Distinguishing veterans with suicidal ideation from suicide attempt history: The role of emotion reactivity

Abstract

Introduction:

Growing evidence suggests emotion reactivity—sensitivity and intensity of emotional experience—may represent a diathesis for suicide risk. However, our understanding of its ability to differentiate risk for suicidal ideation (SI) from suicide attempt (SA) is limited.

Method:

This study compares Veterans with SI (n = 81) to Veterans with SA (n = 177) history on factors relevant to emotion reactivity to determine which variable(s) best differentiate groups. Variables examined are multimodal: (a) self-report: childhood trauma, combat exposure; (b) clinician-assessed: non-suicidal self-injury (NSSI), structured diagnostic interview of psychopathology; and (c) psychophysiological: affect-modulated startle (AMS; proxy for amygdala reactivity and emotion reactivity) to unpleasant pictures was examined in a subset (n = 90).

Results:

SA history was independently predicted by NSSI history, MDD, PTSD, and SUD diagnosis. Childhood trauma and combat exposure did not differentiate groups. The composite risk index demonstrated good accuracy (AUC=0.71, sensitivity=0.90, specificity=0.49). Only AMS independently predicted SA history when added to the model and accuracy was improved (AUC=0.82, sensitivity=0.85, specificity=0.56).

Conclusion:

NSSI history, MDD, PTSD, and SUD diagnosis may be salient risk factors for this population. However, emotion reactivity is a more parsimonious predictor of SA history among Veterans suggesting it is an important treatment target among Veterans with SI.

INTRODUCTION

Suicide is among the leading causes of death worldwide(World Health Organization, 2018). Veterans comprise only 7.9% of the overall U.S. population but account for approximately 13.5% of suicide deaths (Office of Mental Health & Suicide Prevention, 2019). Further, Veteran suicide rates increased by approximately 30% from 2005 to 2017. However, our understanding of what leads Veterans to engage in suicidal behavior remains limited, as efforts have only recently begun attempting to distinguish Veterans with suicidal ideation from those who attempt suicide. Understanding this distinction is essential as up to 29% of individuals who experience SI eventually act on these thoughts (Nock, Borges, et al., 2008). Similar statistics do not exist for Veterans; however, findings suggest 15% of Veterans report SI history and 17 Veterans die by suicide daily (Office of Mental Health & Suicide Prevention, 2019). Differentiating between suicide ideators (individuals who experience suicidal ideation without attempting suicide) and attempters (individuals who attempt suicide at least once) may assist in providing more effective interventions to block this progression from ideation to attempt.

Emotion reactivity

The diathesis–stress model for adverse psychological outcomes suggests risk stems from the interaction between a diathesis (e.g., biomarker, trait-like disposition) and stressors (e.g., interpersonal loss, trauma). Researchers have applied this model to understanding suicidal behavior, proposing that some are born with biological/dispositional traits (diatheses) that place them at higher risk for engaging in suicidal behavior when encountering environmental stress (Brodsky, 2016; van Heeringen & Mann, 2014). Thus, risk for suicidal behavior is highest for individuals with high biological vulnerability and environmental stress and lowest for individuals with low biological vulnerabilities and environmental stress.

In accordance with this model, a commonly proposed diathesis for suicide is emotion reactivity, the sensitivity and intensity with which one experiences emotion (Nock, Wedig et al., 2008). It is distinct, but related to, emotion lability (instability of emotions) and regulation (increasing, maintaining, or decreasing emotions; (Neacsiu et al., 2018). Emotion reactivity is underpinned by limbic system irregularities, including amygdala hyperreactivity (Silbersweig et al., 2007), that may contribute to over-encoding of adverse experiences and increase likelihood of exaggerated emotional and behavioral responses (i.e., engagement in suicidal behavior; Rasch et al., 2009). Given their military experience, Veterans may be particularly susceptible to high emotion reactivity. Exposure to potentially traumatic events (e.g., combat) is commonly associated with greater emotion reactivity (Fitzgerald et al., 2018; Infurna et al., 2015). Evidence also suggests discouraging emotional expression—as is common in military culture (Arkin & Dobrofsky, 1978; Lorber & Garcia, 2010)—may increase emotion reactivity, particularly in individuals with multiple stressors (Davidov & Grusec, 2006; Fabes et al., 2002).

Emotion reactivity is closely associated with both suicidal ideation (SI; Liu et al., 2020; Shapero et al., 2019) and suicide attempts (SA; Nock, 2008b; Shapero et al., 2019). Consistent with conceptualizations of emotion reactivity as a diathesis for suicide, a psychophysiological measure of high emotion reactivity predicted a SA over a 12-month follow-up period among Veterans (Hazlett et al., 2016). Veterans with high emotion reactivity are more likely to experience SI in response to stressors, particularly when they also have poor problem-solving skills (Nezu et al., 2017) and similar patterns are observed in civilians (Dour et al., 2011).

