Posttraumatic Stress and Distress Tolerance

ORIGINAL ARTICLE

Posttraumatic Stress and Distress Tolerance Associations With Suicidality in Acute-Care Psychiatric Inpatients Anka A. Vujanovic, PhD,* Jafar Bakhshaie, MD, MA,* Colleen Martin, MA,† Madhavi K. Reddy, PhD,‡ and Michael D. Anestis, PhD§

Abstract: Trauma and posttraumatic stress disorder (PTSD) symptomatology have been associated with suicidality, including ideation and behavior. The current investigation evaluated, in acute-care psychiatric inpatients, the mediating role of perceived (self-reported) distress tolerance in the association between PTSD symptom severity and suicidality, defined as a) suicidal ideation, intent, or behavior leading to current psychiatric hospitalization; b) self-reported severity of suicidal desire; and c) percentage of days of suicidality during current hospitalization. Participants were composed of 105 adults (55.2% women; mean age, 33.9; SD, 10.9) admitted to a public psychiatric acute-care inpatient hospital in a large metropolitan area; 52.3% of the participants were hospitalized for suicidality. Results indicated that PTSD symptom severity (and severity of each PTSD symptom cluster) may exert an indirect effect on suicidality, specifically suicidality as a basis for current hospital admission and self-reported severity of suicidal desire, through perceived distress tolerance. Effects were documented after controlling for theoretically relevant covariates. Key Words: Trauma, PTSD, distress tolerance, suicide, inpatients (J Nerv Ment Dis 2017;205: 531–541)

I

n the United States, suicide is the 10th leading cause of death, occurring at a rate of approximately 13 per 100,000 individuals (Centers for Disease Control and Prevention, 2016). Suicide rates have been highest among individuals with previous suicide attempts, highlighting the chronic nature of suicidality and the clinical utility of evidence-based prevention programs for such individuals (Yuodelis-Flores and Ries, 2015). The effective prevention of suicide requires a greater empirical understanding of suicidality, defined here as suicidal ideation, intent, plan, or behavior (Krysinska and Lester, 2010). Given the significant public health relevance of suicidality, there has been a focus on examining more specific etiological and maintenance factors related to suicidal ideation and attempts (Han et al., 2015; O'Connor et al., 2015; Yuodelis-Flores and Ries, 2015). Both traumatic life events, defined as those involving “actual or threatened death, serious injury, or sexual violence” (American Psychiatric Association, 2013), and posttraumatic stress disorder (PTSD) symptomatology have been identified as major risk factors for suicidality (Arsenault-Lapierre et al., 2004; Bentley et al., 2016; Brenner et al., 2011; Cavanagh et al., 2003; Jankovic et al., 2013; Krysinska and Lester, 2010; Krysinska and Martin, 2009; Lopez-Castroman et al., 2015; Tarrier and Gregg, 2004). Heightened prevalence or incidence rates of suicidality (i.e., ideation, plan, and/or attempt) have been documented among individuals with PTSD (Krysinska and Martin, *Department of Psychology, University of Houston; †Michael E. DeBakey Veterans Affairs Medical Center; ‡Department of Psychiatry and Behavioral Sciences, McGovern Medical School, University of Texas Health Science Center, Houston, Texas; and §Department of Psychology, University of Southern Mississippi, Hattiesburg, Mississippi. Send reprint requests to Anka A. Vujanovic, PhD, Trauma and Stress Studies Center, Department of Psychology, University of Houston, 126 Heyne Building, 3695 Cullen Boulevard, Houston, TX 77204. E‐mail: [email protected]. Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved. ISSN: 0022-3018/17/20507–0531 DOI: 10.1097/NMD.0000000000000690

