Intervention related reductions in perceived burdensomeness mediates incidence of suicidal thoughts

Abstract

Background:

Interventions aimed at preventing suicidal thoughts target people at risk for suicide based on risk factor elevations. Based on the interpersonal psychological theory of suicide, elevated perceived burdensomeness (PB) and thwarted belongingness (TB) are potential targets for prevention of the occurrence of suicidal thoughts. PB is the belief that one is a burden to others. TB is the perceived lack of social connectedness.

Methods:

This study was designed to examine the effects of a preventative intervention targeting PB and TB on the 6-month incidence of suicide ideation in a sample of 138 people (M =38.01 years, SD =16.40; 50.7% female) with elevated levels of these variables but no current suicidal thoughts at baseline. The three-session intervention included psychoeducation and cognitive bias modification.

Results:

PB was reduced in the intervention condition, compared to the repeated contact control condition (B =2.50, p < .05) and TB was not (B =1.42, p=.43). The likelihood of a reported incident of suicidal thoughts was reduced for those in the active intervention compared to those in the repeated contact control condition, through reductions in PB (B =.12, 95% confidence interval [.01, .32]).

Limitations:

There were two components of the intervention, cognitive bias modification and psychoeducation; thus, it is unclear whether one or both components contributed to these findings.

Conclusions:

This intervention can be used as a preventative intervention for suicidal thoughts by targeting PB. These results further confirm PB as a causal risk factor for suicidal thoughts.

1. Introduction

Suicide deaths have increased during the last two decades (Prevention, 2017), with recent estimates suggesting suicide accounts for upwards of 44,000 deaths in the United States annually (Curtin et al., 2016). Considering this alarming trend, suicide has been declared a public health crisis (US Department of Health and Human Services Office of the Surgeon General and National Action Alliance for Suicide Prevention, 2012), spurring population-level, general programs to promote suicide prevention outreach via public education, screening, hotline support, and means safety counseling; however, there is still need for evidence-based randomized controlled trials (RCTs) that target prevention in specific at-risk populations.

Preventive interventions for suicide can be universal, selective, or indicated. Universal interventions are population-level interventions that do not directly target those at risk. For example, public-service messages on the radio, television, or on billboards would be considered universal interventions. Indicated interventions target people labeled as high-risk who display minimal but detectable signs or symptoms, such as people currently endorsing suicidal thoughts or behaviors. Finally, selective interventions target people with elevations on warning signs or risk factors, and selectively examine the impact of interventions on these risk factors (Muñoz et al., 1996).

Selective interventions aimed at preventing suicidal thoughts are likely to have the most success if they are based on empirically informed theories of suicide. Thus, it would be ideal to focus on risk factors through the lens of well-validated theoretical models of suicide. The Interpersonal Psychological Theory of Suicide (IPTS) is a contemporary and influential empirical model of the development of suicidal thoughts and subsequent progression to suicidal behavior (Joiner, 2005; Van Orden et al., 2010) that has substantial empirical support (e.g., Bryan et al., 2010; Christensen et al., 2014; Cukrowicz et al., 2013; for review, see Ma et al., 2016). The IPTS posits that a desire for death develops when a person experiences: 1) the perception of being a burden to family, friends, and others (i.e., “perceived burdensomeness [PB]”); and 2) insufficient social connectedness through loss of interpersonal contact (i.e., “thwarted belongingness [TB]”). To progress from a desire for death to suicidal behavior, an additional component, the acquired capability for suicide, or the ability to overcome selfpreservation instincts, is also necessary (Joiner, 2005; Van Orden et al., 2010). Whereas acquired capability is perceived as an emergent construct, developing through habituation to life-threatening or painful experiences, PB and TB are posited as dynamic cognitive-affective states that are reciprocally related to social and environmental factors (Joiner, 2005; Van Orden et al., 2010).

The evidence for PB and TB as risk factors for the development of suicidal thoughts is mixed. A recent meta-analysis reported that PB alone was consistently associated with suicidal thoughts but that TB was less consistently associated with suicidal thoughts. Of the studies that examined the synergistic effect of PB and TB, 8 of 12 found a significant interactive effect of PB and TB in relation to suicidal thoughts. However, the majority of these studies were cross-sectional in nature. (Ma et al., 2016).