Our understanding of what differentiates risk at the acute end of the spectrum (i.e., between suicide ideators without attempts and attempters) is more uncertain. Per the diathesis–stress model (Brodsky, 2016), Veterans who attempt suicide (versus those who experience ideation in the absence of an actual attempt) would experience greater psychopathology and high-risk behaviors reflective of heightened emotion reactivity (diatheses), exacerbated by life experiences associated with suicide (stressors). The present study is the first multimodal study (i.e., self-report, clinician, and psychophysiological assessment) to examine risk factors of suicide associated with emotion reactivity to determine which one or combination best distinguish Veterans with SI history only from those with SA history.

Risk factors differentiating suicide ideation from attempt

Consistent with the diathesis–stress model, emotion reactivity is closely associated with various clinically relevant outcomes related to suicide (e.g., exaggerated startle response, non-suicidal self-injury [NSSI], psychopathology) particularly when occurring in combination with potentially traumatic life experiences (e.g., childhood maltreatment, combat exposure).

Potential diatheses for suicide ideation and attempt

Psychophysiology

Preliminary evidence suggests the startle response may provide unique insight into suicide risk. The affect-modulated startle response (eyeblink reflex in response to a brief, startling stimulus; AMS) is a proxy for amygdala activity and physiological measure of emotion reactivity (e.g., Hazlett et al., 2016). To date, few studies have examined startle modulation to differentiate individuals with SI and SA. Among 28 depressed patients, those with SA history demonstrated greater startle in anticipation of predictable shock compared to those without SA history (Ballard et al., 2014). A similar pattern was observed by Smith et al. (2010) when employing unpleasant and suicide-related pictures, yet the sample was small and between-group differences did not reach significance.

Evidence suggests a similar pattern in Veterans (Hazlett et al., 2016). Findings indicate that Veterans with a history of multiple SA demonstrated greater startle potentiation when viewing unpleasant pictures and deficient habituation of AMS over time compared to the those with single SA or SI without SA. Animal and human work show startle potentiation during fearful/unpleasant stimuli is mediated by the brain’s fear/defense circuit, centered primarily on the amygdala (Davis & Whalen, 2001), underscoring the translational aspect of an AMS approach. A novel aspect of the present study is its aim to examine the role of AMS as a diathesis for suicide risk in combination with psychosocial suicide risk factors also associated with emotion reactivity.

Non-suicidal Self-Injury (NSSI)

Non-suicidal self-injury (NSSI), the direct, deliberate destruction of one’s body tissue without intent to die (Klonsky & Muehlenkamp, 2007) has been associated with later SA (Guan et al., 2012). NSSI is associated with emotion reactivity (Kleiman et al., 2014; Nock et al., 2008b), and may mediate the relationship between emotion reactivity and SI by relieving unwanted feelings or emotions (Brown et al., 2002; Liu et al., 2020). NSSI also distinguishes suicidal ideation from attempts among Servicemembers and Veterans (Kiekens et al., 2018; Klonsky et al., 2013). Growing research suggests Army soldiers with NSSI history are more likely to attempt suicide than solders without such history (Nock et al., 2018; Turner et al., 2019). Evidence also suggests NSSI may better predict later SA than SI among military personnel (Bryan et al., 2015). Cumulatively, such findings suggest NSSI, through offering short-term relief from emotion reactivity, confers risk for suicidal behavior after becoming insufficient in managing emotion reactivity.

Psychopathology

Emotion reactivity is also associated with the development and course of certain psychopathologies, for example, post-traumatic stress disorder (PTSD; Fitzgerald et al., 2018; Wang et al., 2018) and substance use disorders (SUDs; Kober, 2014). The role of emotion reactivity in major depressive disorder (MDD) is less consistent: MDD is associated with greater reactivity to positive events and comparable reactivity to negative events (Thompson et al., 2012) but also greater anger reactivity (Ellis et al., 2013). Psychopathology characterized by emotion reactivity may distinguish suicide ideators from attempters (May & Klonsky, 2016; Nock et al., 2013). Among Servicemembers, these include panic disorder, PTSD, intermittent explosive disorder, major depressive episodes, and SUDs (Nock et al., 2014, 2018). Consistent with the diathesis–stress model, evidence suggests the relationship between adverse experiences and suicide risk may be partially mediated by the increasing occurrence of mental disorders (Dube et al., 2001). However, diagnoses are typically viewed in isolation rather than in conjunction with environmental risk factors for suicide, limiting understanding of interactions between potential diatheses and stressors.