2009; Lopez-Castroman et al., 2015; Panagioti et al., 2012a, 2012b; Tarrier and Gregg, 2004). The association between PTSD and suicidal ideation is robust, remaining significant even after controlling for relevant demographic variables and psychiatric diagnoses (e.g., major depression; Mazza, 2000; Ullman and Brecklin, 2002). According to a meta-analysis examining associations between PTSD and suicidal ideation and suicide attempts across various populations (e.g., community members, domestic violence and sexual assault victims, substance abusers, psychiatric inpatients, and war veterans), individuals with PTSD, as compared with those without PTSD, displayed 3.6 to 5.1 times more suicidal ideation and 2.7 to 6.0 times more nonfatal suicidal behavior (Kessler et al., 1999; Krysinska and Lester, 2010; Sareen et al., 2005). Despite the well-established association between PTSD and suicidality, there is a paucity of research on malleable cognitive-affective factors, targetable via cognitive behavioral intervention, underlying the association. This is unfortunate, because a better understanding of the role of such cognitive-affective process in the well-established association between PTSD and suicidality has the potential to meaningfully inform suicide prevention efforts. Distress tolerance, defined as the perceived or actual ability to tolerate negative or aversive emotional or physical states (Leyro et al., 2010), is one such promising factor relevant to the PTSD-suicidality association. The distress tolerance construct tends to be characterized by assessment modality and generally defined as perceived (i.e., selfreport) or behaviorally indexed (e.g., computer tasks, cold pressor) distress tolerance. Furthermore, low and high levels of distress tolerance, measured via either self-report or behavioral measure, have been associated with distinct aspects of suicidality. Low levels of perceived distress tolerance have been associated with heightened suicidal desire (Anestis et al., 2011a, 2011b) or suicidal ideation. Conversely, high levels of behaviorally indexed distress tolerance have been associated with increased capability for suicide (Anestis and Joiner, 2012), defined as elevated tolerance of physical pain and diminished fear of death. Capability for suicide is theorized to develop through repeated exposure to painful and/or provocative events (Anestis et al., 2011a). A combination of suicidal desire and capability for suicide is theorized to confer heightened risk for suicide (e.g., Van Orden et al., 2008). Anestis et al. (2013) documented the moderating role of heightened perceived distress tolerance in the association between nonsuicidal self-injury and suicide potential, or the capacity to engage in suicidal behavior. Furthermore, higher levels of behaviorally indexed distress tolerance significantly moderated the association between PTSD symptom severity and number of past medically attended suicide attempts in a sample of inpatients in a substance use disorder residential treatment program (Anestis et al., 2012). Notably, the investigation by Anestis et al. (2012) represents the only published study to date to document the role of distress tolerance in the association between PTSD and suicidal behavior. The extant empirical and theoretical literature suggests that higher levels of an individual's behaviorally indexed ability to tolerate physical or psychological distress may exacerbate the association between PTSD symptomatology and suicidality, demonstrating associations with heightened capability for suicide (Anestis et al., 2011a,

The Journal of Nervous and Mental Disease • Volume 205, Number 7, July 2017 Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.

www.jonmd.com

531

Vujanovic et al.