Delineating the role of PB and TB as causal risk factors for suicidal thought is of theoretical and clinical import for accurately understanding, and therefore effectively targeting, suicide risk through preventative intervention. A systematic approach to conceptualizing ‘risk factors’ (Kraemer et al., 1997) for suicide necessitates that we: 1) identify the purported mechanisms involved in the progression of suicidality, 2) demonstrate that these mechanisms can be intentionally changed (i.e., are malleable), and 3) examine whether intervening on these mechanisms impacts suicide-related outcomes. To this first point, PB is a known risk marker (i.e., correlate) of suicidal thoughts whereas support for TB is equivocal (Ma et al., 2016; Van Orden et al., 2010). However, it is unclear whether PB and TB are truly causal ‘risk factors’ for suicidal thoughts. One approach, that also addresses the paucity of selective interventions within the realm of suicidal thought prevention, is to examine whether RCTs can effectively reduce PB and TB and whether reductions in these risk markers prevent the emergence of suicidal thoughts.

To date, only two studies have focused on intentionally changing PB and TB. Van Orden et al. (2016) compared a peer companion condition to care-as-usual in a randomized clinical trial (RCT) among older adults at risk for suicide, with initial data suggesting that PB, but not TB, was significantly decreased among the peer companion condition (Van Orden et al., 2016). Another web-based RCT, in an adolescent sample, demonstrated that a preventative intervention targeting PB (LEAP) led to significantly greater reductions in PB, but not TB, among treatment completers compared to a psychoeducation control; though, these reductions were non-significant within intent-to-treat analyses (Hill and Pettit, 2016). Further, there were no significant differences in suicidal ideation between the treatment and control conditions at follow-up. Together, these studies provide initial evidence that PB is malleable through targeted intervention, albeit in specific populations for which findings may not generalize (i.e., older adults, adolescents). Evidence that TB can be reduced or that reductions in PB or TB subsequently mitigates risk of suicidal thoughts is still needed.

In sum, suicide is a public health crisis with extreme social and economic cost (Control & Prevention, 2012), and we presently lack a definitive approach to effective suicide prevention (Zalsman et al., 2016). An empirical approach to preventing suicide should therefore seek to identify and target putative mechanisms grounded in scientific evidence, in a systematic way. Within the framework of the IPTS, attempts to target PB and TB have provided generally positive results (Hill and Pettit, 2016; Van Orden et al., 2016); however, findings regarding reduction in TB are mixed, and positive effects of a PB/TB-specific intervention on suicide-related outcomes have, to our knowledge, not yet been reported.

The current study therefore sought to test whether brief, computerized interventions designed in part to directly target elevated PB and TB could reduce these mechanisms, and prevent the incidence of future suicidal thoughts. It was hypothesized that the active interventions would lead to greater reductions in PB and TB than a repeated-contact control. Given that these interventions were designed to target these risk markers, and not suicide-related outcomes specifically, we did not expect any direct effect of condition on suicidal thought. However, based on the results of prior trials using similar interventions (Schmidt et al., 2014, 2017a, 2017b), we hypothesized that condition-specific changes in PB and TB would reduce the likelihood of subsequent suicidal thoughts during a follow-up period with assessments at one, three, and six months post-intervention.

2. Methods

2.1. Participants

The sample consisted of 138 people selected from a larger randomized clinical trial (RCT) funded by the Department of Defense (DoD) examining the efficacy of two computerized treatments targeting risk factors relevant to suicide (Schmidt et al., 2017). Inclusionary criteria for the larger RCT included scoring at or above the community mean on PB/TB or anxiety sensitivity (AS) cognitive concerns (i.e., the fear that cognitive symptoms of anxiety might be catastrophic; e.g., Taylor et al., 2007) and English speaking to ensure comprehension of the intervention. Exclusionary criteria included: evidence of uncontrolled bipolar or psychotic spectrum disorders, serious suicidal intent that would warrant immediate hospitalization, unstable medication usage (i.e., not on a stable dose for at least three months prior to study entry), and/or participation in current psychotherapy. Only people ages 18 years and older were eligible for participation. Because we were interested in examining the impact of the current intervention as a selective intervention for prevention of the incidence of suicidal thoughts over time, participants were selected if they endorsed no suicidal ideation at baseline, as defined by scoring a 0 on the DSI-SS.

