Posttraumatic Stress Disorder, Suicidal Ideation, and Suicidal Self-Directed Violence Among U.S. Military Personnel and Veterans: A Systematic Review of the Literature From 2010 to 2018

Rates of suicide and posttraumatic stress disorder remain high among United States military personnel and veterans. Building upon prior work, we conducted a systematic review of research published from 2010 to 2018 regarding: (1) the prevalence of suicidal ideation, suicide attempt, and suicide among United States military personnel and veterans diagnosed with posttraumatic stress disorder; (2) whether posttraumatic stress disorder was associated with suicidal ideation, suicide attempt, and suicide among United States military personnel and veterans. 2,106 titles and abstracts were screened, with 48 articles included. Overall risk of bias was generally high for studies on suicidal ideation or suicide attempt and low for studies on suicide. Across studies, rates of suicidal ideation, suicide attempt, and suicide widely varied based on study methodology and assessment approaches. Findings regarding the association between posttraumatic stress disorder diagnosis with suicidal ideation and suicide were generally mixed, and some studies reported that posttraumatic stress disorder was associated with lower risk for suicide. In contrast, most studies reported significant associations between posttraumatic stress disorder and suicide attempt. These findings suggest complex associations between posttraumatic stress disorder and suicidal ideation, suicide attempt, and suicide, which are likely influenced by other factors (e.g., psychiatric comorbidity). In addition, most samples were comprised of veterans, rather than military personnel. Further research is warranted to elucidate associations between posttraumatic stress disorder and suicidal ideation, suicide attempt, and suicide, including identification of moderators and mediators of this relationship. Addressing this among United States military personnel, by gender, and in relation to different trauma types is also necessary.