Potential stressors for suicide ideation and attempt

Potentially traumatic life experiences

Certain life experiences appear to interact with emotion reactivity to increase risk for suicidal behavior. Previous research suggests the relationship between emotion reactivity and suicidal thoughts and behavior (Shapero et al., 2019) is stronger in individuals with higher versus lower childhood trauma severity. Among new Servicemembers, frequent exposure to multiple forms of childhood maltreatment was associated with progression from SI to planning for suicide, and from planning to SA (Stein et al., 2018). Studies examining the link between service-related experiences and suicide risk, however, yield mixed results. Research consistently suggests deployment is not associated with elevated SA risk (Nock et al., 2014; Reger et al., 2015); however, SA history has been both positively and negatively associated with combat exposure intensity (Bryan, Griffith, et al., 2015). Additional research is necessary to understand the role of combat exposure as a stressor to differentiate Veterans with SI from those with SA history, particularly in combination with other suicide risk factors.

Present study

The present study is the first to concurrently and multimodally examine suicide risk factors associated with emotion reactivity—a psychophysiological measure of startle response (AMS), NSSI, clinician-assessed diagnoses (i.e., PTSD, SUD, anxiety disorders, depressive disorders), childhood trauma, and combat exposure—to determine which best discriminate Veterans with SA history from those with SI history in the absence of SA. This is a particularly important line of research among Veterans as active-duty Servicemembers often underreport suicide risk in fear of stigma and occupational repercussions of disclosure (Nock et al., 2018; Turner et al., 2019).

A prior publication examined the subgroup who completed the AMS paradigm (Hazlett et al., 2016); however, data from the larger sample have not been reported on. The current study uses different independent/grouping variables than used in previous and includes psychosocial variables associated with emotion reactivity separate from and in combination with the AMS variable. We hypothesized, compared to Veterans with history of SI only (SA−), Veterans with SA history (SA+) would endorse greater incidence of childhood trauma, combat exposure, NSSI history; be more likely to have a diagnosis of depressive disorders, anxiety disorders, PTSD, and SUD; and display exaggerated AMS in response to unpleasant stimuli. We expected all hypothesized diagnostic and psychosocial predictors would predict SA history.

METHOD

Participants

Data were obtained from a larger transdiagnostic sample of 310 U.S. military Veterans with varying levels of suicide risk. Participants were recruited from inpatient/outpatient psychiatry referrals at the James J. Peters VA Medical Center (JJPVAMC) between 2010 and 2015. Inclusion criteria included Veterans receiving mental health services between the ages of 18–57, and meeting “higher-risk” suicide status per the following criteria: SA or SI resulting in psychiatric hospitalization or emergency room presentation within the past 3 months; chronic SI (i.e., >3 months); and/or assignment to JJPVAMC’s “high-risk” suicide list. Exclusion criteria included meeting DSM-IV criteria for any schizophrenia-related psychotic disorder or current/history of organic brain impairment (e.g., stroke, CNS tumor, demyelinating disease, severe head trauma). Undomiciled participants and those engaging in substance use were eligible. Participants who denied SI history (n = 52) on interview were excluded from current analyses, yielding a sample size of 258. A subset of participants completed the AMS paradigm (n = 90). Sample demographics are reported in Table 1 (total) and 2 (AMS-only).

TABLE 1.

Demographic characteristics in suicide attempt groups

	SI (n = 81)		SA (n = 177)		Chi-Square	p
	N	%	N	%
Sex					4.72	.03
Female	16	19.8	58	33.0
Male	65	80.2	118	67.0
Marital Status					3.78	.15
Never Married	32	39.5	64	36.4
Married	22	27.2	33	18.8
Divorce/Sep/Widow	27	33.3	79	44.9
Highest Education					0.47	.79
<12 years	26	32.1	50	28.4
12–16 years	51	63.0	115	65.3
16+ years	4	4.9	11	6.3
Employed	17	21.0	33	18.8	0.18	.67
Race					3.67	.30
Caucasian	18	22.2	31	17.6
African American	38	46.9	70	39.8
Hispanic	18	22.2	59	33.5
Other	7	8.6	16	9.1
Military Branch					3.38	.34
Army	41	58.6	101	63.5
Air Force	4	5.7	17	10.7
Navy	16	22.9	24	15.1
Marines	9	12.9	17	10.7
Combat Exposure (Y/N)	30	37.5	61	35.9	0.06	.80
OEF/OIF	28	35.0	67	39.0	0.36	.55
Re-enlisted (Y/N)	8	32.0	23	37.7	0.25	.62
Deployments					0.78	.68
None	13	52.0	26	42.6
One	8	32.0	21	34.4
Two +	4	16.0	14	23.0
Self-Directed Violence (Y/N)
Preparatory Behavior	4	4.9	59	33.7	24.71	<.01
NSSI	9	11.3	59	34.1	14.54	<.01
	SI		SA
	M	SD	M	SD	F	p
Age	42.31	10.03	40.10	11.14	1.58	.12
Income	13,862	20,072	12,559	19,438	0.12	.73
Military Service (years)	5.21	5.15	5.26	4.06	0.01	.93

Some categories do not contain complete information for some participants due to missing data.