Participants’ ages ranged from 18 to 79 (M =38.01 years, SD =16.40) and gender was evenly distributed (50.7% female). The majority of the sample identified as White (51.4%) followed by Black (36.2%), other (e.g., biracial; 8.1%), Asian (3.6%), and American Indian (.7%). Regarding ethnicity, the majority of the sample identified as non-Hispanic (94.2%). A little over half of the sample completed some college (54.3%) with 20.3% completing a bachelor’s degree, 13.8% completing high school or the equivalent, 10.1% completing graduate school, and 1.5% obtaining less than a high school education. In terms of diagnoses, 36.1% met for a primary anxiety disorder followed by 26.1% with a primary depressive disorder, 14.5% with a primary trauma-and stressor-related disorder, 4.3% with a primary obsessive-compulsive and related disorder, 4.3% with a primary substance-related and addictive disorder, .7% with a primary feeding and eating disorder, and 14% with no primary diagnosis. Lastly, 36.2% of the sample identified as a veteran having served in the US military.

2.2. Procedures

All people were recruited from the community via newspaper advertisements, website listings, and community mail outs to local medical and mental health care providers. Additionally, as this was a DoD funded project we conducted targeted recruitment for military veterans and/or people with a trauma history in order to increase military relevance. This included contacting local veteran organizations such as the Vet Center and Veterans Healthcare Administration. Interested participants phoned an outpatient clinic to complete a brief screening instrument. People deemed eligible based on the initial telephone screening were invited to complete a more comprehensive intake where they completed a thorough suicide risk assessment (Joiner et al., 1999), a brief battery of self-report questionnaires, and were interviewed using the Structured Clinical Interview for the DSM-5 (SCID; First et al., 2015). All qualified people were then scheduled for a baseline appointment and randomized, using an online random number generator, to one of four possible study conditions (see description of conditions below). Only people randomized to receive the active mood intervention (i.e., PB or TB), the combined intervention (i.e., PB/TB and AS cognitive concerns), or a repeated contact control intervention, with no active suicidal thoughts (based on a DSI-SS score of 0 at baseline) were included in the present investigation. All ineligible people were thanked for their time and given relevant community referrals based on their needs. Study procedures were approved the by university’s institutional review board.

2.3. Timepoints

2.3.1. Baseline appointment

Participants first read and signed an informed consent. Next, they completed a battery of self-report questionnaires and were awarded any monetary compensation they earned.

2.3.2. Intervention appointments (3 sessions)

Participants received the intervention at a rate of one session per week for three weeks. During sessions 1, 2, and 3, participants completed their assigned intervention followed by the intervention assessment measures. In the active treatment conditions, session 1 included the relevant psychoeducation and CBM interventions. Sessions 2 and 3 included only CBM.

2.3.3. Follow-up appointments (month 1, month 3, and month 6)

Participants first completed a brief battery of self-report questionnaires and were then scheduled for their next follow-up appointment and awarded any monetary compensation they earned. At the six-month follow-up, participants were also debriefed and given the opportunity to receive the combined intervention protocol, that was then scheduled at their convenience if they decided to participate.

2.4. Description of intervention conditions

2.4.1. Anxiety intervention condition

Participants who received the anxiety intervention completed the Cognitive Anxiety Sensitivity Treatment (CAST; Schmidt et al., 2014) and an AS-focused CBM program (CBM-I for AS; Capron and Schmidt, 2016). CAST is a fully computerized, 45-min intervention designed to closely model the educational and behavioral techniques that are commonly used in anxiety treatments. CAST contains 50 slides and was programed using Articulate Presenter. CAST contains video animation and audio narration throughout, as well as some interactive features (e.g., brief quizzes to promote comprehension). The psychoeducation portion of CAST focuses on providing corrective information about the nature of stress and its effects on the body. The program is designed to dispel myths regarding the immediate dangers of stress on the body. The program focuses on the supposition that anxiety symptoms are not dangerous, and that participants may have developed a conditioned fear to these sensations. After the psychoeducation component, IE exercises were introduced as a way to correct the conditioned fear response to psychological sensations. With the programs direction and assistance, participants completed a brief hyperventilation exercise and rated the level of fear/distress experienced on a ten-point scale. Participants then completed ten 60-second hyperventilation trials. After each trial, participants rated the intensity of the sensations experienced as well as their subjective distress, which were graphed for them by the program to demonstrate how they changed over the course of the trials. For more information about CAST please see Schmidt et al. (2014).