INTRODUCTION
Within the United States (U.S.), suicide remains a significant public health concern, with the Centers for Disease Control and Prevention (CDC) recently reporting it as the tenth overall leading cause of death (Heron, 2018). Risk for suicide is especially pronounced among U.S. military personnel and veterans, among whom adjusted suicide rates have, at times, outpaced suicide rates in the general U.S. non-veteran adult population (Reimann and Mazuchowski, 2018; Department of Veterans Affairs, 2019). As such, preventing suicide among military personnel and veterans remains a top clinical priority of the Departments of Defense and VA.
Efforts have been made to understand why U.S. military personnel and veterans are at increased risk for suicide, compared to civilians. Although suicide is understood to be etiologically complex, one potential conduit of increased risk is through traumatic experiences and their sequelae. Military personnel and veterans experience trauma with heightened propensity, both in terms of military-and non-military-related trauma, such as combat-related experiences, military sexual assault, childhood abuse, and intimate partner violence (Gates et al., 2012;Lehavot et al., 2018). Military personnel and veterans also experience high rates of posttraumatic stress disorder (PTSD) (Gates et al., 2012;Lehavot et al., 2018). As such, researchers have posited elevated rates of PTSD as a potential explanation for suicidal ideation (SI), suicide attempt (SA), and suicide among service members and veterans (Pompili et al., 2013). Indeed, prior systematic reviews and meta-analyses of the relationship between PTSD and suicide have reported significant associations between PTSD and suicide risk (Krysinska and Lester, 2010;Kanwar et al., 2013;Panagioti et al., 2015). However, to date, only one such review has focused specifically on military personnel and veterans (Pompili et al., 2013). Pompili et al. (2013) conducted a systematic review of literature published from 1980 to 2010 regarding the association between PTSD with SI and suicidal self-directed violence (S-SDV; e.g., suicide attempt, suicide) among U.S. and Canadian military personnel and veterans. Based on their review of 18 studies, they concluded that PTSD was associated with SI, SA, and suicide. Pompili et al. further noted, however, that PTSD was associated with several mental health outcomes (e.g., psychiatric comorbidity), which may have partially accounted for the reported associations. As such, it remains difficult to ascertain to what extent PTSD independently explains heightened risk for SI, SA, and suicide among U.S. military personnel and veterans.
While the aforementioned systematic review was seminal in its focus on military personnel and veterans, the review focused on "war-related PTSD" (e.g., combat-related), limiting inference regarding PTSD from other prevalent military-or non-military related traumatic exposures (e.g., military sexual assault, childhood abuse). Another limitation of the prior review was the inclusion of studies focused on PTSD symptoms, limiting the ability to draw precise inferences regarding PTSD diagnosis, as those without a diagnosis of PTSD may have been included. In addition, since 2010, veterans from the recent conflicts in Afghanistan and Iraq have experienced high rates of deployment-related experiences, such as combat and sexual assault (Street et al., 2009;Vasterling et al., 2010;Barth et al., 2016). While some of these experiences are not necessarily unique to deployment (e.g., sexual assault can occur during training or while stateside), deployment can disrupt pre-deployment life and family functioning. For example, Paley et al. (2013) noted that deployment can introduce a number of challenges, including disruptions in family routine, extended separation from friends and family, and parenting challenges following return due to mental health sequelae. These disruptions appear particularly salient in driving PTSD symptomatology following militaryrelated trauma (Polusny et al., 2014). This, in turn, may result in differing clinical presentations in more recent research with military personnel and veterans that could impact associations of PTSD with SI and S-SDV. As such, an updated review of the literature published since 2010 is critical.
Finally, a number of prior reviews discussed SI, SA, and suicide as categorically-similar constructs, rather than differentiating between these outcomes. This is problematic to discerning optimal clinical care to mitigate risk among service members and veterans with PTSD . The VA and Department of Defense have mandated a specific classification system and nomenclature for suicide, the Centers for Disease Control and Prevention's (CDC) Suicidal Self-Directed Violence Classification System (Crosby et al., 2011), to distinguish these constructs. The CDC defines suicidal ideation as "thoughts of engaging in suicide-related behavior" (p. 90), suicide attempt as "a non-fatal self-directed potentially injurious behavior with any intent to die as a result of the behavior. . . which may or may not result in injury" (p. 21), and suicide as "death caused by self-directed injurious behavior with any intent to die as a result of the behavior" (p. 23). As these constructs have distinct underlying theoretical underpinnings, as well as divergent risk factors (Joiner, 2007;Nock et al., 2008;Klonsky and May, 2015;Klonsky et al., 2016;May and Klonsky, 2016), delineating the extent to which a diagnosis of PTSD is associated with SI, SA, and suicide remains crucial.
The current systematic review aimed to address these limitations and provide an enhanced update of research examining the association between PTSD diagnosis and SI, SA, and suicide among U.S. military personnel and veterans. To update prior work (c.f. Pompili et al., 2013) (Moher et al., 2009;Matchar, 2012). In particular, inclusion criteria were as follows: Population(s): U.S. military personnel and/or veterans; Intervention/Exposure(s): (1) Assessment and diagnosis of PTSD and (2) assessment or documentation of SI, SA, or suicide; Comparator(s): A comparison group was not required for KQ1; however, for KQ2, a comparator of no PTSD diagnosis was required; Outcome(s): Prevalence of SI, SA, or suicide among those with PTSD (KQ1); reported on the association between PTSD and SI, SA, or suicide (KQ2); and Timing/Setting: No restrictions based on timing, setting, or study design. Additional criteria for inclusion were: (1) presentation of original study data in a peer-reviewed journal article; (2) adequate data to address KQ1 and/or KQ2 (i.e., reported rate and/or association between PTSD and SI, SA, or suicide within text); (3) the full-text article was in English; and (4) published between January 1, 2010 and April 25, 2018. Exclusion criteria were as follows: (1) duplicate datasets (i.e., re-analysis of a previously reported dataset) and (2) inadequate data (i.e., inability to calculate rates or associations based on content reported within the manuscript); (3) dissertations, conference proceedings, commentaries, editorials, letters, books, book chapters, duplicate datasets, and reviews.
Databases were searched sequentially on the same day. Complete citations were exported and de-duplicated using EndNote X8 reference management software (Thomson Reuters, New York City, NY, USA). References were then exported into Covidence review software. Review of studies for inclusion was based on previously-used systematic review frameworks, including those used by the study team (Hoffberg et al., 2020).
For the PRISMA screening stage, at least two reviewers (RH, LMB, KSY, LAB) independently screened each title and abstract for retrieval. When not in agreement, a third reviewer (LLM) evaluated the record for the final retrieval decision. All coauthors have experience conducting research on PTSD and SI and S-SDV among U.S. veterans and have previously published in this domain (e.g., Brenner et al., 2011a;Holliday et al., 2018b;Barnes et al., 2019;Monteith et al., 2019).
Covidence was also used to evaluate records selected for the PRISMA eligibility stage of the review. Each full-text record was assessed by at least two reviewers (RH, LMB, KSY, LAB). The decision process was stepwise and based on the PICOTS model. Each reviewer progressed through the PICOTS decision tree until either all inclusion criteria were met or the record was excluded for a particular reason. Any disagreements at this stage were similarly resolved by a third blind reviewer (LLM).
Data from full-text articles selected for inclusion were abstracted into tables by three authors (RH, LMB, and LLM). Conflicts were resolved by group consensus, with RH making the final determination for inclusion. The preliminary data abstraction template was tested before being finalized. The following information was extracted: source article, population/sample (e.g., composed of military personnel or veterans, proportion of males and females within the sample, index trauma), measurement of PTSD, measurement of SI/SA/suicide, prevalence of SI/SA/suicide among those diagnosed with PTSD (KQ1), association of PTSD to SI/SA/suicide (KQ2), and additional relevant results. As variability of study designs and outcome measurements precluded a meta-analytic approach, a descriptive synthesis approach was conducted (McKenzie and Brennan, 2019). For each study included, the sample was described based on composition (i.e., military personnel or veteran) and proportion of males/females; however, due to infrequent reporting across studies, index trauma could not consistently be assessed and was thus not reported.
Included articles were classified by study design using the Taxonomy of Study Design Tool (Hartling et al., 2010) in a custom Research Electronic Data Capture (REDCap) database (Harris et al., 2009). Study design was assessed independently (KSY, ASH), with conflicts resolved by a third author (RH). Each of these authors had prior experience conducting systematic reviews or meta-analyses (Bahraini et al., 2013;Hoffberg et al., 2018Hoffberg et al., , 2020Holliday et al., 2018a;Creech et al., 2019).
After reaching consensus on study design, risk of bias was assessed using the Effective Public Health Practice Project (EPHPP) quality assessment tool for quantitative studies (Thomas et al., 2004). Bias items included selection bias, study design, confounders, blinding, data collection, withdrawals/dropouts, and other sources (e.g., no disclosure of conflicts of interest). Each of these domains, if applicable, were rated as having a low, moderate, or high risk of bias based on standard guidelines (see Thomas et al., 2004 for additional information). An overall risk of bias rating was then generated using standard guidelines (Effective Public Health Practice Project, 1998a,b), such that studies rated as having no high risk of bias domain ratings were classified as low overall risk of bias, studies having up to one high risk of bias domain rating were classified as moderate overall risk of bias, and studies having two or more high risk of bias domain ratings were classified as high overall risk of bias. Risk of bias was evaluated independently by KSY and ASH, with discrepancies discussed among KSY, ASH, and RH until achieving consensus. Both KSY and ASH had prior experience evaluating risk of bias in prior systematic reviews (e.g., Hoffberg et al., 2018Hoffberg et al., , 2020.