SI, Suicidal Ideation; SA, Suicide Attempt.

Significant associations (p < .05) are in bold

TABLE 2.

Demographic characteristics in suicide attempt groups, AMS subgroup only

	SI (n = 27)		SA (n = 63)		Chi-square	p
	N	%	N	%
Sex					0.35	.56
Female	3	11.1	10	15.9
Male	24	88.9	53	84.1
Marital Status					5.32	.07
Never Married	11	40.7	18	28.6
Married	8	29.6	10	15.9
Divorce/Sep/Widow	8	29.6	35	55.6
Highest Education					2.25	.33
<12 years	10	37.0	18	28.6
12–16 years	14	51.9	42	66.7
16+ years	3	11.1	3	4.8
Employed	3	11.1	12	19.0	0.86	.36
Race					3.39	.34
Caucasian	6	22.2	15	23.8
African American	15	55.6	23	36.5
Hispanic	5	18.5	22	34.9
Other	1	3.7	3	4.8
Military Branch					0.88	.83
Army	17	68.0	36	60.0
Air Force	1	4.0	5	8.3
Navy	5	20.0	12	20.0
Marines	2	8.0	7	11.7
Combat Exposure (Y/N)	7	25.9	22	36.1	0.87	.35
OEF/OIF	6	22.2	22	35.5	1.53	.22
Re-enlisted (Y/N)	8	32.0	23	37.7	0.25	.62
Deployments					0.78	.68
None	13	52.0	26	42.6
One	8	32.0	21	34.4
Two +	4	16.0	14	23.0
Self-Directed Violence (Y/N)
Preparatory Behavior	2	7.4	21	33.9	6.87	<.01
NSSI	2	7.4	23	37.1	8.21	<.01
	SI		SA
	M	SD	M	SD	F	p
Age	45.48	9.69	41.70	11.41	2.27	.14
Income	8,667	19,946	12,561	24,306	0.28	.60
Military Service (years)	5.20	5.89	5.33	3.68	0.02	.90

Some categories do not contain complete information for some participants due to missing data.

SI, suicidal ideation; SA, suicide attempt.

Significant associations (p < .05) are in bold

Current and lifetime psychopathology was assessed with the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I; First et al., 1996),¹ a semi-structured diagnostic interview conducted by a trained clinical psychologist. Weekly consensus and diagnostic meetings were led by a second clinical psychologist. Rates of diagnoses between groups are provided (Table 3). All participants provided written informed consent approved by JJPVAMC’s Institutional Review Board.

TABLE 3.

Description of Psychopathology in SI vs. SA

	SI		SA		Chi-square	Odds ratio	Confidence intervals
Diagnosisa	N	%	N	%			LLb	UL
Depression
Current	33	40.7	111	62.7	10.88**	2.45	1.43	4.19
Lifetime	44	54.3	115	65.0	2.67	1.56	0.91	2.66
Anxiety
Current	20	24.7	32	18.1	1.51	0.67	0.36	1.27
Lifetime	20	24.7	33	18.6	1.25	0.70	0.37	1.31
PTSD
Current	24	29.6	86	48.6	8.17**	2.25	1.28	3.93
Lifetime	31	38.3	93	52.5	4.53*	1.79	1.04	3.05
SUD
Current	44	54.3	114	64.4	2.38	1.52	0.89	2.60
Lifetime	62	76.5	137	77.4	0.02	1.05	0.56	1.96

Some categories do not contain complete information for some participants due to missing data. SI, n = 81; SA, n = 177. PTSD=post-traumatic stress disorder.

SUD, substance use disorder.

^aDiagnoses derived from SCID.

^bLL=lower limit, UL=upper limit.

^*p <.05,

^**p <.01.

Measures

Lifetime suicidal ideation and behavior history was assessed using the Columbia Suicide Severity Rating Scale (C-SSRS; Posner et al., 2011). The C-SSRS is a semi-structured interview measuring the presence/absence of prior SI, suicide behavior, and NSSI. Participants were included in the suicidal ideation (SI) group if they endorsed history of either passive (e.g., a “wish to be dead”) or active SI (e.g., “I’ve thought about killing myself”) but denied SA history. Those in the suicide attempt (SA) group endorsed lifetime history of 1+ actual, interrupted, or aborted SA. The C-SSRS has demonstrated good convergent and divergent validity with other suicidal ideation and behavior scales (Posner et al., 2011).

