus, although full 
mediation was not present, PB partially accounted for the relationship between numbing, 
reexperiencing, and P-SI in their respective models.  
Although Hypothesis 3 was not supported, results from the structural model containing 
numbing did provide additional support for Hypothesis 2.  Specifically, the magnitudes of the 
standardized direct path coefficients from numbing to PB, TB, and P-SI were larger than the 
standardized coefficients for corresponding paths in any other individual cluster model (see 
Table A9). For example, the value for the direct path coefficient from numbing to TB was ? = 
.62, compared to ? = .38 for the path from dysphoric arousal to TB, which was the next largest 
(full results from the dysphoric arousal model are discussed below). Additionally, a slightly 
greater proportion of the variance in P-SI was accounted for in the numbing model (R2 = .64) 
compared to any other individual cluster models (R2 values of .60-.62, see Table A8). This 
suggests the relationship between numbing and both P-SI and the IPTS constructs was stronger 
than any other cluster, including reexperiencing.  
 Results from the models containing anxious arousal, dysphoric arousal, and avoidance 
were also similar to one another. Each symptom cluster had significant direct paths to both PB 
and TB, and in each model, both PB and TB had significant direct paths to SI. Unlike numbing 
and reexperiencing, anxious arousal, dysphoric arousal, and avoidance did not have significant 
 
 
23 
 
direct paths to SI. However, each was indirectly related to P-SI through either both PB and TB 
(dysphoric arousal and avoidance), or PB alone (anxious arousal). These findings were again 
partially consistent with study hypotheses. Specifically, the relationships between P-SI and all 
three symptom clusters were partially mediated by one or both IPTS constructs. As with 
reexperiencing, this finding is inconsistent with Hypothesis 3, which specified that of the PTSD 
clusters only numbing would have a significant indirect relationship with SI. 
  
 
 
24 
 
 
 
Discussion   
This study examined differential relationships between specific PTSD symptom clusters 
and P-SI, using the IPTS as a theoretical framework. This is the first study to simultaneously 
address two key limitations of previous research. First, in contrast with previous studies that 
either examined PTSD as a syndrome, used the rationally derived three-factor DSM-IV model of 
PTSD symptoms, or failed to consider all PTSD symptom clusters, this study examined all 
clusters based on the empirically derived dysphoric arousal model. Second, in contrast with 
previous studies that were primarily atheoretical, this study involved systematic application of 
the IPTS, a comprehensive and specific model of suicidal behavior with considerable potential to 
account for the association between PTSD and P-SI. In addition, this study is the first to use  
SEM to simultaneously assess relationships between PTSD clusters, IPTS constructs, and P-SI, 
and to explicitly assess for potential indirect effects of PTSD symptoms on SI through PB and 
TB.  
 Based on the IPTS assumption that PB and TB are sufficient and possibly necessary 
proximal causes of P-SI, the primary study hypothesis (Hypothesis 3) was that PB and TB would 
fully mediate the relationship between numbing and P-SI. As noted, due to multicollinearity this 
hypothesis could not be evaluated as intended. Specifically, instead of a single structural model 
containing all PTSD symptom clusters, it was necessary to run separate structural models for 
each symptom cluster. Contrary to expectations, full mediation was not observed in the structural 
model involving numbing. Further, an implied hypothesis was that numbing would have a 
significant (indirect) association with P-SI while other PTSD clusters would not. Instead, both 
 
 
25 
 
numbing and reexperiencing had significant direct paths to P-SI in their respective structural 
models. Additionally, the relationship between each PTSD symptom cluster and P-SI was 
partially mediated by one or both of the IPTS constructs. Thus, in this sample, every PTSD 
cluster was indirectly related to P-SI. 
 The IPTS provides some explanation for the fact that this hypothesis was not supported. 
Specifically, the IPTS acknowledges that the nature of the relationship between PB, TB and P-SI 
is somewhat unclear. The theory conservatively describes PB and TB as sufficient proximal 
causes for the occurrence of P-SI. A more extreme competing hypothesis is that the presence of 
PB, TB, or both is not only sufficient but necessary for the development of P-SI, i.e.,  that PB 
and TB lead to P-SI across situations (Van Orden et al., 2010). If proven, this hypothesis would 
greatly enhance the clinical usefulness of the theory. This study attempted to assess the validity 
of this competing hypothesis, and failed to support it. Instead, both the noted presence of partial 
mediation through PB and TB, and direct relationships between PB, TB, and P-SI in all but the 
numbing model support the original IPTS assertion that both constructs are proximally related to 
P-SI. One possible explanation is that the INQ did not adequately measure PB and TB. However, 
in their review paper on the IPTS, Van Orden and colleagues (2010) noted that the majority of 
existing theory (e.g., Baechler?s taxonomy of suicide) suggests that causes of various forms of 
suicidal behavior (including P-SI) are situationally dependent. This implies that no construct or 
set of constructs should be capable of universally predicting SI. Thus, this study?s findings are 
actually consistent with the majority of existing theory and not explicitly inconsistent with the 
IPTS itself.  
At least two explanations may account for the inaccuracy of predictions concerning 
numbing and other PTSD clusters. First, results may have been affected by the inability to 
 
 
26 
 
consider all clusters simultaneously in a single model. That is, the relationships between other 
PTSD symptom clusters and P-SI might not have been significant in a model that included 
numbing as a covariate. Second, given previous research, these findings should arguably not 
have been unexpected. Study hypotheses assumed that PB and TB were the sole proximal causes 
of SI. As such, symptom clusters that did not fit theoretically as indicators of either construct 
were not focused on, even though in some cases (i.e. reexperiencing and hyperarousal), mixed 
research demonstrated a relationship between those clusters and SI (Bell & Nye, 2007; Tarrier & 
Gregg, 2004; Ben Yaakov & Amir, 2004). Reexperiencing in particular was both directly and 
indirectly related to P-SI. Similar future studies should broaden focus and consider clusters that 
are less theoretically consistent with IPTS constructs and potentially search for explanations for 
the relationship between those clusters and P-SI outside the IPTS. 
Notably, other findings were more consistent with study hypotheses. Specifically, as 
hypothesized the dysphoric arousal model had significantly better fit than both numbing and 
dysphoria models, adding to a growing body of research supporting this model. Additionally, 
various findings supported the hypothesis that numbing would have stronger relationships with 
PB, TB, and SI than other PTSD clusters. In the full measurement model, numbing had stronger 
bivariate relationship with TB, PB, and P-SI than any other cluster, and in all but three cases 
(i.e., dysphoric and anxious arousal with P-SI; dysphoric arousal with PB), those differences 
were statistically significant. Further, the magnitudes of the relationships between numbing and 
TB, PB, and P-SI were larger than corresponding values in any other individual cluster structural 
model. The magnitude of the relationship between numbing and TB in the individual structural 
model in particular seemed to confirm the hypothesized theoretical consistency between the two. 
Finally, the lack of a significant direct relationship between TB and P-SI in the numbing model 
 