Study Selection
As depicted in Figure 1, 2,106 titles and abstracts identified in electronic searches were screened after removing duplicates.
Following screening, 399 full-text articles were assessed for eligibility. Based on our PICOTS criteria, 351 of these fulltext articles were subsequently excluded, resulting in 48 articles included for KQ1 and/or KQ2. Risk of bias for these articles is reported in Table 1.   (Table 2). However, the overwhelming majority of included studies (n = 13) had a high overall risk of bias. Three had an overall moderate risk of bias and reported rates of SI ranging from 35.71 to 44.90% (Magruder et al., 2012;Denneson et al., 2014;Cox et al., 2016). Only two studies had a low overall risk of bias, reporting that 3.21-34.81% of veterans with PTSD had SI (Corson et al., 2013;Cohen et al., 2015). Of studies relying on interview or assessment measures, current/recent rates of SI ranged from 1.98 to 44.9%, whereas lifetime rates varied more broadly, from 17.39 to 71.89%. All of these studies focused on veterans; none focused exclusively on military personnel. Studies evaluating lifetime SI often comprised fairly small clinical samples, whereas studies reporting on more recent SI involved much larger samples, typically comprising treatmentseeking Operation Enduring Freedom/Operation Iraqi Freedom (OEF/OIF) veterans.

KQ1: Prevalence of SA Among Military Personnel and Veterans With PTSD
The eight studies on rates of SA among those diagnosed with PTSD reported rates ranging from 0.31 to 38.71% (Table 3). Only one included study was determined to have a low overall risk of bias (Gradus et al., 2014), and one had a moderate overall risk of bias . Of note, all of the studies on the prevalence of SA among those with PTSD focused on veterans; none reported on rates of SA among military personnel with PTSD.