The Childhood Trauma Questionnaire (CTQ; Bernstein & Fink, 1998), a 28-item self-report questionnaire, assessed severity of maltreatment experiences from childhood and adolescence using a 5-point scale from “1 (Never True)” to “5 (Very Often True).” The CTQ produces a total score and five clinical subscale scores: Sexual, Physical, and Emotional Abuse; Physical and Emotional Neglect. It has demonstrated excellent internal consistency (Bernstein et al., 2003), including in the present sample (α=0.88). To preserve statistical power, only the total score was included in predictive models.

The Combat Exposure Scale (CES), a 7-item self-report questionnaire (Keane et al., 1989), was used to assess exposure to various combat situations (e.g., firing rounds at the enemy, having very dangerous duty) during deployment(s) using a 5-point Likert scale. Total score is calculated as a sum of weighted scores, classified according to five categories of combat exposure ranging from “light” to “heavy.” It demonstrated excellent internal consistency in the present sample (α=0.90).

Affect-modulated startle paradigm

A subset of demographically matched participants (n = 90) completed an AMS paradigm, lasting approximately 32 minutes (Table 2). Participants were read standard instructions, and electrodes were attached for psychophysiological recording. Before task initiation, two examples were presented to familiarize participants with stimuli and ensure adequate elicitation of the startle-eyeblink reflex. Participants viewed a series of pictures on the screen and were instructed to look at each picture the entire time it appeared and think about its personal meaning until each picture’s offset. They sometimes heard a brief burst of static noise through headphones during and between viewing of pictures.

The 84 pictures (28 unpleasant, e.g., a man hitting a woman; 28 neutral, e.g., a neutral face; 28 pleasant, e.g., people embracing) were selected from the International Affective Picture Show (IAPS; Lang et al., 2008), primarily focusing on social contexts. Pictures were presented in a mixed, semirandom order for 6 seconds each on a 22-inch computer monitor, ~36 inches from the participant’s face. Pictures were separated by an inter-trial interval (ITI; 12–18 s, mean ITI=15 s). Of the 28 pictures in each picture condition, 20 were accompanied by a startle-eliciting stimulus/probe—105 dB, 50 msec white noise burst with a near-instantaneous (< 1 msec) rise/fall. Of the 20 pictures in each picture condition that included a startle stimulus, half were probed 4000 msec after picture onset and half at 5000 msec. Twenty additional startle probes were presented during the ITI (10 early, 10 late block). Half of startle probes were presented during the first half of the experiment; the remaining in the second half. Picture valence and arousal levels were matched across early and late trials. Further details of the paradigm can be found elsewhere (Hazlett et al., 2016).

Digitized startle data were scored offline in microvolt units using standard in-house methods. Our program scores responses occurring within a latency window of 21–120 msec and reaching a peak within 150 msec of response onset. This scoring system has high inter-rater reliability (ICC range=0.80–0.99, median=0.95). To minimize between-subject variability, individual raw data were standardized into t-scores using blinks of each participant, per Psychophysiology committee guidelines (Blumenthal et al., 2005). Present analyses included AMS in the unpleasant picture condition, a proxy for amygdala activity and emotion reactivity.

Statistical methods

Experimental group membership was determined using participant C-SSRS responses. SA− Veterans (n = 81) were compared to SA+Veterans (n = 177) using analysis of variance and chi-square analyses. Multiple imputation (MI) was used to account for missing data contributing to self-report sum scores (CES, n = 3; CTQ, n = 5). MI is considered the preferred method for addressing missing data in large datasets as it is less biased than traditional complete case analysis (Sterne et al., 2009) and provides a wider range of possible outcomes (Sinharay et al., 2001). A false detection rate (FDR) of 5% was used to control for multiple comparisons.

Hypothesized predictors (CTQ, CES, NSSI history, current depressive disorder, anxiety disorder, SUD, and PTSD) were selected based on their association with emotion reactivity and entered into a binary logistic regression model to identify unique predictors of SA-SI status. Demographic variables that significantly differentiated groups were included as covariates to adjust for potential demographic differences.

A separate predictive model was tested with inclusion of AMS with aforementioned variables, run without MI due to limited power to detect effects within this smaller sample. Performance of the resulting composite risk index with sensitivity and specificity for each model was evaluated using the area under the receiver operating characteristics curve (AUC).