 
27 
 
only may in part have been due to conceptual overlap between numbing and TB. Taken together 
these findings support the idea that compared to other PTSD clusters, numbing has a unique and 
important relationship with SI. 
  This study had several methodological limitations. First, and most crucially, apparent 
multicollinearity prevented the consideration of all PTSD symptom clusters together as part of a 
single model. In addition to preventing evaluation of study hypotheses as intended, the need to 
model clusters separately limits the validity and generalizability of these findings. PTSD clusters 
are distinct but highly intercorrelated, and would be unlikely to present in isolation (e.g., the 
modal response to trauma would not be to experience only symptoms of reexperiencing, or only 
symptoms of hyperarousal; Wilson, 2004). Without examining a model containing all symptom 
clusters, it is impossible to speak to the unique relationship between specific clusters and SI as 
the disorder typically presents. For example, although in the individual model reexperiencing 
was both directly and indirectly related to SI, inclusion of numbing and other clusters might have 
rendered those relationships non-significant. Notably, it is possible that the occurrence of 
multicollinearity was affected by a second limitation, which was the non-clinical nature of the 
sample. Specifically, limited variability in symptom endorsement may have contributed to the 
unusually high inter-item and inter-cluster correlations in the sample (i.e. r = .91 between 
reexperiencing and avoidance), which led to muticollinearity.  
 A third limitation concerned the choice to control for sadness rather than all depressive 
symptoms in study analyses. According to the IPTS, PB and TB should be directly related to SI 
even when controlling for MDD (Van Orden et al., 2010). The inclusion of all depressive 
symptoms might arguably have constituted a more rigorous test of the theory. Further, although 
consensus holds that co-occurrence of PTSD and MDD is associated with increased SI compared 
 
 
28 
 
to PTSD alone, researchers disagree on why (i.e. whether comorbid MDD mediates or moderates 
the relationship between PTSD and SI; Panagioti et al., 2012). Excluding all depressive 
symptoms prevented this study from contributing to the ongoing debate. That said, some 
researchers have suggested that the high rate of comorbidity between PTSD and MDD is 
attributable to symptom overlap, which can contribute to multicollinearity (Brady, Kileen, 
Brewerton, & Lucerini, 2000). Research supporting this hypothesis is currently mixed (Tabet, 
Abed, & Vostanis, 2004; Taft, Resick, Watkins, & Panuzio, 2009). However, given the aims of 
this study the majority of depressive symptoms were excluded to avoid multicollinearity 
apparent in similar studies (e.g., Monteith et al., 2013).  
  A fourth limitation was that this study was cross-sectional and correlational in nature, 
which prevents inferences concerning causal relationships between study variables (Pearl, 2000), 
and data were collected exclusively via self-report. Ideally, future research would be longitudinal 
to provide information about temporal relationships between variables and would rely more 
heavily on clinician-administered measures. Fifth, our sample was, predominantly Caucasian, 
and predominantly female. Although gender was controlled for in all analyses, findings should 
be replicated in samples with greater ethnic diversity to determine generalizability. Sixth, for the 
purposes of this study all trauma types were combined, masking potential differences in the 
presentation of SI or IPTS constructs across type of trauma. Finally, the use of pairwise deletion 
as a strategy for handling missing data is a potential limitation. Specifically, this strategy can 
lead to biased parameter estimates, particularly when data are missing at random rather than 
MCAR, as was the case in our study. Ideally, a more robust strategy, such as multiple imputation 
in the case of a WLSMV estimator, would be used (Brown, 2006).  
 
 
29 
 
 Despite these limitations, this study represents a significant contribution to literature on 
the relationship between PTSD and SI both methodologically and practically. Focusing on 
methodology, this was the first study to study the relationships between PTSD clusters and SI 
using SEM. Although this approach has clear advantages, it would ideally be used to consider 
PTSD symptom clusters as they occur phenomenologically, together. Future research should 
focus on identifying and eliminating specific causes of multicollinearity that prevent the 
simultaneous consideration of PTSD clusters. The choice to focus only on P-SI was another 
important advance. Both research and theory suggest that various forms of suicidal behavior (e.g. 
P-SI, desire, intent, attempts) occur at different rates, and for different reasons (Van Orden et al., 
2010). Thus, deliberate focus on identifying predictors for specific forms of suicidal behavior 
should provide more useful and specific information.  
The relationship between PB, TB, and P-SI in this sample was clear, which implies the 
utility of the IPTS in predicting P-SI among those with PTSD. Clinicians should consider 
assessing for the presence of PB and TB in this population to identify patients at risk for 
developing SI. Additionally, future research should investigate the applicability of the IPTS in 
this population with other forms of suicidal behavior (i.e., suicidal intent and attempts). 
However, contrary to hypotheses, PB and TB were not the only variables proximally related to 
P-SI. This suggests that although the IPTS is helpful, it does not provide a comprehensive 
explanation for the occurrence of P-SI in this population. Findings also implicated PTSD?s 
numbing and reexperiencing clusters and sadness as variables with direct relationships to SI. 
Assuming that these findings are replicated, future research should focus on providing 
empirically and theoretically justifiable explanations for the relationship between these variables 
and SI. The finding concerning numbing also highlights the importance of separating avoidance 
 
 
30 
 
and numbing symptoms. Notably, with the recent publication of the DSM-5 the core symptoms 
of PTSD have changed (APA, 2013). While these changes must be taken into account, they 
should not be considered to invalidate existing findings. DSM-5 clusters corresponding to 
reexperiencing and numbing should be prioritized as the focus of future research, and likely 
intervention. More broadly, this study provides additional confirmation for the idea that 
differential relationships exist between specific PTSD symptom clusters and SI. Future research 
should continue to consider PTSD and its correlates at symptom cluster level after suitable 
investigation of the latent structure of the DSM-5 symptoms.  
 
  
 
 
31 
 
 
 
References 
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders 
(4th ed.). Washington, DC: Author. 
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders 
(5th ed.). Washington, DC: Author. 
Anestis, M. D., Bryan, C. J., Cornette, M. M., & Joiner, T. E. (2009). Suicidal behavior in the 
military: An evaluation of Joiner?s interpersonal-psychological theory of suicide in two 
case studies of active duty post-deployers. Journal of Mental Health Counseling, 31, 60-
75. 
Asmundson, G. J. G, Stapelton, J. A., & Taylor, S. (2004). Are avoidance and numbing distinct 
PTSD symptom clusters? Journal of Traumatic Stress, 17, 467?475.  
Asparouhov T. & Muth?n, B. (2010). Weighted least squares estimation with missing data. 
Unpublished technical report. 
Baumeister, R. F. & Leary, M. R. (1995). The need to belong: Desire for interpersonal 
attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529. 
Beck, A. T., & Steer, R. A. (1991). Manual for Beck Scale for Suicide Ideation. New York: 
Psychological Corporation. 
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Beck Depression Inventory-II Manual (2nd 
ed.). San Antonio, TX: Psychological Corporation. 
 
 
32 
 
Beck, A. T., Brown, G. K., & Steer, R. A. (1997). Psychometric characteristics of the Scale for 
Suicide Ideation with psychiatric outpatients. Behaviour Research and Therapy, 35, 
1039-1046. 
Bell, J. B., & Nye, E. C. (2007). Specific symptoms predict suicidal ideation in Vietnam combat 
veterans with chronic post-traumatic stress disorder. Military Medicine, 172, 1144?1147. 
Bernal, M., Haro, J.M., Bernert, S., Brugha, T., de Graaf, R., Bruffaerts, R., et al.(2007). Risk 
factors for suicidality in Europe: Results from the ESEMED study. Journal of Affective 
Disorders, 101, 27?34.  
Ben-Ya?acov, Y., & Amir, M. (2004). Posttraumatic symptoms and suicide risk. Personality and 
Individual Differences, 36, 1257?1264.  
Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Charney, D. S., & Keane, T. M. 
(2004). Clinician-Administered PTSD Scale. Los Angeles: Western Psychological 
Services. 
Brady, K. T., Killeen, T. K., Brewerton, T., & Lucerini, S. (2000). Comorbidity of psychiatric 
disorders and posttraumatic stress disorder. The Journal of Clinical Psychiatry, 61, 22?
32. 
Brenner, L. A., Gutierrez, P. M., Cornette, M. M., Betthauser, L. M., Bahraini, N., & Staves, P. 
J. (2008). A qualitative study of potential suicide risk factors in returning combat 
veterans. Journal of Mental Health Counseling, 30, 211-225. 
Breslau, N., Davis, G. C., Peterson, E. L., & Schultz, L. (2000). A second look at comorbidity in 
victims of trauma: The posttraumatic stress disorder-major depression connection. 
Society of Biological Psychology, 48, 902? 909. 
 