KQ1: Prevalence of Suicide Among Military Personnel and Veterans With PTSD
Eight studies were included that reported on the prevalence of suicide among military personnel and veterans diagnosed with PTSD ( Table 4). Strength of evidence appeared strong for these studies, as the majority of included studies were determined to have a low overall risk of bias (n = 5; Ilgen et al., 2010b;Brenner et al., 2011b;Bohnert et al., 2013;Louzon et al., 2016;Shen et al., 2016). Four studies reported the percent of military personnel or veterans with PTSD who died by suicide, with a range of <0.01-0.30% (Brenner et al., 2011b;Bohnert et al., 2013;Conner et al., 2014;Louzon et al., 2016). An additional four studies reported rates per person-years, ranging from 55.1 to 159.9 per 100,000 (Ilgen et al., 2010b;Black et al., 2011;Bachynski et al., 2012;Louzon et al., 2016), with rates ranging based on timeframe (e.g., per-person year, per-person quarter).

KQ2: Association of PTSD With SI Among Military Personnel and Veterans
Nineteen studies reported on the association between PTSD and SI. The majority of included studies were determined to have a high overall risk of bias (n = 13), four had a moderate risk of bias (Magruder et al., 2012;Denneson et al., 2014;Monteith et al., 2016;Kimbrel et al., 2018a), and only two studies had a low overall risk of bias (Corson et al., 2013;Finley et al., 2015). Bivariate associations between PTSD and SI were generally significant (significant in ten studies, non-significant in three studies; Table 2). In contrast, results regarding the association between PTSD and SI at the multivariate level were mixed. Of the fourteen studies reporting multivariate results regarding the association between PTSD and SI, seven reported significant associations, and seven reported non-significant associations. All of the studies reporting significant bivariate or multivariate associations between PTSD and SI reported that these were positively, rather than inversely, associated.

KQ2: Association of PTSD With SA Among Military Personnel and Veterans
Nine studies examined the relationship between PTSD and SA. Only three had a low overall risk of bias (Brenner et al., 2011a;Gradus et al., 2014;Finley et al., 2015), all of which reported significant associations between PTSD and SA ( Table 3). In fact, all but two studies reported significant positive associations at the bivariate and/or multivariate level. In addition, all but one study focused on veterans (versus military personnel).
PTSD was not associated with SI at a bivariate level, χ 2 = 0.08, p = 0.773.

DISCUSSION
Given the continued rise in rates of suicide among U.S. military personnel and veterans (Reimann and Mazuchowski, 2018; Department of Veterans Affairs, 2019), understanding the extent to which common mental health diagnoses, such as PTSD, are associated with SI, SA, and suicide is important. Building upon the systematic review by Pompili et al. (2013), the current systematic review provides an update of literature spanning 2010-2018, focused on U.S. service members and veterans. Importantly, this review is inclusive of cohorts spanning multiple service eras, including those who served in the recent conflicts based in Afghanistan and Iraq, a population with notably high rates of suicide (Department of Veterans Affairs, 2019).

KQ1
This systematic review is among the first to examine the prevalence of SI, SA, and suicide among U.S. military personnel and veterans with PTSD. Attempts to determine rates of SI, SA, and suicide across studies were largely impacted by differences in study methodology. Studies varied in their methods of assessing PTSD and suicide-related constructs (e.g., use of electronic medical records vs. validated semi-structured interviews), as well as in the temporality of these variables (e.g., lifetime vs. past month). This resulted in highly variable rates, especially among studies examining SI. The majority of included studies examining KQ1 for SI and SA were also rated as having high bias in study methodology. This is likely driven, in part, by the fact that many of these studies were not designed to examine the prevalence of SI or SA. Rather, a number of studies examined PTSD, SI, or SA as secondary outcomes or covariates. In contrast, studies on suicide tended to have a low overall risk of bias, although suicide rates were still quite variable. Stronger methodology and consistent reporting regarding assessment and timeframe of PTSD, SI, and SA is needed for future research to ensure accurate depiction of reported rates.