The Journal of Nervous and Mental Disease • Volume 205, Number 7, July 2017

2011b; Bender et al., 2012). However, heightened perceived emotional distress tolerance has been associated with increased suicidal desire (Anestis et al., 2011a) but a theoretically lower capability of actually engaging in potentially lethal behaviors. Theoretically, trauma-exposed individuals with lower levels of the perceived ability to tolerate negative emotional states may be especially likely to struggle with PTSD symptomatology, to experience such symptomatology as intolerable and unmanageable, and to manifest suicidal ideation or desire to “escape” or “end” the distress. Lower levels of the perceived (versus actual) ability to withstand negative emotionality may account for—or mediate—the association between PTSD symptom severity and certain forms of suicidality, particularly suicidal desire. Conversely, one's actual, or behaviorally indexed, ability to tolerate distress or physical pain may be a separate—though related—construct that confers greater risk for suicidal behavior. Because the extant literature has found little to no convergence among self-report and behavioral indices of distress tolerance (Leyro et al., 2010; Marshall-Berenz et al., 2010; McHugh et al., 2011), it is important to examine the variously defined dimensions of the distress tolerance construct in relation to suicidality among trauma-exposed, at-risk populations. Taken together, there are several limitations inherent in the extant literature on PTSD symptomatology and suicidality. First, although the association between trauma exposure and suicidality has been well established (Beristianos et al., 2016; Dias de Mattos Souza et al., 2016; Stein et al., 2010), more research focused on examining suicidality in trauma-exposed samples (with various types of psychopathology) is necessary to better understand the role of trauma exposure in suicidal ideation and risk. Second, few studies to date have examined the associations between PTSD symptomatology and suicidality in psychiatric inpatient samples (Anestis et al., 2012; Dore et al., 2012; Huang et al., 2012; Oquendo et al., 2003; Ramberg et al., 2015). This shortage of data is unfortunate because psychiatric inpatient settings harbor an especially high-risk population with regard to suicidal ideation and behavior (Anestis et al., 2012; Pompili et al., 2014) as well as trauma exposure and PTSD symptomatology (Dore et al., 2012; Havens et al., 2012; McCormack and Adams, 2016; Muskett, 2014). Most relevant studies examining PTSD and suicidality in inpatient settings to date have been based on adolescent (Brand et al., 1996; Havens et al., 2012) and/or substance-abusing samples (Anestis et al., 2012; Dore et al., 2012). The empirical literature on general adult acute-care psychiatric inpatient samples is significantly limited. Third, no studies to date have examined the mediating role of perceived distress tolerance in the association between PTSD symptom severity and suicidality in acute-care psychiatric inpatients. Finally, few relevant studies to date have focused on low-income, inner-city adults, a historically underrepresented and understudied group in the empirical literature. The overarching aim of the current investigation was to address these gaps in the literature by evaluating the indirect effect of PTSD symptom severity on suicidality via perceived (self-reported) distress tolerance among a sample of low-income, inner-city acute-care psychiatric inpatients. In the current study, suicidality was defined as a) suicidal ideation, intent, or behavior leading to current psychiatric hospitalization; b) self-reported severity of suicidal desire; and c) percentage of days of suicidality during current hospitalization. First, it was hypothesized that PTSD symptom severity would be significantly (positively) associated with each of the suicidality outcomes. Second, it was hypothesized that perceived distress tolerance would be negatively (inversely) associated with each of the suicidality outcomes. Third, it was hypothesized that the association between PTSD symptom severity and each of the suicidality outcomes would be indirect through distress tolerance. Specifically, PTSD symptom severity was expected to be associated with distress tolerance, which, in turn, would be associated with each of the suicidality outcomes. Finally, post hoc exploratory analyses were conducted to examine each of the four PTSD symptom clusters (i.e., intrusions, avoidance, negative alternations in cognitions 532

www.jonmd.com

and mood, and arousal) as predictors in the models. Rigorous models were tested, as all effects were expected above and beyond the variance accounted for by theoretically relevant covariates, including sex, number of trauma exposure types, number of psychiatric diagnoses, substance use (i.e., positive urine drug screen at admission), and number of prior suicide attempts. These variables were selected as covariates because of past work demonstrating associations between each of these variables and suicidality (Ferrada-Noli et al., 1998; Goldstein et al., 1991, 2009; Heydari et al., 2014; Ilgen et al., 2007; Jeon et al., 2013; Kim et al., 2016).

METHOD Participants Participants were composed of 105 adults (55.2% women; mean age, 33.9; SD, 10.9) admitted to a public, university-affiliated, psychiatric acute-care inpatient hospital in a large metropolitan area in the southern United States. Participants were hospitalized for an average of 8.7 days (SD, 4.4; range, 2–32). Inclusionary criteria were composed of being between 18 and 65 years of age and reporting a history of trauma exposure, consistent with the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5; American Psychiatric Association, 2013) PTSD Criterion A (i.e., participants did not need to meet criteria for a full PTSD diagnosis per electronic medical records). Exclusionary criteria were composed of an inability to provide verbal and written informed consent and significant cognitive impairment, as defined below. Please see Tables 1 and 2 for information regarding participants' clinical and sociodemographic characteristics.