2.4.2. Cognitive bias modification (CBM) – interpretation bias (IB): anxiety intervention condition

CBM-IB was programmed using E-Prime software (Schneider et al., 2002) and was based on previous work on CBM for anxiety (Beard and Amir, 2008; Brosan, Hoppitt, Shelfer, Sillence, and Mackintosh, 2011). In the program, on each trial, participants were presented with a word for 1 s (e.g., “excited”) followed by a sentence (e.g., “You notice your heart is beating faster”). On half the trials, the combination of the word and sentence created a benign meaning (as in the previous example). On the other half of trials, this combination created an anxious meaning (e.g., “stressful” followed by the sentence – “Your mind is full of thoughts”). Participants were required to judge the relatedness of the word and the sentence, pressing “yes” if they think the word and sentence were related and “no” if they think they were unrelated. Participants were given feedback during training such that judging the anxious combinations as “unrelated” and the benign combinations as “related” would produce a “correct” response. Furthermore, if they judged anxious combinations as being “related” and benign combinations as being “unrelated” they would hear a horn blast (approximately 85 dB) and be given feedback that the response was “incorrect”. IB is measured by the number of trials in which participants endorsed benign pairings and rejected anxious pairings. Participants began with 40 test trials with no reinforcement to measure initial IB. They then completed 80 training trials in which their response was reinforced. Afterward, participants took a 5-min break followed by another 80 trials of training. Finally, they were given 40 test trials of word/sentence pairs that they have never seen before to measure change in IB. This procedure is based on previous IB literature (Brosan et al., 2011).

2.4.3. Mood intervention condition

The mood condition paralleled the anxiety condition in that it included a top-down psychoeducational portion as well as a bottom-up CBM portion. Participants who received the mood intervention completed a fully computerized, 50-min intervention designed to closely model the educational and behavioral techniques that are commonly used in depression treatments. Similar to CAST, the mood protocol contained video animation and narration throughout, as well as some interactive features (e.g., brief quizzes to promote comprehension). The mood intervention was designed in consultation with clinicians experienced in treating suicidal individuals dealing with issues related to PB/TB. The psychoeducation portion used CBT principles to correct problematic ideas and behaviors related to PB/TB. More specifically, the program was designed to dispel “myths” regarding PB/TB (e.g., “If you are around other people, you shouldn’t feel lonely”, “Talking to others about your problems makes you a burden”). The program emphasizes the idea that social interaction is a critical need, just like the need for food and water. Participants were taught that negative beliefs about being isolated and being a burden are usually inaccurate. Following this, behavioral activation techniques were introduced as a way to decrease isolation and feelings of burdensomeness. Specifically, participants were taught that the best way to combat these negative feelings is to take opposite action, such as reaching out, sharing your feelings with someone you trust, volunteering, and so forth.

2.4.4. Cognitive bias modification (CBM): mood intervention condition

Utilizing the positive CBM-I paradigm developed by Holmes, Mathews, Dalgleish, and Mackintosh (2006), 100 scenarios were presented across five training blocks containing 20 scenarios each. The descriptions, presented in E-Prime, were read aloud (via headphones) using the same female voice (lasting 10–13 s) with 2-second gaps in between each description. While initially ambiguous regarding outcomes, all scenarios were designed to be resolved positively. An example is as follows: “At your computer lesson you finish your work early so the lecturer gives you a new tasks to do. You don’t understand the task so you ask for help. The lecturer tells you that your request is the sign of being a good student.” Participants were instructed to imagine the events while listening to them. In order to further focus participants, after each paragraph they were asked to rate the vividness of imagery (“How vividly could you imagine the situation that was described?”) using a 5-point Likert-like scale ranging from1 (not at all vivid) to 5 (extremely vivid). All scenarios had more than one possible outcome with the overarching aim of training participants to generate positive outcomes to situations that could have developed in less desired ways. Participants began by completing an emotionality rating ranging from 1 (extremely unpleasant) to 9 (pleasant) using ambiguous test paragraphs (“You have recently taken an important exam. Your results arrive with an unexpected letter of explanation about your grade”). Following this, the participants completed the training phase (described above) followed by a five minute break and then a testing phase in which they once again rated the emotional valence of ambiguous test descriptions (e.g., “You buy a new outfit for a party. Other people’s reactions show how you look”). For more information regarding the CBM-I used see Holmes et al. (2006).

2.4.5. Combined intervention condition

Participants assigned to the combined condition completed both the anxiety and mood conditions. As such, the combined intervention was not matched for length. However, the order of the mood and CAST interventions, as well as their respective CBM’s, were counterbalanced across participants at each session.