KQ2
Research examining if PTSD is associated with SI and S-SDV was also mixed, suggesting a complex relationship. Only PTSD Among those with PTSD, the suicide mortality rate per 100,000 person-years was 159.5.
Among those with PTSD, the RR was 5.6 (95% CI: 3.6-8.7, p < 0.05). VA National Patient Care Database NDI 226 (.11%) of those with PTSD died by suicide determined to be an "intentional overdose." * Unadjusted analyses found PTSD was associated with a relative risk of 4.46 (95% CI: 3.85-5.18, p < 0.05) as it relates to suicides determined to be an "intentional overdose." Veterans were followed through FY 2006. Study classified suicide by overdose as "intentional", "indeterminate intent", and "unintentional." Only "intentional was reported based on purpose of paper.
Brenner et al.  and SA appeared to have a significant association that was largely maintained in the presence of covariates. In contrast, for studies reporting on the association between PTSD and SI or suicide, multivariate findings diverged, with some studies reporting nonsignificant associations. Further, in some studies, PTSD was even associated with decreased risk for suicide, further complicating conceptualization of this relationship. Mixed findings regarding the association between PTSD with SI or suicide were maintained even when restricting studies to those with low overall risk of bias. This suggests that while study methodology and rigor are pertinent to synthesizing the literature base, factors outside of study quality may produce varied findings regarding the relationship between PTSD with both SI and suicide.
Several potential explanations can be posited as to why the relationship between PTSD and SI or suicide may be less robust. First and foremost, inclusion of other important correlates of SI and suicide may be more explanatory of risk than PTSD. Several included studies found other variables to be significant correlates of SI, SA, and suicide, including psychiatric comorbidities, psychosocial functioning (e.g., socioeconomic status, social support), and sociodemographic factors. In particular, across several studies, depression was a robust predictor of SI and S-SDV and thus may be more explanatory of suicide risk among U.S. military personnel and veterans with PTSD.
It is also important to note that a number of the studies which reported non-significant relationships between PTSD and SI or S-SDV relied on electronic medical records. There is sizable variability in the diagnostic validity of data obtained from medical records compared to diagnostic interviews , suggesting that this may be an important factor to consider when interpreting these findings. The recency (e.g., past-month vs. lifetime) of PTSD, SI, and SA also can be difficult to ascertain from electronic medical records, which may further impact findings.
While our systematic review focused on the presence or absence of a PTSD diagnosis, a dichotomous focus on the presence or absence of a PTSD diagnosis likely does not provide sufficiently nuanced information regarding factors inherent to, or associated with, PTSD that may be driving risk within this population. The clinical presentation of PTSD can vary largely between patients based on heterogenous symptom profiles. In addition, specific symptoms of PTSD, such as guilt, social isolation, and trauma-related beliefs, have been posited as risk factors for suicide among military personnel and veterans (Bryan et al., 2013DeBeer et al., 2014;Legarreta et al., 2015;McLean et al., 2017;Holliday et al., 2018a;Borges et al., 2020), but were not the focus of the current review. More research is needed to understand if specific components of PTSD are more strongly related to SI and suicide among military personnel and veterans.
Despite these potential explanations, none provide insight as to why the overwhelming majority of studies on SA found that PTSD was associated with SA, while many studies on SI and suicide did not. Research noting inherent differences in what motivates progression from SI to S-SDV (Nock et al., 2008) may explain some of these differences in findings regarding SI and SA, but would not necessarily explain inconsistencies between findings on SA vs. suicide. One alternate explanation is that, for KQ2, the number of SA studies, particularly low risk of bias studies, were far fewer; as such, inclusion of additional studies may regress toward a more confident understanding of the relationship between PTSD diagnosis and SA. Overall risk of bias for studies on SA were largely rated as moderate to high, leading to potentially spurious findings in comparison to the majority of studies on suicide, which were generally rated as low. This further reinforces the need for additional investigation of the relationship between PTSD and SA, using consistent, sound methodology.
Additional factors underlying S-SDV may also explain variance. For instance, differing means of attempting suicide (e.g., firearms vs. overdose) vary in lethality, as the overwhelming majority of individuals who use a firearm to attempt suicide die (Spicer and Miller, 2000). It is possible that those who survive a SA using a less lethal means differ in their precipitants and drivers of suicide risk in comparison to those who use more lethal means (e.g., firearms). Given that military personnel and veterans have high rates of access to highly lethal means (e.g., firearms; Cleveland et al., 2017) as well as the potential interrelationship between trauma and firearm access Sadler et al., 2020;Simonetti et al., 2020;Stanley et al., 2020), longitudinal research must be prioritized to further understand if PTSD is associated with using different types of lethal means to enact S-SDV.