 
33 
 
Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY: 
Guilford Press. 
Browne, M.W., & Cudeck, R. (1993) Alternative ways of assessing model fit. In K.A.Bollen & 
J.S. Long (Eds.), Testing structural equation models (pp. 136-162). Newbury Park, CA: 
Sage. 
Bryan, C., Cukrowicz, K.C., West, C.L., & Morrow, C.E. (2010). Combat experience and the 
acquired capability of suicide. Journal of Clinical Psychology, 66, 1044-1056. 
Bryan, C., & Anestis, M. (2011). Reexperiencing symptoms and the interpersonal-psychological 
theory of suicidal behavior among deployed service members evaluated for traumatic 
brain injury. Journal of Clinical Psychology, 67, 1-10. 
Bryan, C. J., Hernandez, A. M., Allison, S., & Clemans, T. (2012). Combat exposure and suicide 
risk in two samples of military personnel. Journal of Clinical Psychology. Advance 
online publication. Doi: 10.1002/jclp.21932 
Carter, B. (2011). Measurement Invariance of Posttraumatic Stress Disorder Symptoms 
(Unpublished doctoral dissertation). Auburn University, Auburn, AL.  
Cheung, G.W. & Lau, R.S. (2008). Testing mediation and suppression effects of latent variables: 
Bootstrapping with structural equation models. Organizational Research Methods, 11, 
296-325. 
Davis, M.T., Witte, T.K., & Weathers, F.W. (under review) Posttraumatic Stress Disorder and 
Suicidal Ideation:  The Role of Specific Symptoms within the Framework of the 
Interpersonal-Psychological Theory of Suicide. Psychological Trauma: Research, 
Practice, and Policy. 
 
 
34 
 
Elhai, J., Biehn, T.L., Armour, C., Klopper, J., Frueh, C., & Palmieri, P. (2011). Evidence for a 
unique PTSD construct represented by PTSD?s D1-D3 symptoms. Journal of Anxiety 
Disorders, 25, 340-345. 
Elhai, J., & Palmieri, P. (2011). The factor structure of posttraumatic stress disorder: A literature 
update, critique of methodology, and agenda for future research. Journal of Anxiety 
Disorders, 25, 840-854. 
Flora, D.B., & Curran, P.J. (2004). An empirical investigation of alternative methods of 
estimation for confirmatory factor analysis with ordinal data. Psychological Methods, 9, 
466-491. 
Gray, M., Litz, B., Hsu, J., & Lombardo, T. (2004). Psychometric properties of the Life Events 
Checklist. Assessment, 11, 330-341. 
Grewal, R., Cote, J.A., & Baumgartner, H. (2004). Multicollinearity and measurement error in 
structural equation models: Implications for theory testing. Marketing Science, 23, 519-
529. 
Guerra, V. S., & Calhoun, P. S. (2011). Examining the relation between posttraumatic stress 
disorder and suicidal ideation in an OEF/OIF veteran sample. Journal of Anxiety 
Disorders, 25, 12-18. 
Hayduk, L.A. (1996). LISREL issues, debates and strategies. Baltimore, MD: Johns Hopkins 
University Press. 
Hellmuth, J., Stappenbeck, C.A., Hoerster, K.D., & Jakupcak, M. (2012) Modeling PTSD 
symptom clusters, alcohol, and depression as they relate to aggression and suicidality in 
returning U.S. veterans. Journal of Traumatic Stress, 25, 527-534. 
 
 
35 
 
Holden, R.R., & DeLisle, M.M. (2005). Factor analysis of the beck suicide scale for suicidal 
ideation with female suicide attempters. Assessment, 12, 232-238. 
Hu, L., & Bentler, P.M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: 
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. 
Inoue, K., Tanii, H., Kaiya, H., Abe, S., Nishimura, Y., Masaki, M., & Fukunaga, T. (2007). The 
correlation between unemployment and suicide rates in Japan between 1978 and 2004. 
Legal Medicine, 9, 139?142. 
Joiner, T. (2005). Why people die by suicide. Cambridge, MA, US: Harvard University Press. 
Joiner Jr., T.E., Rudd, M.D., & Rajab, M.H. (1997). The Modified Scale for Suicidal Ideation: 
factors of suicidality and their relation to clinical and diagnostic variables. Journal of 
Abnormal Psychology, 106, 260?265. 
J?reskog, K.G. (1999, June 22). How large can a standardized coefficient be? Unpublished 
Technical Report. Retrieved from: 
http://www.ssicentral.com/lisrel/techdocs/HowLargeCanaStandardizedCoefficientbe.pdf 
Kaplan, M. S., McFarland, B. H., Huguet, N., & Newsom, J. T. (2007). Physical illness, 
functional limitations, and suicide risk: A population based study. American Journal of 
Orthopsychiatry, 77, 56?60. 
King, D. W., Leskin, G. A., King, L. A., & Weathers, F. W. (1998). Confirmatory factor analysis 
of the clinician-administered PTSD Scale: evidence for the dimensionality of 
posttraumatic stress disorder. Psychological Assessment, 10, 90-96. 
Kline, R.B. (2011). Principles and Practice of Structural Equation Modeling (3rd Ed.). New 
York, NY: Guilford Press. 
 
 
36 
 
Krysinska, K, & Lester, D. (2010). Post-Traumatic Stress Disorder and Suicide Risk: A 
Systematic Review. Archives of Suicide Research, 14, 1-23. 
McDonald, S.D., & Calhoun, P. (2010). The diagnostic accuracy of the PTSD checklist: A 
critical review. Clinical Psychology Review, 30, 976-987. 
Monteith, L., Menefee, D., Petit, J., Leopoulos, W., & Vincent, J. (2013). Examining the 
interpersonal-psychological theory of suicide in an inpatient veteran sample. Suicide and 
Life-Threatening Behavior, DOI: 10.1111/sltb.12027. 
Muth?n, B., du Toit, S.H.C. & Spisic, D. (1997). Robust inference using weighted least squares 
and quadratic estimating equations in latent variable modeling with categorical and 
continuous outcomes. Unpublished technical report. 
Muth?n, L. (2008, May 16). Warning Message for EFA [Msg 2]. Message posted to 
http://www.statmodel.com/discussion/messages/8/3204.html?1323016665. 
Muth?n, L.K., & Muth?n, B.O. (1998-2012) Mplus User?s Guide (Seventh Edition). Los 
Angeles, California: Muth?n & Muth?n. 
Nock, M. K., Hwang, I., Sampson, N., Kessler, R. C., Angermeyer, M., Beautrais, A., Williams, 
D. R. et al. (2009). Cross-national analysis of the associations among mental disorders 
and suicidal behavior: Findings from the WHO World Mental Health Surveys. PlOS 
Medicine, August 2009. 
Oquendo, M., Brent, D.A., Birmaher, B., Greenhill, L., Kolko, D., Stanley, B., Zelazny, J., 
Burke, A.K,, Firinciogullari, S., Ellis, S.P., & Mann, J.M. (2005). Posttraumatic stress 
disorder comorbid with major depression: Factors mediating the association with suicidal 
behavior. The American Journal of Psychiatry, 162, 560-566. 
 