2.4.6. Repeated contact control condition

There is relatively little empirical work on suicide prevention. Of the few trials that have shown positive effects, data suggests that a relatively simple intervention, such as repeated contact, may be effective in reducing suicide (Fleischmann et al., 2008; Motto and Bostrom, 2001). As such, participants in the repeated contact control condition were assigned a personal study coordinator at their baseline appointment. Participants were informed that the study coordinator would be contacting them once per week during the next three weeks (corresponding to the treatment session intervals for those in the active treatment conditions) for a brief phone check in which suicide risk would be evaluated.

3. Assessments

3.1. Diagnostic interview

3.1.1. Structured clinical interview for DSM-5, research version (SCID-5RV)

The SCID-5-RV is a semi-structured clinical interview used to assess for the presence of DSM-5 psychiatric diagnoses (First et al., 2015). The SCID-5-RV was administered by doctoral student therapists who underwent a systematic training procedure including reviewing SCID training tapes, observing live SCIDs, and conducting practice interviews. Therapists only began conducting diagnostic interviews once they demonstrated high levels of reliability. A licensed clinical psychologist reviewed all diagnostic decisions to ensure high levels of diagnostic accuracy. Diagnostic raters were blind to experimental conditions. Prior studies in our lab using the same procedures have demonstrated excellent inter-rater reliability (κ = .77–.83; Keough and Schmidt, 2012; Schmidt et al., 2014). In the current study a subsample of subjects were used for reliability coding which yielded excellent inter-rater reliability (κ = .86).

3.2. Self-report measures

3.2.1. Beck depression inventory-II (BDI-II)

The BDI-II is a 21-item measure of depressive symptoms (Beck et al., 1996). The BDI-II is widely used and has been shown to have strong psychometric properties among nonclinical and clinical samples (Endler et al., 1999). In the present study, the BDI-II demonstrated excellent internal consistency at baseline (α = .93).

3.2.2. Depressive symptom inventory–suicide subscale (DSI-SS)

The DSI-SS is a 4-item questionnaire assessing suicidal ideation, suicidal plans, control of suicidal thoughts, and suicidal impulses (Metalsky and Joiner, 1997). The DSI-SS has been shown to have strong psychometric properties (Joiner et al., 2002). In the current study, reporting beyond a 0 on any item indicated an incident of suicidal thought.

3.2.3. Interpersonal needs questionnaire-revised (INQ-R)

The INQ-R is a 15-item self-report questionnaire including six items designed to measure perceived burdensomeness and nine items designed to measure thwarted belongingness (Van Orden et al., 2012). The INQ-R has been shown to have strong psychometric properties (Van Orden et al., 2012). In the current study, the burdensomeness and belongingness scales demonstrated good reliability (α’s=.88 and .84, respectively).

3.3. Data analytic plan

Analyses were conducted in Mplus version 8 (Muthén and Muthén, 1998–2017). Treatment condition was modeled as 0=Combined Intervention, 1=Repeated Contact Control. To adjust for potential data nonnormality, the robust maximum likelihood estimator was applied in direct effect models. This approach provides Huber-White sandwich-adjusted standard errors. For mediation models, the maximum likelihood estimator was used to allow for bootstrapped confidence intervals (CIs; with 5000 data resamples) to provide the most accurate testing of the indirect effect (Hayes and Scharkow, 2013; Preacher and Hayes, 2008). To assess overall model fit, the χ² value was used to determine if the model provided exact fit to the tested data. A nonsignificant χ² value indicated that the model fit the data. Model fit indices also provided support for adequate, though not exact fit. Model fit indices assessed included the comparative fit index (CFI) and the root mean square error of approximation (RMSEA). The RMSEA also provides accompanying 90% CIs. A CFI value greater than .95 and an RMSEA value below .05 suggest good fit. A lower bound RMSEA CI containing .05 indicates that good fit cannot be ruled out and an upper bound RMSEA CI containing .10 indicates that poor fit cannot be ruled out (Bentler, 1990; Browne et al., 1993; Hu and Bentler, 1999; MacCallum et al., 1996).