Measures Medical Records Review Two suicidality outcome variables were obtained from medical records: a) suicidality as a basis for current hospital admission (yes/no) and b) percentage of days (of total days hospitalized) of psychiatrist- or nurse-coded suicidality during the hospitalization period. Suicidality as a basis for current hospital admission was defined as significant suicidal ideation, including intent or plan, and/or suicidal behavior that necessitated hospitalization for stabilization. The number of self-reported past suicide attempts also was extracted from the medical record and evaluated as a covariate within each mediation model. Psychiatric diagnoses were determined via clinical interviews between psychiatrists and patients. Discharge psychiatric diagnostic data were used to inform the diagnostic composition of the present sample, because discharge diagnostic data are more comprehensive and provide a standardized time point for extraction of diagnostic data. The number of psychiatric diagnoses variable was entered as a covariate within each model. In addition, urine drug screen data at admission were obtained from medical records. At admission, all patients were screened for cocaine, cannabis, phencyclidine, amphetamines, opiates, benzodiazepines, barbiturates, methadone, propoxyphene, and methaqualone. A dichotomous urine drug screen variable was created and entered as a covariate within each model. A summary of sample characteristics is reported in Table 1.

Mini Mental State Examination The Mini Mental State Examination (MMSE; Folstein et al., 1975) is an 11-item instrument used as an objective screening assessment for general mental status. The MMSE provides a brief screening of abilities in the areas of attention, memory orientation (recall of words, recognition of sentences), and initiation and maintenance of verbal and motor responses. Scores range from 0 to 30. Scores of 20 or below are indicative of moderate cognitive impairment (Crum et al., 1993); thus, a score of 20 or below was used as an exclusionary © 2017 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.


TABLE 1. Participant Characteristics

Posttraumatic Stress and Suicidality

TABLE 1. (Continued)

Variable

Mean (SD) or % (n) a

Race/ethnicity White Black/African American Hispanic Asian “Other” Marital statusa Single/never married Married Divorced Separated Widowed Not reported Socioeconomic statusb Less than high school High school/GED Some college College degree No. days with paid work (past 30 days) Monthly income (past 30 days)

43.8% (46) 39.0% (41) 13.3% (14) 2.9% (3) 1.0% (1) 65.7% (67) 12.4% (13) 11.4% (12) 8.6% (9) 1.9% (2) 2.9% (3) 27.8% (29) 25.3% (26) 21.8% (23) 9.9% (10) 6.2 (14.2) $409.50 (SD, $1,500.60)

Hospitalization and suicidality characteristics MMSE total scoreb No. days—current hospitalizationa Suicidality as precipitant to current hospitalization (yes/no)a Suicidality—during current hospitalization (yes/no)a No. days of suicidality—current hospitalizationa Positive urine drug screen at admission (yes/no)a Traumatic event exposure typesc Physical assault Transportation accident Natural disaster Childhood physical abuse Serious accident Assault with a weapon Witnessed sudden violent death Sexual assault Life-threatening illness or injury Childhood sexual abuse Other unwanted sexual experiences Fire or explosion Captivity Causing serious injury to someone else Combat war zone exposure Other stressful event or experience (e.g. childhood neglect) DSM-IV diagnoses (axis I)a Substance use disorders Cocaine abuse/dependence Cannabis abuse/dependence Alcohol abuse/dependence Cannabis abuse/dependence

Variable

Mean (SD) or % (n)

Amphetamine abuse/dependence Polysubstance abuse/dependence Opioid abuse/dependence Anxiolytic abuse/dependence Amphetamine abuse/dependence Hallucinogen abuse/dependence Mood disorders Bipolar I disorder with or without psychosis Bipolar disorder NOS Major depressive disorder with or without psychosis Bipolar II disorder Mood disorder NOS Psychotic disorders Schizoaffective disorder Schizophrenia Schizophreniform disorder Substance-induced psychosis Anxiety disorders Generalized anxiety disorder PTSD