RESULTS

Groups were comparable demographically (Table 1); however, male Veterans were less likely to endorse SA history (OR=0.51). SA+Veterans were more likely to endorse history of NSSI (OR=4.08) and preparatory behavior for suicide (OR=9.79). Among SA+Veterans, 83.6% reported history of actual suicide attempt, 44.6% aborted attempt, and 27.1% interrupted attempt. Average SA number was 3.32 (SD=7.63), 39.8% of which resulted in at least moderate physical injury. Among those completing the AMS paradigm, groups were also comparable demographically (Table 2); however, SA+Veterans were also more likely to endorse history of NSSI (OR=8.21) and preparatory behavior for suicide (OR=6.40).

All but one Veteran met criteria for an Axis I diagnosis. SA status was associated with current diagnosis of MDD (OR=2.45) and PTSD (OR=2.25; Table 3). Additionally, SA+Veterans reported more severe childhood maltreatment experiences (Table 4). SA-SI distinction was not significantly associated with degree of combat exposure, physical abuse history, or presence of anxiety disorder or SUD.

TABLE 4.

Clinical traits and startle magnitude in suicidal ideation (SI) vs. suicide attempt (SA) groups

	SI		SA		F	p	d
	M	SD	M	SD
Combat Exposure	6.97	9.46	7.69	9.80	0.29	.59	0.07
Childhood Trauma	61.72	17.91	69.84	19.88	9.52	<.01	0.43
AMS (Negative)	51.20	2.06	53.38	3.27	10.20	.00	0.80

Overall sample: SI, n = 81; SA, n = 177. AMS paradigm: SI, n = 27; SA, n = 63. AMS (Negative) = affect-modulated startle magnitude in response to negative stimuli. Childhood Trauma, as measured by the Childhood Trauma Questionnaire (CTQ). Overall scores are the sum of all 28 items, which are each rated on a scale of “1” to “5” to yield a sum score between 28 to 140. Higher scores are indicative of greater severity of different types of childhood trauma (e.g., abuse and/or neglect). Combat Exposure, as measured by the Combat Exposure Scale (CES). Overall scores are the sum of all 7 items, which are each rated on a scale of “1” to “5” to yield a sum score between 7 and 35. Higher scores are indicative of greater degree of exposure to various combat situations during deployment. Cohen's d compares a mean difference in scores between SI and SA groups. d = 0.20–0.49 is indicative of a small effect size, d = 0.50–0.80 is a medium effect size, and d > 0.80 is a large effect size.

Significant associations (p < .05) are in bold

Pearson correlations among predictors are included in Table 5. Predictors (along with sex, given its significant association with SA status) were included in binary logistic regression models to predict SA history (Table 6). Model accuracy was calculated for complete case analysis and pooled results of the five generated imputation models. Contrary to hypotheses, anxiety disorder diagnosis and combat exposure did not significantly predict group membership. Furthermore, while SA-SI distinction was associated with childhood trauma, it also was not a significant predictor. Significant independent predictors in the final pooled MI model were sex, NSSI history, current depressive disorder, PTSD, and SUD diagnoses such that SA+Veterans were more likely to be female, endorse NSSI history, and meet criteria for a depressive disorder, PTSD, and SUD compared to SA− Veterans. Results were comparable using complete case analysis. The receiver operating characteristic curve (ROC) for the complete case analysis risk index containing predictors of SA history achieved good accuracy in distinguishing SA+from SA− Veterans (AUC=0.71; sensitivity=.90, specificity=.49; Figure 1) and was comparable to the average risk model of all imputation models (AUC=0.71, sensitivity=0.90, specificity=0.48).

TABLE 5.

Correlations among predictors

	Sex	NSSI	Dep	Anx	PTSD	SUD	CTQ	CES	AMS-N
Sex
NSSI	0.08
Dep	−0.06	0.03
Anx	0.07	0.03	0.06
PTSD	0.08	0.05	0.10	0.04
SUD	−0.26**	−0.12	−0.02	−0.16*	0.03
CTQ	0.20**	0.07	0.13*	−0.06	0.10	0.06
CES	−0.16*	−0.05	0.06	−0.05	0.42**	0.04	−0.02
AMS-N	0.20*	0.25**	0.19*	−0.02	0.05	0.02	0.19*	0.02

Sex coded “1” for female. NSSI=non-suicidal self-injury (“yes” =1).

AMS-N=affect-modulated startle in response to negative stimuli. Current diagnoses for psychopathology were derived using the SCID.

Anx, any anxiety disorder; CES, Combat Exposure Scale sum score; CTQ, Childhood Trauma Questionnaire sum score; Dep, any depressive disorder; PTSD, post-traumatic stress disorder; SUD, substance use disorder.

^*p < .05,

^**p < .01.

TABLE 6.