4. Results

4.1. Descriptive statistics, normality, and missing data

Examination of all baseline and mediator variables revealed acceptable levels of skew and kurtosis for the use of robust maximum likelihood (i.e., all skew values < 3, kurtosis < 10; (Curran et al., 1996)). Descriptive statistics for baseline variables by condition are provided in the top panel of Table 1. There were no significant differences found. Incidences of new reported suicidal thoughts by follow-up session are provided in the bottom panel of Table 1. There was a total of 9 participants in the intervention condition (9.89%) and 5 participants in the control condition (10.64%) reporting suicidal thoughts following the intervention.

Table 1.

Descriptive statistics for pre-intervention and post-intervention measures by condition.

	Intervention		Control		F
	Mean	SD	Mean	SD
Pre-intervention
INQ perceived burdensomeness	12.51	8.28	11.36	6.48	.68
INQ thwarted belongingness	33.53	11.89	34.26	11.66	.12
BDI-2 depression Post-intervention	19.34	11.85	18.66	11.79	.10
INQ perceived burdensomeness	9.59	6.23	11.74	8.05	2.96
INQ thwarted belongingness New suicidal thoughts cases	28.75	12.08	31.11	13.59	1.07
Month 1 suicidal thoughts	2	2.30%	3	6.52%
Month 3 suicidal thoughts	2	2.53%	1	6.82%
Month 6 suicidal thoughts	5	7.94%	1	9.38%
Total suicidal thoughts	9	9.89%	5	10.64%

Note. INQ = Interpersonal Needs Questionnaire. BDI-2 = Beck Depression Inventory-2. Month 1 Intervention N = 87, Control N = 46. Post-Intervention Intervention N = 91, Control = 46. Month 3 Intervention N = 79, Control N = 41. Month 6 Intervention N = 63, Control N = 29. Total Intervention N = 91, Control N = 47.

4.2. Treatment condition predicting suicidal thoughts

The direct effect of treatment condition on suicidal thoughts post-intervention was first examined. As expected, the effect of treatment condition on suicidal thoughts was not significant (B =.04, p=.89). Next a mediation model, including PB immediately following the intervention as a mediator from treatment condition to suicidal thoughts was examined. In this model, gender was included as a covariate on post-intervention PB and on suicidal thoughts. Baseline PB was included as a control variable for post-intervention PB. This model provided excellent fit to the data (χ² =.16, df =1, CFI =1.00, RMSEA =.00, 95% CI [.00, .17]). Treatment condition significant predicted reductions in PB (B =2.50, p < .05), such that on average levels of PB were 2.50 points lower in the treatment condition compared to the control condition. Treatment condition remained a nonsignificant predictor of suicidal thoughts (B =−.06, p=.89). Post-intervention PB marginally predicted suicidal thoughts (B =.05, p=.06). Finally, there was a significant indirect effect from treatment condition to suicidal thoughts through PB (B =.12, 95% CI [.01, .32]). This model accounted for 10.4% of the variance in suicidal thoughts at follow-up.

A mediation model including TB instead of PB was then examined, including the same or commensurate covariates. This model also provided excellent fit to the data (χ² =.10, df =1, CFI =1.00, RMSEA =.00, 95% CI [.00, .16]). Treatment condition was not a significant predictor of post-intervention TB (B =1.42, p=.43) or of follow-up suicidal thoughts (B =−.02, p=.96). Post-intervention TB was not a significant predictor of suicidal thoughts (B=.02, p=.11). Finally, the indirect effect from treatment condition to suicidal thoughts through TB was not significant (B =.03, 95% CI [− .04, .18]).

Finally, the model including PB as a predictor was examined, including baseline depression as a covariate (predicting post-intervention PB and follow-up suicidal thoughts). This model provided excellent fit to the data (χ² =.04, df =1, CFI =1.00, RMSEA =.00, 95% CI [.00, .13]). Fig. 1 contains unstandardized parameters for this model. Treatment condition significantly predicted post-intervention PB (B =2.38, p=.02) as did gender (B =−2.10, p=.04) and baseline depression (B =.23, p < .001). Baseline PB only marginally predicted post-intervention PB (B =.20, p=.08). Only post-intervention PB predicted follow-up suicidal thoughts (B =.05, p=.04). Further, the mediation pathway through post-intervention PB was significant (B =.11, 90% CI [.00, .31]).

Fig. 1.

The effect of treatment condition (0=Active, 1=Control) on suicidal thoughts post-intervention (i.e., month 1 through month 6) through post-intervention perceived burdensomeness, controlling for gender and baseline depression. Solid lines indicate significance; dotted lines indicate nonsignificance.