 
37 
 
Panagioti, M., Gooding, P., & Tarrier, N. (2009). Post-traumatic stress disorder and suicidal 
behavior: A narrative review. Clinical Psychology Review, 29, 471?482. 
Panagioti, M., Googing, P., Dunn, G., & Tarrier, N. (2011). Pathways to suicidal behavior in 
posttraumatic stress disorder. Journal of Traumatic Stress, 24, 137-145. 
Panagioti, M., Gooding, P., & Tarrier, N. (2012). A meta-analysis of the association between 
posttraumatic stress disorder and suicidality: The role of comorbid depression. 
Comprehensive Psychiatry, 53, 915-930. 
Pearl, J. (2000). Causality: Models, Reasoning, and inference. New York: Cambridge University 
Press. 
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and 
comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 
879-891.  
Reddy, M.K., Anderson, B.J., Liebschultz, J., & Stein, M.D. (in press). Factor structure of PTSD 
symptoms in opioid-dependent patients rating their overall trauma history. Drug and 
Alcohol Dependence. Retrieved from http://dx.doi.org/10.1016/j.drugalcdep.2013.04.009. 
Ruggiero, K. J., Del Ben, K., Scotti, J. R., & Rabalais, A. E. (2003). Psychometric properties of 
 the PTSD-Checklist, civilian version. Journal of Traumatic Stress, 16, 495-502. 
Simms, L. J., Watson, D., & Doebbeling, B. N. (2002). Confirmatory factor analyses of   
posttraumatic stress symptoms in deployed and non-deployed veterans of the Gulf  
War. Journal of Abnormal Psychology, 111, 637-647. 
Steer, R.A., Rissmiller, D.J., Ranieri, W.F., & Beck, A.T. (1993). Dimensions of suicidal 
ideation in psychiatric in patients. Behavior Research and Therapy, 31, 229-239. 
 
 
38 
 
Steryn, R., Vawda, N., Wyatt, G.E., Williams, J.K., & Madu, S.N. (2013). Posttraumatic stress 
disorder criteria and suicidal ideation in a South African police sample. African Journal 
of Psychiatry, 16, 19-22. 
Taft, C., Resick, P., Watkins, L., & Panuzio, J. (2009). An investigation of posttraumatic stress 
disorder and depressive symptomology among female victims of interpersonal trauma. 
Journal of Family Violence, 24, 407-415. 
Tarrier, N., & Gregg, L. (2004). Suicide risk in civilian PTSD patients. Predictors of suicide 
ideation, planning and attempts. Social Psychiatry and Psychiatric Epidemiology, 39, 
655?661. 
Van Orden, K.A., Cukrowicz, K.C., Witte, T.K., & Joiner, T.E. (2012). Thwarted belongingness 
and perceived burdensomeness: Construct validity and psychometric properties of the 
Interpersonal Needs Questionnaire. Psychological Assessment, 24, 197-215. 
Wang, L., Long, D., Li, Z., & Armour, C. Posttraumatic stress disorder symptom structure in 
Chinese adolescents exposed to a deadly earthquake. Journal of Abnormal Child 
Psychology, 39, 749-758. 
Weathers, F., Litz, B., Herman, D., Juska, J., & Keane, T. (1993, November). The PTSD 
Checklist (PCL): Reliability, validity, and diagnostic utility. Presentation at the annual 
meeting of the International Society for Traumatic Stress Studies, San Antonio, TX. 
Weathers, F. W., Keane, T. M., & Davidson, J. R. T. (2001). Clinician-Administered PTSD 
Scale: A review of the first ten years of research. Depression and Anxiety, 13, 132-156. 
Wilson, J. P. (2004). PTSD and complex PTSD: Symptoms, syndromes and diagnoses. In J. P. 
Wilson & T. M. Keane (Eds.), Assessing psychological trauma and PTSD (pp. 1-46). 
New York: Guilford. 
 
 
39 
 
Witte, T.K., Joiner, T.E., Brown, G.K., Beck. A.T., Beckman, A., Duberstein, P., & Yeates, C. 
(2006) Factors of suicidal ideation and their relation to clinical and other indicators in 
older adults. Journal of Affective Disorders, 94, 165-172.  
 
 
40 
 
 
 
Appendix A: Tables 
 
  
 
 
41 
 
Table A1   
Descriptive Statistics for Items Used in Structural Analyses. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(continued) 
 
Variable Minimum Maximum Mean  SD Skew Kurtosis 
Reexperiencing       
PCL item 1 1.00 5.00 1.90 1.02 1.27 1.20 
PCL item 2 1.00 5.00 1.67 0.97 1.63 2.34 
PCL item 3 1.00 5.00 1.52 0.89 1.81 2.67 
PCL item 4 1.00 5.00 2.17 1.17 0.84 -0.14 
PCL item 5 1.00 5.00 1.74 1.07 1.48 1.50 
Avoidance       
PCL item 6 1.00 5.00 2.11 1.25 0.96 -0.19 
PCL item 7 1.00 5.00 1.87 1.16 1.25 0.50 
Emotional Numbing       
PCL item 8 1.00 5.00 1.71 1.07 1.53 1.54 
PCL item 9 1.00 5.00 1.49 0.92 0.95 4.10 
PCL item 10 1.00 5.00 1.64 1.05 1.23 1.67 
PCL item 11 1.00 5.00 1.59 1.05 1.73 1.96 
PCL item 12 1.00 5.00 1.47 0.92 2.05 3.54 
Dysphoric Arousal       
PCL item 13 1.00 5.00 1.95 1.27 1.20 0.24 
PCL item 14 1.00 5.00 1.62 1.02 1.62 1.71 
PCL item 15 1.00 5.00 1.98 1.24 1.05 -0.10 
 
 
42 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(continued) 
 
Variable Minimum Maximum Mean SD Skew Kurtosis 
Anxious Arousal       
PCL item 16 1.00 5.00 1.86 1.24 1.34 0.60 
PCL item 17 1.00 5.00 1.67 1.08 1.58 1.54 
Sadness (BDI-II item 1) 1.00 4.00 1.39 0.56 1.23 1.11 
Perceived Burdensomeness       
INQ item 1 1.00 6.00 1.43 0.97 2.60 6.54 
INQ item 2 1.00 6.00 1.39 0.92 2.56 6.17 
INQ item 4 1.00 6.00 1.20 0.64 3.84 15.80 
INQ item 5 1.00 6.00 1.29 0.77 3.30 11.80 
INQ item 6 1.00 7.00 1.64 1.24 2.28 5.06 
Thwarted Belongingness       
INQ item 7 1.00 7.00 1.90 1.41 1.80 3.70 
INQ item 8 1.00 7.00 2.52 1.78 1.10 0.23 
INQ item 9 1.00 7.00 1.91 1.47 1.83 2.63 
INQ item 10 1.00 7.00 1.93 1.57 1.92 3.00 
INQ item 11 1.00 7.00 2.58 1.86 0.94 -0.37 
INQ item 12 1.00 7.00 2.65 1.91 0.90 -0.50 
INQ item 13 1.00 7.00 2.05 1.69 1.74 2.10 
INQ item 14 1.00 7.00 2.29 1.72 1.30 0.70 
INQ item 15 1.00 7.00 2.11 1.69 1.55 1.44 
 
 
43 
 
Variable Minimum Maximum Mean SD Skew Kurtosis 
Suicidal Ideation       
BSS item 1 1.00 1.00 0.02 0.20 6.22 36.90 
BSS item 2 1.00 2.00 0.09 0.30 3.20 9.65 
BSS item 3 0.00 1.00 0.04 0.20 5.00 22.82 
BSS item 4/6 combined 0.00 3.00 0.03 0.26 9.00 78.13 
BSS item 5 0.00 2.00 0.07 0.26 4.01 16.34 
Note. SD = Standard Deviation; PCL= PTSD Checklist ? Specific Version; PB = perceived 
burdensomeness; TB = thwarted belongingness; BDI-II = Beck Depression Inventory-Second Edition; 
BSS = Beck Suicide Scale.
 