Regression models predicting suicide attempt history

Predictor	β	SE	p	OR	95% confidence interval
					LLa	UL
Complete case analysis
Sex	−0.86	0.39	.03	0.42	0.20	0.91
NSSI	1.40	.42	<.01	4.04	1.79	9.14
Depression	0.89	.31	<.01	2.42	1.32	4.46
Anxiety	−0.47	.38	.22	0.62	0.29	1.32
PTSD	0.76	.38	.04	2.13	1.02	4.46
SUD	0.73	.33	.03	2.08	1.08	3.98
Combat Exposure	−0.01	.02	.69	0.99	0.96	1.03
Childhood Trauma	0.01	.01	.16	1.01	1.00	1.03
MI (Pooled)
Sex	−0.86	.39	.03	0.42	0.20	0.91
NSSI	1.38	.42	<.01	3.98	1.76	8.97
Depression	0.89	.31	<.01	2.44	1.33	4.47
Anxiety	−0.43	.38	.26	0.65	0.31	1.37
PTSD	0.75	.37	<.05	2.11	1.02	4.39
SUD	0.75	.33	.02	2.11	1.11	4.01
Combat Exposure	0.01	.02	.70	0.99	0.96	1.03
Childhood Trauma	0.01	.01	.15	1.01	1.00	1.03
Psychophysiology subgroup
Sex	−0.50	.96	.60	0.61	0.09	3.96
NSSI	1.54	0.86	.07	4.67	0.87	25.14
Depression	1.03	.59	.08	2.81	0.88	9.00
Anxiety	0.78	.76	.30	2.19	0.50	9.64
PTSD	0.40	.72	.58	1.49	0.37	6.05
SUD	1.13	.69	.10	3.11	0.81	11.95
Combat Exposure	−0.01	.04	.85	0.99	0.92	1.07
Childhood Trauma	0.01	.02	.64	1.01	0.98	1.04
AMS (Negative)	0.32	.13	.01	1.38	1.08	1.76

Combat exposure was measured using the Combat Exposure Scale (CES) sum score. Childhood trauma was measured using the Childhood Trauma Questionnaire (CTQ) sum score. Current diagnoses for psychopathology were derived using the SCID. “Complete Case Analysis” model (n = 242) represents results without the use of MI to correct for missing data. “MI (pooled)” model (n = 258) represents pooled results of 5 models calculated using MI. “Psychophysiology Subgroup” model (n = 85) represents results without the use of MI, including only the subset of Veterans who completed the AMS paradigm.

AMS, affect-modulated startle in response to negative stimuli; MI, multiple imputation; NSSI, non-suicidal self-injury; PTSD, post-traumatic stress disorder; SUD, substance use disorder.

^aLL=lower limit, UL=upper limit.

Significant associations (p < .05) are in bold

FIGURE 1.

ROC curve for final model predicting suicide attempt status using complete case analysis without multiple imputation (AUC=0.71; sensitivity=0.90, specificity=0.49, n = 242). Results are comparable to that of the average risk model of all imputation models (AUC=0.71, sensitivity=0.90, specificity=0.48, n = 258).

Affect-modulated startle during the presentation of unpleasant pictures (AMS-N) was added to this model to explore whether it uniquely contributed to prediction of SA history among those who completed the paradigm (Table 6). Greater AMS-N amplitude was the only significant independent predictor. The risk profile achieved very good accuracy in distinguishing SA+from SA− Veterans (AUC=0.82; sensitivity=0.85, specificity=0.56; Figure 2).

FIGURE 2.

ROC curve for final model predicting suicide attempt status using complete case analysis data without multiple imputation, including affect-modulated startle (AMS) in response to negative stimuli (AUC=0.82; sensitivity=0.85, specificity=0.56; n = 85).

DISCUSSION

This study is the first to concurrently examine a psychophysiological measure of emotion reactivity and psychosocial risk factors for suicide among Veterans. We examined whether AMS-N, NSSI history, psychopathology, combat exposure, and childhood trauma differentiated Veterans with SA history from those with SI history in the absence of SA. Consistent with prior research, this SA-SI distinction was uniquely predicted by sex, NSSI history, PTSD, SUD, and MDD diagnosis; the resulting risk index demonstrated good ability to discriminate SA+from SA− Veterans. When a psychophysiological measure of emotion reactivity—startle amplitude in response to unpleasant pictures—was added to the model, accuracy to distinguish groups improved, yet psychosocial predictors were no longer significant. Findings support greater emotion reactivity as a likely diathesis for suicidal behavior. Consistent with previous research (Shapero et al., 2019), SA+Veterans may possess a biological vulnerability to emotionally respond more intensely to stressors, increasing risk for suicidal behavior as a means of escape. Suicide prevention efforts should promote treatment strategies aimed at reducing and responding to emotion reactivity and encouraging emotion regulation among Veterans with these diatheses. If replicated, findings may improve understanding of suicide etiology and inform prevention approaches.