12.3% (13) 10.4% (11) 1.9% (2) 1.9% (2) 0.9% (1) 0.9% (1) 63 24.7% (26) 18.0% (19) 13.2% (14) 2.8% (3) 0.9% (1) 20 7.6% (8) 5.7% (6) 3.8% (4) 1.9% (2) 4 1.9% (2) 1.9% (2)

a

27.8 (1.9) 8.7 (4.4) 52.3% (55) 14.2% (15) 0.3 (0.6) 41.9% (44)

Data derived from electronic medical records. MMSE. c Data derived from LEC-5. b

criterion for the present study. In the current study, the range of MMSE scores was 22 to 30.

Life Events Checklist-5 62.9% (66) 61.0% (64) 59.0% (62) 53.3% (56) 44.8% (47) 41.0% (43) 36.6% (38) 36.2% (38) 35.2% (37) 30.5% (32) 24.8% (26) 24.8% (26) 19.0% (20) 17.8% (19) 11.0% (12) 38.1% (40)

111 25.7% (27) 20% (21) 16.1% (17) 15.2% (16)

The Life Events Checklist-5 (LEC-5; Weathers et al., 2013). The LEC-5 is a self-report measure used to screen for potentially traumatic events experienced anytime during the life span. The LEC-5 presents respondents with 16 potentially traumatic events (e.g., natural disaster, combat, sexual assault) and includes an additional item assessing for “other” potentially traumatic events not listed. In the current study, respondents were asked to indicate (via check mark) whether each listed event “has happened to you at some point in your life.” Two items from the LEC-5 (i.e., “exposure to toxic substance” and “severe human suffering”) were removed, and two items were added (i.e., “childhood physical abuse” and “childhood sexual abuse”). These modifications were instituted to simplify the measure, to reduce potential for falsepositive reports (e.g., endorsement of “exposure to toxic substance” by patients with psychotic-spectrum psychopathology), and to add a childhood time frame for endorsement of sexual or physical abuse/ assault. Study staff verbally asked participants to identify a “worst event” (i.e., most distressing event) on the LEC-5 to complete the PTSD Checklist–Civilian Version–5 (PCL-5). In the current study, the LEC-5 was used to determine (1) trauma exposure, an inclusionary criterion for the study, and (2) the number of traumatic life event types each participant experienced, a covariate.

PTSD Checklist–Civilian Version–5 The PCL-5 (Blevins et al., 2015) is a 20-item self-report measure of PTSD symptom severity. Each of the 20 items reflects a DSM-5 symptom of PTSD. Respondents were verbally asked by study staff to complete the PCL-5 with regard to the “worst event” identified on the

© 2017 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.

www.jonmd.com

533


Vujanovic et al.

TABLE 2. Descriptive Statistics and Bivariate Correlations Between Study Variables (N = 105) Variable 1. Sex (female)a 2. No. trauma typesa,b 3. No. psychiatric diagnosesa,c 4. Substance usea,d 5. No. past suicide attemptse,f 6. PCL-5g,h 7. DTSg,i 8. Suicide admissionj,k,l 9. BSS-5j,m 10. % Suicidal days hospitalizationj,n

Mean (SD)/% (n)

1

2

3

4

5

6

7

8

9

10

55.2% (58) 5.93 (3.19) 1.97 (1.13) 41.5% 1.11 (1.48) 40.2 (21.5) 42.8 (14.8) 52.3% (55) 1.33 (2.39) 0.05 (0.09)