 
44 
 
 
Table A2 
PTSD Checklist Inter-item Correlations 
 
Note. PCL= PTSD Checklist ? Specific Version (PCL-S).  
* p ? .05. 
Variable PCL1 PCL2 PCL3 PCL4 PCL5 PCL6 PCL7 PCL8 PCL9 PCL10 PCL11 PCL12 PCL13 PCL14 PCL15 PCL16 
PCL 1 1.00                
PCL 2 .70* 1.00               
PCL 3 .53* .52* 1.00              
PCL 4 .65* .59* .50* 1.00             
PCL 5 .59* .58* .49* .69* 1.00            
PCL 6 .53* .51* .34* .64* .54* 1.00           
PCL 7 .52* .52* .45* .63* .57* .59* 1.00          
PCL 8 .26* .31* .23* .31* .36* .41* .33* 1.00         
PCL 9 .36* .32* .34* .38* .39* .44* .44* .27* 1.00        
PCL 10  .32* .29* .22* .36* .32* .44* .34* .29* .61* 1.00       
PCL 11 .31* .28* .26* .32* .30* .43* .32* .28* .60* .71* 1.00      
PCL 12 .26* .29* .38* .38* .37* .33* .36* .31* .49* .52* .55* 1.00     
PCL 13 .31* .33* .31* .39* .32* .37* .35* .32* .55* .53* .52* .49* 1.00    
PCL 14 .40* .36* .41* .48* .42* .47* .49* .32* .60* .56* .65* .53* .68* 1.00   
PCL 15 .38* .39* .37* .44* .41* .46* .44* .32* .58* .60* .54* .50* .42* .68* 1.00  
PCL 16 .39* .38* .46* .38* .43* .35* .36* .26* .30* .36* .30* .33* .50* .36* .38* 1.00 
PCL 17 .39* .38* .42* .35* .41* .35* .39* .27* .31* .36* .33* .34* .43* .47* .52* .66* 
 
 
45 
 
Table A3 
Bivariate Correlations Between PTSD Checklist Items and Other Study Variables 
 
Note. PCL= PTSD Checklist ? Specific Version (PCL-S); INQ = Interpersonal Needs Questionnaire; BSS = Beck Suicide Scale; 
Sadness = Beck Depression Inventory-Second Edition (Item 1). 
* p ? .05. 
  
Variable PCL1 PCL2 PCL3 PCL4 PCL5 PCL6 PCL7 PCL8 PCL9 PCL10 PCL11 PCL12 PCL13 PCL14 PCL15 PCL16 PCL17 
INQ 1 .17* .28* .23* .22* .22* .28* .26* .23* .38* .50* .49* .40* .47* .47* .42* .24* .25* 
INQ 2 .08 .17* .17* .19* .12* .23* .23* .16* .28* .47* .43* .36* .40* .40* .33* .15* .19* 
INQ 3 .10 .17* .13* .10 .12* .20* .15* .15* .26* .41* .35* .32* .35* .32* .33* .17* .24* 
INQ 4 .11 .18* .20* .20* .15* .21* .23* .21* .27* .42* .38* .45* .38* .39* .34* .20* .30* 
INQ 5 .13* .19* .20* .20* .16* .17* .29* .20* .23* .32* .35* .39* .30* .31* .26* .13* .14* 
INQ 6 .10 .18* .16* .12* .14* .14* .15* .10 .29* .39* .37* .31* .37* .35* .32* .23* .20* 
INQ 7 .07 .10 .10 .12* .06 .14* .15* .13* .32* .36* .40* .37* .28* .27* .21* .06 .08 
INQ 8 .11* .10 .07 .12* .10 .17* .17* .11* .33* .50* .43* .37* .28* .29* .26* .17* .13* 
INQ 9 .06 .06 .14* .08 .09 .19* .17* .01 .31* .28* .33* .23* .29* .27* .22* .16* .16* 
INQ 10  .15* .15* .10 .13* .11 .17* .24* .13* .34* .39* .31* .27* .18* .33* .19* .05 .11* 
INQ 11 .18* .11 .15* .24* .20* .29* .24* .10 .39* .51* .47* .34* .34* .37* .37* .26* .26* 
INQ 12 .16* .11 .10 .21* .21* .22* .24* .11* .33* .46* .40* .28* .28* .31* .34* .26* .30* 
INQ 13 .13* .14* .10 .10 .04 .13* .17* .10 .30* .35* .32* .30* .21* .26* .19* .06 .06 
INQ 14 .18* .15* .11 .18* .10 .17* .21* .11 .33* .45* .43* .31* .25* .34* .21* .13* .11* 
INQ 15 .16* .17* .10* .11 .09 .16* .20* .10 .30* .28* .40* .24* .22* .30* .20* .10 .10 
BSS 1 .10 .12* .14* .13* .15* .13* .10 .14* .22* .23* .23* .27* .20* .14* .23* .20* .13* 
BSS 2 .15* .21* .15* .16* .13* .20* .15* .12* .30* .31* .34* .40* .22* .26* .23* .24* .12* 
BSS 3 .10 .20* .13* .08 .14* .05 .10 .10 .20* .22* .20* .20* .15* .17* .15* .09 .11* 
BSS 4/6 .15* .20* .14* .14* .14* .12* .20* .20* .15* .23* .15* .23* .17* .13* .15* .14* .14* 
BSS 5 
 
 
.10 .16* .05 .10 .10 .05* .07 .11* .20* .22* .29* .24* .13* .12* .14* .03 -.01 
 Sadness .19* .18* .21* .25* .17* .21* .23* .08 .34* .38* .45* .40* .31* .38* .34* .24* .24* 
 
 
46 
 
Table A4 
Bivariate Correlations Between Sadness (BDI-II Item 1), and Items of the Interpersonal Needs Questionnaire, and Beck Suicide Scale  
 
Note. INQ = Interpersonal Needs Questionnaire; BSS = Beck Suicide Scale.* p ? .05.
Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 
1. INQ 1 1.00                    
2. INQ 2  .83* 1.00                   
3. INQ 3 .72* .64* 1.00                  
4. INQ 4 .76* .73* .60* 1.00                 
5. INQ 5 .58* .59* .53* .62* 1.00                
6. INQ 6 .59* .54* .64* .46* .49* 1.00               
7. INQ 7 .38* .45* .28* .30* .34* .28* 1.00              
8. INQ 8 .36* .44* .36* .30* .32* .40* .68* 1.00             
9. INQ 9 .36* .35* .32* .29* .24* .29* .23* .28* 1.00            
10.INQ 10 .27* .32* .27* .24* .22* .26* .61* .60* .26* 1.00           
11.INQ 11 .42* .49* .41* .33* .39* .42* .30* .54* .47* .30* 1.00          
12.INQ 12 .36* .39* .36* .31* .35* .36* .27* .48* .38* .26* .81* 1.00         
13.INQ 13 .40* .47* .36* .29* .30* .34* .66* .63* .29* .68* .38* .33* 1.00        
14.INQ 14 .37* .42* .33* .27* .30* .31* .64* .65* .36* .65* .44* .39* .78* 1.00       
15. INQ 15 .32* .28* .30* .24* .20* .32* .57* .51* .31* .54* .27* .22* .63* .70* 1.00      
16. BSS 1 .34* .28* .35* .32* .17* .26* .13* .12* .25* .18* .24* .21* .12* .19* .13* 1.00     
17. BSS 2 .40* .33* .30* .40* .33* .30* .27* .34* .21* .26* .38* .29* .29* .38* .20* .48* 1.00    
18. BSS 3 .32* .22* .25* .35* .27* .23* .18* .18* .10 .23* .24* .28* .17* .20* .14* .39* .38* 1.00   
19. BSS 4/6 .30* .29* .30* .40* .35* .18* .10* .11* .10 .21* .19* .20* .16* .17* .14* .51* .39* .47* 1.00  
20. BSS 5 .19* .18* .16* .21* .25* .22* .24* .24* .09 .15* .27* .20* .25* .26* .17* .34* .31* .33* .15* 1.00 
21.Sadness  .39* .38* .35* .40* .35* 41* 30* .37* .27* .24* .47* .43* .32* .35* .29* .32* .40* .20* .23* .29* 
 