Limitations and Future Research
While this systematic review provides important insight into research published since 2010, several factors limit comprehensive inferences regarding this body of literature. A number of studies did not report rates of SI, SA, or suicide specific to those diagnosed with PTSD, precluding inclusion for KQ1. Included study samples were also overwhelmingly male and veteran, which generally reflects the compositions of the U.S. military and veteran populations. Nonetheless, as most studies reviewed did not adequately sample women or report rates or associations separately based on gender, future studies should consider oversampling women to facilitate exploring whether associations between PTSD with SI, SA, and suicide differ by gender.
Moreover, while our review expanded beyond war-related trauma (Pompili et al., 2013) to also include non-combat-related trauma (e.g., military sexual assault), only a limited number of studies specifically assessed and reported on trauma type or focused on non-combat-related PTSD. Because of this, it was not possible to differentiate rates and associations based on types of trauma. Therefore, given research suggesting that risk (e.g., for SI) may differ based on trauma type (Blais and Monteith, 2019), researchers should assess and report type of index trauma among their samples, especially in research pertaining to SI, SA, and suicide.
The current systematic review also focused on PTSD diagnosis, excluding studies relying on self-report symptom inventories (e.g., PTSD Checklist). While these inventories are psychometrically valid and can determine probable diagnoses, they should not be used to infer a formal diagnosis. As this was a departure from prior systematic reviews (e.g., Pompili et al., 2013), it may contribute to differences in findings. A number of studies also used single-item self-report measures of SI and SA. While these items were face valid, their psychometric properties relative to a clinical interview or formal assessment measure is debatable. Further inquiry is needed to understand how method of assessing SI and SA may impact findings.
Included studies were also predominantly cross-sectional. While still informative, such studies were rated as having higher risk of bias as this type of design precludes understanding the temporal relationship between PTSD with SI and SA (i.e., if PTSD precedes subsequent SI and SA). Longitudinal research was limited in this body of literature, and in particular, prospective cohort studies are warranted to further elucidate drivers of suicide risk.
Finally, studies largely differed in the factors that they accounted for in multivariate analyses. Because analytic approaches differed across studies, it is difficult to infer if associations between PTSD with SI, SA, and suicide would have differed based on inclusion of additional variables. As such, researchers should ensure multivariate analyses with PTSD also include correlates of SI, SA, and suicide identified in this review (e.g., depression) to better understand the magnitude and direction of the association with PTSD. Consistent inclusion of these factors would also facilitate future research focused on understanding potential moderators and mediators of the association between PTSD and SI, SA, and suicide (e.g., meta-regression).

CONCLUSIONS
When interpreting findings within the context of these limitations, this systematic review provides insight into the prevalence of SI, SA, and suicide among U.S. military personnel and veterans. Results provide continued, yet tentative, support that PTSD diagnosis is likely associated with SI, SA, and suicide at a bivariate level. Because of this, clinicians should continue to assess for PTSD, as well as associated psychiatric comorbidities (e.g., depression), in the context of suicide risk assessment when working with trauma-exposed U.S. military personnel and veterans. Similarly, clinicians should continue to screen for SI and prior SA when working with military personnel and veterans diagnosed with PTSD. For many service members and veterans, PTSD may not be the sole driver of suicide risk, as SI, SA, and suicide are likely the result of an accumulation of numerous comorbid risk factors (e.g., depression) and other traumarelated sequelae (e.g., low social support; Lemaire and Graham, 2011). It is important to note that PTSD likely also increases risk for psychiatric comorbidity and decreased psychosocial functioning (Pietrzak et al., 2010;Hefner and Rosenheck, 2019). Therefore, PTSD may increase risk through indirect pathways (e.g., PTSD increases risk for depression, which, in turn, increases risk for suicide; e.g., McKinney et al., 2017).
Given the lack of consistency in multivariate results, further synthesis of the literature remains warranted. Analytic approaches that account for methodological differences between studies (e.g., diagnostic interview vs. electronic medical record) and also synthesize the role of potential covariates and other notable risk factors (e.g., depression) across studies is an integral next step. Additional understanding of specific aspects of PTSD (e.g., guilt, social isolation, trauma-related beliefs) that potentially moderate or mediate the relationship between PTSD diagnosis with SI, SA, and suicide would also further elucidate factors potentially impacting risk among U.S. military personnel and veterans. Identification of specific drivers of risk would be particularly important to inform clinical assessment and evidence-based treatment to prevent SI and S-SDV among U.S. military personnel and veterans with PTSD.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.