5. Discussion

The current study examined whether an RCT focused on reducing PB and TB could reduce incidence of suicidal thoughts in a high-risk sample of people reporting no suicidal thoughts at baseline. Consistent with hypotheses, reductions in PB predicted lower incidence of suicidal ideation over time. In contrast, the intervention had no effect on TB nor on incidence of suicidal thoughts via TB reductions.

In the current study, PB was reduced through a brief intervention targeting top-down (psychoeducation) and bottom-up (CBM-I) processes, in a sample of people with elevated levels of PB and TB. Van Orden et al. (2016) also found that PB could be reduced in an RCT focused on increasing social connectedness in older adults through pairing them with untrained peer companions. Hill and Pettit (2016) found that PB could be reduced in a sample of adolescents completing a two-week online intervention, using CBT principles to target PB. The wide range of approaches, sample characteristics, and length of interventions across these three studies provide evidence that PB is a malleable risk factor that can be reduced through multiple intervention strategies.

In contrast, there was no evidence that TB could be reduced using the same brief intervention. These findings are consistent with those of Van Orden et al. (2016) and Hill and Pettit (2016), who also found no effects of their interventions on TB. Further, as Ma et al. (2016) report, there is mixed evidence of an association between TB and suicidal thoughts, especially when controlling for the effects of PB. Therefore, using the framework provided by Kraemer et al. (1997), there is no evidence for TB as a causal risk factor for suicidal thoughts. These findings are in contrast to the theoretical underpinnings of the IPTS, which describes TB as a causal risk factor (e.g., Joiner, 2005; Van Orden et al., 2010). These findings suggest a need to reconcile the role of TB in the IPTS with current evidence.

To our knowledge, this study is the first to demonstrate that reductions in PB could limit the incidence of suicidal thoughts up to six months later. This finding is especially noteworthy when considering the ease of implementing the intervention. This intervention requires three computer-guided sessions with minimal clinician contact. Given the current suicide crisis, researchers have increasingly called for scalable, cost-efficient, and technology-enhanced interventions (e.g., Kreuze et al., 2016). The current intervention meets these criteria and suggests that it would seem worthwhile to explore approaches to disseminate this treatment to those in need, particularly those presenting with high levels of PB.

Targeting risk factors for suicidal thoughts such as PB in a preventative framework may be particularly advantageous given that PB is more likely to be reported than suicidal thoughts. Despite efforts to destigmatize suicidal thoughts (National Action Alliance for Suicide Prevention: Research Prioritization Task Force, 2014), suicidal thoughts and behaviors are still underreported due to the associated stigma (Rimkeviciene et al., 2015; Rudd et al., 2013). Therefore, through this and similar treatments, clinicians may be able to effectively target suicidal thoughts for people reporting PB, regardless of whether people are willing to admit current suicidal thoughts.

Although these findings are promising, there are several limitations to consider. First, there were two elements to the current intervention, a CBM and a psychoeducation component. Therefore, it is unclear whether these results were due to one or both components and whether the effects were additive or synergistic. Addressing these questions will aid in further refining treatment efficacy. In addition, all measures used in the current study were self-report measures. Future studies should integrate additional units of measurement. For example, performance on implicit association tasks and difficulties modulating emotions as measured by neurophysiological recordings have both been linked to suicidal thoughts (Cha et al., 2010; Kudinova et al., 2016) and might be useful outcomes to consider in addition to self-report. Further, the incidence of any suicidal thoughts was examined as the outcome variable. It would be relevant to consider severity of these thoughts in future studies involving larger samples to allow for variance in severity to be modeled. Finally, suicidal thoughts were only assessed during relatively brief periods during in the six months following the intervention. Given that suicidal thoughts can be fleeting, approaches such as ecological momentary assessment (EMA; Shiffman et al., 2008) might be valuable in capturing whether the current intervention also reduces day-to-day variability in suicidal thoughts.

In spite of these limitations, there are multiple strengths of the current study. First, this study was the first to suggest that reducing perceived burdensomeness may impact later incidence of suicidal thoughts. Therefore, the current study provides the most compelling evidence to date for a selective intervention targeting PB to mitigate the likelihood of developing suicidal thoughts. Further, these findings provide evidence of theoretically and clinically important specificity in that they suggest that PB, but not TB, can be considered a causal risk factor for suicidal thoughts. Future studies are needed, especially studies confirming these results using additional methods of assessing suicidal ideation.