 
 
Table A5 
Fit Statistics for Three Empirically Supported Models of PTSD Symptom Structure. 
Note. Numbing Model = four-factor numbing model proposed by King et al., 1995; Dysphoria 
model = four factor model proposed by Simms et al., 2002; Elhai model = five-factor model 
proposed by Elhai et al., 2011; RMSEA = Root Mean Square Error of Approximation; CFI = 
Bentler Comparative Fit Index; TLI = Tucker-Lewis Index.
Model  ?2  (p-value) Degrees of 
Freedom 
RMSEA Estimate 
(90% CI) 
CFI TLI 
Numbing Model  
(4-Factors) 
359.16 (p ? .01) 113 .08 (.07-.09) .97 .96 
Dysphoria Model  
(4-Factors) 
252.16 (p ? .01) 113 .06 (.05-.07) .98 .98 
Dysphoric Arousal 
Model (5-Factors) 
216.78 (p ? .01) 109 .06 (.04-.06) .98 .98 
 
 
 
Table A6 
Standardized and Unstandardized Covariance Estimates for the Full Measurement Model 
(continued) 
Covariance  Estimated Estimate /S.E. (p-value) 95 % CI STDYX (p-value) 95 % CI 
P-SI with Rxp 0.30 / 0.07 (p < .01) (0.15 - 0.44) .40 (p < .01) (.23 - .57) 
P-SI with Avoid 0.27 / 0.08 (p < .01) (0.11 - 0.43) .36 (p < .01) (.17 - .56) 
P-SI with Numb 0.32 / 0.05 (p < .01) (0.21 - 0.42) .70 (p < .01) (.58 - .82) 
P-SI with Anx arousal 0.32 / 0.08 (p < .01) (0.15 - 0.48) .40 (p < .01) (.21 - .58) 
P-SI with Dys arousal 0.39 / 0.09 (p < .01) (0.21 - 0.56) .50 (p < .01) (.32 - .68) 
P-SI with Gender -0.03 / 0.10 (p = .74) (-0.23 - 0.17) -.04 (p = .74) (-.26 - .18) 
P-SI with Sadness 0.66 / 0.07 (p < .01) (0.53 - 0.79) .72 (p < .01) (.62 - .81) 
P-SI with TB 0.52 / 0.06 (p < .01) (0.49 - 0.80) .67 (p < .01) (.56 - .78) 
P-SI with PB 0.65 / 0.08 (p < .01) (0.39 - 0.64) .73 (p < .01) (.63 - .83) 
TB with Rxp 0.17 / 0.04 (p < .01) (0.09 - 0.26) .25 (p < .01) (.14 - .37) 
TB with Avoid 0.26 / 0.04 (p < .01) (0.17 - 0.34) .37 (p < .01) (.26 - .49) 
TB with Numb 0.28 / 0.04 (p < .01) (0.21 - 0.36) .68 (p < .01) (.61 - .76) 
TB with Anx arousal 0.23 / 0.05 (p < .01) (0.15 - 0.32) .33 (p < .01) (.21 - .44) 
TB with Dys arousal 0.35 / 0.04 (p < .01) (0.27 - 0.42) .49 (p < .01) (.40 - .58) 
TB with Gender 0.04 / 0.06 (p = .58) (-0.09 - 0.16) .04 (p = .58) (-.23 - .20) 
TB with Sadness 0.48 / 0.04 (p < .01) (0.40 - 0.56) .57 (p < .01) (.48 - .66) 
TB with PB 0.61 / 0.03 (p < .01) (0.54 - 0.67) .75 (p < .01) (.69 - .80) 
PB with Rxp 0.24 / 0.05 (p < .01) (0.13 - 0.34) .30 (p < .01) (.17 - .43) 
PB with Avoid 0.32 / 0.06 (p < .01) (0.21 - 0.43) .41 (p < .01) (.37 - .54) 
PB with Numb 0.33 / 0.04 (p < .01) (0.24 - 0.43) .70 (p < .01) (.61 - .79) 
PB with Anx arousal 0.31 / 0.06 (p < .01) (0.18 - 0.43) .37 (p < .01) (.23 - .51) 
PB with Dys arousal 0.50 / 0.05 (p < .01) (0.41 - 0.60) .62 (p < .01) (.52 - .72) 
PB with Gender 0.03 / 0.08 (p = .74) (-0.14 - 0.19) .03 (p = .74) (-.14 - .20) 
 
 
49 
 
Note. CI = confidence interval; PB = perceived burdensomeness; TB = thwarted belongingness; P-SI = 
passive suicidal ideation, desire and intent; Rxp = reexperiencing; Avoid = avoidance; Numb = emotional 
numbing; Anx arousal = anxious arousal; Dys arousal = dysphoric arousal.  
*p ? .05.**p? .01.  
Covariance Estimated Estimate /S.E. (p-value) 95% CI STDYX (p-value) 95% CI 
PB with Sadness 0.60 / 0.05 (p < .01) (0.50 - 0.69) .62 (p < .01) (.52 - .72) 
Rxp with Avoid 0.61 / 0.04 (p < .01) (0.54 - 0.68) .91 (p < .01) (.86 - .96) 
Rxp with Numb 0.26 / 0.04 (p < .01) (0.19 - 0.33) .64 (p < .01) (.56 - .73) 
Rxp with Anx arousal 0.48 / 0.04 (p < .01) (0.39 - 0.56) .68 (p < .01) (.59 - .76) 
Rxp with Dys arousal 0.44 / 0.04 (p < .01) (0.37 - 0.52) .64 (p < .01) (.55 - .72) 
Rxp with Gender 0.12 / 0.06 (p = .05) (0.00 - 0.25) .15 (p = .05) (.00 - .30) 
Rxp with Sadness 0.26 / 0.05 (p < .01) (0.15 - 0.36) .31 (p < .01) (.19 - .44) 
Avoid with Numb 0.30 / 0.04 (p < .01) (0.22 - 0.36) .75 (p < .01) (.67 - .83) 
Avoid with Anx arousal 0.44 / 0.05 (p < .01) (0.35 - 0.53) .63 (p < .01) (.53 - .73) 
Avoid with Dys arousal 0.49 / 0.04 (p < .01) (0.31 - 0.56) .71 (p < .01) (.62 - .79) 
Avoid with Gender 0.16 / 0.07 (p = .03) (0.02 - 0.30) .20 (p = .03) (.02 - .36) 
Avoid with Sadness 0.30 / 0.06 (p < .01) (0.19 - 0.41) .37 (p < .01) (.24 - .50) 
Numb with Anx arousal 0.27 / 0.04 (p < .01) (0.19 - 0.34) .63 (p < .01) (.53 - .72) 
Numb with Dys arousal 0.37 / 0.04 (p < .01) (0.28 - 0.46) .89 (p < .01) (.85 - .93) 
Numb with Gender 0.17 / 0.04 (p = .69) (-0.07 - 0.10) .03 (p = .70) (-.14 - .21) 
Numb with Sadness 0.30 / 0.04 (p < .01) (0.22 - 0.38) .61 (p < .01) (.50 - .71) 
Anx arousal with  
Dys arousal 
0.50 / 0.04 (p < .01) (0.42 - 0.58) .69 (p < .01) (-.15 - .18) 
Anx arousal with Gender 0.14 / 0.08 (p = .07) (-0.01 - 0.29) .16 (p = .07) (-.01 - .33) 
Anx arousal with Sadness 0.33 / 0.06 (p < .01) (0.21 - 0.46) .39 (p < .01) (.25 - .52) 
Dys arousal with Gender 0.01 / 0.07 (p = .84) (-0.13 - 0.16) .02 (p = .84) (-.15 - .18) 
Dys arousal with Sadness 0.43 / 0.06 (p < .01) (0.31 - 0.53) .50 (p < .01) (.38 - .63) 
Gender with Sadness 0.20 / 0.09 (p = .03) (0.02 - 0.37) .20 (p = .03) (.02 - .37) 
 