Notably, however, exposure to potentially traumatic life experiences (e.g., combat exposure and childhood trauma) did not contribute to the SI-SA distinction. Per the diathesis–stress model, environmental stresses are most impactful when they occur within the context of high vulnerability stemming from diatheses (e.g., experienced by a Veteran with a history of NSSI and/or current psychopathology; Brodsky, 2016). Thus, because the current sample of Veterans had varying levels of vulnerability, effects of environmental stresses were likely muted, limiting ability to detect these effects statistically. Future research should examine potential moderating effects of NSSI history and psychopathology on the relation between environmental stressors and the SI-SA distinction.

Clinical implications

Findings highlight the salient role of AMS in distinguishing Veterans with SI only from those with SA. AMS is not reliant on self-report and may be particularly useful in identifying Veterans at greatest risk, who may have concerns about stigma associated with disclosure. While impractical for Veterans with SI to complete an AMS paradigm during suicide screening, results inform understanding of the role of emotion reactivity in the presence of suicidal behavior and further research is necessary to identify utility of alternative assessment methods.

Findings may support suicide theories that propose transition from suicide ideation to action results from one’s capability to make a suicide attempt (Joiner, 2005; Klonsky & May, 2015). This is a consequence of habituation to pain, fear, and death (i.e., exposure to painful, provocative events), dispositional factors (i.e., genetics, trait impulsivity), and practical factors (i.e., access to/familiarity with lethal means). Emotion reactivity may confer risk for suicidal behavior by indirectly increasing suicide capability, via greater engagement in painful behavior (e.g., NSSI, substance use) to escape/alleviate emotion reactivity (Law et al., 2015). NSSI and substance use may also increase familiarity with and access to potentially lethal instruments used for suicide (e.g., means for cutting, dangerous substances; Kaiser et al., 2012). Findings should be replicated and expanded upon to better understand the role of emotion reactivity in etiology of Veteran suicide risk and potential target for prevention and risk management.

Limitations and future directions

The current study has several strengths: large, transdiagnostic sample; interview assessment of psychopathology and NSSI; inclusion of a biomarker of suicide risk; and broad coverage of risk factors (clinical traits, adverse life experiences, and psychopathology). Whereas previous research has largely utilized samples with overall low risk for suicide, the present sample included Veterans identified as at high risk for suicide. As is often the case, these individuals had a high prevalence of mental health diagnoses. The present study also expands upon literature that highlights differences in Servicemembers with suicidal ideation and behavior (Nock et al., 2018) by utilizing a sample of Veterans from several military branches.

However, there are notable limitations, including utilization of cross-sectional data and reliance on retrospective report of suicide history absent of a timeline of psychopathology onset and SA history. Consequently, application of results is limited in understanding risk for future suicidal behavior as it was not prospectively predicted. Emotion reactivity may be a diathesis present prior to suicidal behavior or result from associated adverse life events, psychopathology, NSSI, or suicidal behavior history. Only Veteran SA survivors were included, limiting application of results to suicide risk broadly. Veterans with greater emotion reactivity may be more likely to attempt suicide and survive, but unlikely to make a lethal SA, whereas those with less emotion reactivity may be more likely to die by suicide. Although beyond the present study’s scope, future longitudinal studies can aid in exploring this possibility. Potential moderating variables reflecting affective vulnerability (e.g., emotion regulation) were not assessed but likely also play an important role in development of suicide risk. Future longitudinal studies should examine these variables as early in military service as possible—ideally preceding basic training—to understand their role in suicide etiology. Furthermore, as is common with psychophysiological data, only a subgroup completed the startle paradigm limiting statistical power to detect small yet potentially meaningful effects. Future studies should continue to explore distinct profiles for suicide risk utilizing multimodal approaches, including biological measures, among large Veteran samples.

Although the presence of psychiatric diagnoses was assessed using a structured diagnostic interview (i.e., SCID), a single-item interview assessed NSSI. Future studies should incorporate interview questions assessing qualities of NSSI history (e.g., recency, frequency, method, function) in differentiating across the SI-SA spectrum. Given overlap between predictors identified by the present study and experiences of individuals with BPD, other variables that may be of interest to future researchers include dissociation and impulsivity.

CONCLUSION

Consistent with research in civilian samples, various variables representative of emotion reactivity differentiated SA+ Veterans from SA− Veterans: sex, NSSI history, MDD, PTSD, and SUD diagnosis. Only startle reflex to negative stimuli remained significant when included in the risk model. These are areas deserving of continued research to better understand etiology of Veteran suicide risk.