 
50 
 
Table A7 
Standardized and Unstandardized Parameter Estimates for the Full Structural Model 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(continued) 
Path Estimated Estimate /S.E. (p-value) STDYX (p-value) 
TB on Gender -0.02 / 0.22 (p = .81) .02 (p = .81) 
P-SI on Gender -0.02 / 0.16 (p = .69) -.04 (p = .67) 
PB on Gender -0.06 / 0.10 (p = .85) .01 (p = .85) 
TB on Reexperiencing -0.27 / 0.25 (p = .28) -.26 (p = .28) 
P-SI on  Reexperiencing 0.86 / 0.52 (p = .10) .77 (p = .09) 
PB on  Reexperiencing -0.23 / 0.30 (p = .43) .38 (p = .43) 
TB on Avoidance -0.02 / 0.30 (p = .96) -.02 (p = .96) 
P-SI on Avoidance -1.09 / 0.62 (p = .08) -.97 (p = .07) 
PB on Avoidance -0.05 / 0.36 (p = .89) -.04 (p = .89) 
TB on Emotional Numbing 2.11 / 0.45 (p < .01) 1.24 (p < .01) 
P-SI on Emotional Numbing 2.32 / 0.93 (p = .01) 1.24 (p = .02) 
PB on Emotional Numbing 1.41 / 0.49 (p < .01) .72 (p < .01) 
TB on Anxious Arousal 0.06 / 0.11 (p = .57) .06 (p = .67) 
P-SI on Anxious Arousal 0.02 / 0.20 (p = .91) .02 (p = .91) 
PB on  Anxious Arousal -0.04 / 0.12 (p = .71) -.04 (p = .71) 
TB on Dysphoric Arousal -0.52 / 0.20 (p < .01) -.53 (p = .01) 
SI on Dysphoric Arousal -0.80 / 0.44 (p = .07) -.73 (p = .08) 
PB on Dysphoric Arousal 0.10 / 0.24 (p = .69) .09 (p = .69) 
TB on Sadness 0.19 / 0.08 (p = .02) .13 (p = .02) 
SI on Sadness 0.30 / 0.12 (p = .02) .18 (p = .02) 
PB on Sadness 0.32 / 0.09 (p <.01) .19 (p <.01) 
P-SI on TB -0.02 / 0.22 (p = .92) -.02 (p = .92) 
P-SI on PB 0.36 / 0.16 (p = .03) .38 (p = .02) 
Numbing with Gender 0.01 / 0.02 (p = .70) .03 (p = .70) 
Numbing with Dysphoric Arousal 0.37 / 0.05 (p <.01) .89 (p <.01) 
 
 
51 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Note. PCL= PTSD Checklist ? Specific Version; PB = perceived burdensomeness; TB = thwarted belongingness; P-
SI = passive suicidal ideation. 
* p ? .05. ** p ? .01. 
Path Estimated Estimate /S.E. (p-value) STDYX (p-value) 
Numbing with Reexperiencing 0.26 / 0.04 (p <.01) .64 (p < .01) 
Numbing with Avoidance 0.30 / 0.20 (p <.01) .75 (p < .01) 
Numbing with Anxious Arousal 0.27 / 0.04 (p <.01) .63 (p < .01) 
Numbing with Sadness 0.13 / 0.02 (p <.01) .47 (p < .01) 
Gender with Dysphoric Arousal 0.01 / 0.02 (p = .85) .01 (p = .85) 
Gender with Reexperiencing 0.04 / 0.02 (p = .05) .11 (p = .05) 
Gender with Avoidance 0.05 / 0.02 (p = .03) .14 (p = .03) 
Gender with Anxious Arousal 0.05 / 0.03 (p = .08) .12 (p = .08) 
Dysphoric Arousal with Reexperiencing 0.44 / 0.04 (p <.01) .12 (p < .01) 
Dysphoric Arousal with Avoidance 0.49 / 0.04 (p <.01) .71 (p < .01) 
Dysphoric Arousal with Anxious Arousal  0.50 / 0.04 (p <.01) -.04 (p < .01) 
Dysphoric Arousal with Sadness 0.19 / 0.03 (p <.01) .40 (p < .01) 
Reexperiencing with Avoidance 0.61 / 0.04 (p <.01) .91 (p < .01) 
Reexperiencing with Anxious Arousal 0.48 / 0.04 (p <.01) .68 (p < .01) 
Reexperiencing with Sadness 0.12 / 0.03 (p <.01) .25 (p <.01) 
Avoidance with Anxious Arousal 0.44 / 0.05 (p <.01) .63 (p < .01) 
Avoidance with Sadness 0.13 / 0.03 (p <.01) .29 (p < .01) 
Anxious Arousal with Sadness 0.15 / 0.03 (p <.01) .31 (p < .01) 
PB with TB 0.17 / 0.04 (p <.01) .49 (p < .01) 
Gender with Sadness 0.03 / 0.02 (p = .05) .11 (p = .05) 
 
 
 
Table A8 
Model Fit Statistics for Models Including Single PTSD Symptom Clusters. 
Note. RMSEA = root mean square error of approximation; CFI = Bentler comparative fit index; TLI = Tucker-Lewis index; R2 
= coefficient of determination; TB = thwarted belongingness; PB = perceived burdensomeness.
PTSD Symptom Cluster Included ?2  (p-value) Degrees of 
Freedom 
RMSEA Estimate 
(90% CI) 
CFI TLI R2 
(TB) 
R2 
(PB) 
R2 
(SI) 
Reexperiencing 561.63 (p?.01) 310 .05 (.04-.06) .98 .98 .22 .27 .62 
Avoidance 526.75 (p ? .01) 238 .06 (.05-.07) .97 .97 .26 .31 .60 
Numbing 598.09 (p ?. 001) 310 .05 (.05-.06) .97 .97 .49 .52 .64 
Dysphoric Arousal 554.70 (p ? .01) 261 .06 (.05-.07) .97 .97 .32 .45 .60 
Anxious Arousal 542.66 (p ? .01) 238 .06 (.06-.07) .97 .96 .23 .29 .61 
 
 
53 
 
Table A9 
Standardized and Unstandardized Parameter Estimates for Structural Models Containing Individual PTSD Symptom Clusters 
Note. TB = thwarted belongingness; P-SI = passive suicidal ideation; PB = perceived burdensomeness; STDYX = standardized solution outputted by Mplus; S.E. 
= standard error; Sx = symptom. * p ? .05. ** p ? .01.  
Parameter Estimated Reexpereincing 
Model 
Avoidance Model Numbing 
Model 
Anxious Arousal 
Model 
Dysphoric Arousal 
Model 
 Estimate (S.E.) STDY
X 
Estimate 
(S.E.) 
STDYX Estimate 
(S.E.) 
STDYX Estimate 
(S.E.) 
STDY
X 
Estimate 
(S.E.) 
STDYX 
TB on Gender -0.06 (0.10) -.03 -0.09 (0.10) -.05 -0.01 (0.10) -.01 -0.07 (0.10) -.04 -0.01 (0.01) -.01 
P-SI on Gender -0.20 (0.16) -.10 -0.18 (0.16) -.09 -0.17 (0.15) -.08 -0.19 (0.16) -.09 -0.17 (0.15) -.08 
PB on Gender -0.11 (0.13) -.05 -0.15 (0.13) -.07 -0.04 (0.13) -.02 -0.12 (0.13) -.06 -0.04 (0.12) -.02 
TB on PTSD Sx Cluster 0.16 (0.06)* .16 0.29 (0.07)** .28 1.33 (0.24)** .62 0.21 (0.06)** .21 0.36 (0.05)** .38 
P-SI on PTSD Sx Cluster 0.17 (0.09)* .16 0.04 (0.12) .03 0.73 (0.30)* .31 0.11 (0.10) .10 0.03 (0.13) .03 
PB on PTSD Sx Cluster 0.22 (0.07)* .20 0.35 (0.09)** .30 1.49 (0.27)** .61 0.28 (0.08)** .25 0.55 (0.06)** .51 
TB on Sadness 0.61 (0.06)** .41 0.55 (0.06)* .37 0.22 (0.07)* .15 0.57 (0.07)** .38 0.44 (0.07)** .29 
P-SI on Sadness 0.39 (0.07)** .24 0.41 (0.08)** .26 0.35 (0.08)** .21 0.40 (0.08)** .24 0.41 (0.08)** .25 
PB on Sadness 0.75 (0.07)** .44 0.69 (0.07)** .41 0.33 (0.08)** .20 0.71 (0.08)** .42 0.49 (0.08)** .28 
P-SI on TB 0.29 (0.11)** .26 0.29 (0.11)* .27 0.17 (0.13) .16 0.29 (0.11)* .26 0.29 (0.11)* .27 
P-SI on PB 0.33 (0.10)* .35 0.36 (0.10)** .37 0.25 (0.11)* .27 0.34 (0.10)** .36 0.35 (0.10)** .37 
TB with PB 0.41 (0.03)** .67 0.37 (0.04)** .65 0.19 (0.04)** .49 0.39 (0.04)** .66 0.30 (0.10)** .61 
Gender with PTSD Cluster 0.04 (0.02) .11 0.05 (0.03)* .14 0.01 (0.01) .03 0.05 (0.03) .12 0.01 (0.03) .01 
Gender with Sadness 0.03 (0.02) .11 0.03 (0.02) .11 0.03 (0.02) .11 0.03 (0.02) .11 0.03 (0.02) .11 
PTSD Cluster with Sadness 0.12 (0.03)** .25 0.13 (0.03)** .29 0.10 (0.02)** .48 0.14 (0.03)** .31 0.20 (0.03)** .40 
 
 
54 
 
Table A10 
Indirect Effects of Each PTSD symptom cluster on Suicidal Ideation Through Perceived Burdensomeness and Thwarted 
Belongingness. 
 
 
 
 
 
 
 
 
 
Note. PTSD = posttraumatic stress disorder; SI = suicidal ideation; PB = perceived burdensomeness; TB = thwarted belongingness.  
* p ? .05. ** p ? .01.
PTSD Symptom Cluster 
Included 
Specific Indirect Effect on 
SI through PB 
Specific Indirect Effect on 
SI through TB 
 Estimate (95 % CI) STDYX (95 % CI) Estimate (95 % CI) STDYX (95 % CI) 
Reexperiencing 0.07 (0.00 ? 0.14) 0.07* (0.01 ? 0.13) 0.04 (0.00 ? 0.11) 0.04 (-0.01 ? 0.09) 
Avoidance 0.13 (0.01 ? 0.23) 0.11* (0.02 ? 0.20) 0.08 (0.02 ? 0.20) 0.07* (0.01 ? 0.14) 
Numbing 0.38 (-0.10 ? 0.76) 0.16* (0.00 ? 0.32) 0.10 (-0.20 ? 0.65) 0.10 (-0.05 ? 0.25) 
Dysphoric Arousal 0.19 (0.01 ? 0.33) 0.19* ( .04 ? 0.33) 0.10 (0.02 ? 0.21) 0.10* (.02 ? 0.18) 
Anxious Arousal 0.10 (0.01 ? 0.19) 0.09* (0.01 ? 0.17) 0.06 (0.01 ? 0.16) 0.05 (0.00 ? 0.12) 
 
 
55 
 
 
 
Appendix B: Figures 
 
     56 
 
56 
 
Figure B1 
Full Structural Model 
Note. Covariance values between all PTSD symptom clusters, gender, and sadness, and between PB and 
TB were also estimated, but are not depicted in this diagram to enhance readability. BDI-II = Beck 
Depression Inventory-Second Edition; SI = suicidal ideation.
 
57 
 
Figure B2 
Individual Symptom Cluster Models with all Specified Paths.  
Note. BDI = Beck Depression Inventory-Second Edition; PTSD = posttraumatic stress disorder; 
Sx = symptom.
   
 
 
 
Appendix C: Item Content 
  
 
59 
 
Table C1 
PCL Item Mapping for Numbing, Dysphoria, and Dysphoric Arousal CFA Models 
PCL-S 
Item 
DSM-IV PTSD Symptom Description 
 
Numbing 
Model 
Dysphoria 
Model 
Dysphoric 
Arousal Model 
1 B1 Intrusive thoughts of trauma R I R 
2 B2 Recurrent dreams of trauma R I R 
3 B3 Flashbacks R I R 
4 B4 Emotional reactivity to trauma cues R I R 
5 B5 Physiological reactivity to trauma cues R I R 
6 C1 Avoiding thoughts of trauma A A A 
7 C2 Avoiding reminders of trauma A A A 
8 C3 Inability to recall aspects of trauma N D N 
9 C4 Loss of interest N D N 
10 C5 Detachment N D N 
11 C6 Restricted affect N D N 
12 C7 Sense of foreshortened future N D N 
13 D1 Sleep disturbance H D DA 
14 D2 Irritability H D DA 
15 D3 Difficulty Concentrating H D DA 
16 D4 Hypervigilance H H AA 
17 D5 Exaggerated startle response H H AA 
Note. Numbing Model = four-factor numbing model proposed by King et al., 1995; Dysphoria model = 
four factor model proposed by Simms et al., 2002; Dysphoric Arousal model = five-factor model 
proposed by Elhai et al., 2011; PCL-S= PTSD Checklist, specific version; PTSD = posttraumatic stress 
disorder; I = Intrusions; R = Re-experiencing; A = Avoidance; N = Emotional Numbing; H = 
Hyperarousal; D = Dysphoria; DA = Dysphoric Arousal; AA = Anxious Arousal. 
 
60 
 
Table C2 
Content Descriptors For All Item From the Interpersonal Need Questionnaire, Beck Suicide 
Scale, and Beck Depression Inventory (Second Edition) Used in Analyses. 
(continued) 
Scale (Latent Factor) Item Number Item Content 
INQ (Perceived Burdensomeness)   
 INQ item 1 Better off without me 
 INQ item 2 Happier without me 
 INQ item 3 Burden to society 
 INQ item 4 Death as a relief 
 INQ item 5 Rid of me 
 INQ item 6 Makes things worse 
INQ (Thwarted Belongingness)   
 INQ item 7* Others care about me 
 INQ item 8* Feel like I belong 
 INQ item 9 Rarely interact with caring others 
 INQ item 10* Have caring/supportive friends 
 INQ item 11 Feel disconnected 
 INQ item 12 Feel like an outsider 
 INQ item 13* People I can turn to 
 INQ item 14* Close to others 
 INQ item 15* One satisfying interaction/day 
 
61 
 
 Note. INQ = Interpersonal Needs Questionnaire; BSS = Beck Suicide Scale; BDI-II = Beck 
Depression Inventory (Second Edition); * Denotes items that were reverse coded for analyses. 
 
 
 
 
Scale (Latent Factor) Item Number Item Content 
BSS (Suicidal Ideation)   
 BSS item 1 Wish to live 
 BSS item 2 Wish to die 
 BSS item 3 Reasons for living ( > for dying) 
 BSS item 4/6 Desire for suicide/ Length and duration 
of SI 
 BSS item 5 Passive desire for suicide 
Sadness   
 BDI-II  item 1 Sadness