1 0.05 0.02 0.12 0.17 0.14 −0.10 0.01 −0.10 0.19

1 0.03 0.07 0.15* 0.49** −0.22* −0.07 0.25** 0.11

1 0.40** 0.28 0.13 0.02 0.11 0.18 0.05

1 0.02 −0.02 0.14 0.10 −0.03 −0.09

1 0.38** −0.15 0.35** 0.19* 0.16

1 −0.50** 0.24* 0.42** 0.07

1 −0.33** −0.41** −0.05

1 0.17* 0.09

1 0.02

1

**p < 0.01, *p < 0.05. All significance levels are based on 95% bootstrapped CIs of correlation coefficients. a Covariate; bnumber of traumatic event types, LEC-5; cnumber of psychiatric diagnoses (DSM-IV, axis I), electronic medical records; dpositive urine drug screen at admission, electronic medical records; epredictor; fnumber of self-reported prior suicide attempts, electronic medical record; gindirect explanatory variable; hPTSD Checklist-5; iDTS—total score; joutcome variable; ksuicidality as reason for current hospital admission, electronic medical records (0 = no; 1 = yes); lpoint-biserial correlation/categorical variable; mBeck Scale for Suicide Ideation-5; npercentage of days (of total number of days hospitalized) of self-reported suicidality during current hospitalization.

LEC-5, and the study staff oriented the participants to the LEC-5 when completing the PCL. Each PCL-5 item is rated on a 5-point scale (0 [not at all] to 4 [extremely]) in terms of how often they have been bothered by the symptom in the past month (e.g., “In the past month, how much have you been bothered by repeated, disturbing, and unwanted memories of the stressful experience?”). Total symptom severity scores range from 0 to 80, where higher scores indicate higher symptom severity, with a preliminary cutoff score of 33 recommended in the current literature. The PCL-5 has demonstrated good internal consistency, test-retest reliability, and convergent and discriminant validity (Armour et al., 2015; Bovin et al., 2015). In the current study, the PCL-5 demonstrated good internal consistency (Cronbach's α = .94). The PCL-5 total score was used to represent level of PTSD symptom severity, the predictor variable. The PCL-5 also was used to derive the PTSD symptom cluster severity scores, predictor variables in the post hoc exploratory analyses.

Distress Tolerance Scale The Distress Tolerance Scale (DTS; Simons and Gaher, 2005) is a 15-item self-report inventory measured on a 5-point scale (1 [strongly agree] to 5 [strongly disagree]), which evaluates the extent to which respondents believe they can experience and withstand distressing emotional states (Simons and Gaher, 2005). The DTS contains four subscales: tolerance (“I can't handle feeling distressed or upset”), appraisal (“Being distressed or upset is always a major ordeal for me”), absorption (“When I'm distressed or upset, I cannot help but concentrate on how bad the distress actually feels”), and regulation (“I'll do anything to stop feeling distressed or upset”). The total score ranges from 15 to 75, with higher values indicating greater distress tolerance. The DTS demonstrates good psychometric properties, including good internal consistency, test-retest reliability, convergent validity, and discriminant validity with established measures of mood (Simons and Gaher, 2005). For the current study, the DTS demonstrated good levels of internal consistency (Cronbach's α = .92). The DTS total score was used to represent level of distress tolerance, as consistent with relevant past literature (Simons and Gaher, 2005; Vujanovic et al., 2013). The DTS total score was evaluated as a mediator variable.

Beck Scale for Suicide Ideation The Beck Scale for Suicide Ideation (BSS-5; Beck and Steer, 1991) is a five-item self-report screening measure that measures the 534

www.jonmd.com

severity of an individual's suicidal ideation or desire to die (Lam et al., 2004). The five-item screening version of the BSS was implemented to reduce participant burden for nonsuicidal patients (Lam et al., 2004). Using a 3-point scale, the BSS-5 assesses the most accurate statement for the intensity of past-week suicidal ideation (e.g., “I have no wish to live”). The BSS has demonstrated high internal consistency (α = .96) and strong test-retest reliability over a 1-week period (r = 0.88) in psychiatric inpatient populations (Beck et al., 1988; Pinninti et al., 2002). In the current study, the BSS-5 demonstrated good internal consistency (Cronbach's α = .90). The BSS-5 total score was utilized as a measure of self-reported suicidal desire severity, an outcome variable.

Procedure All individuals assigned to one unit in an acute-care psychiatric inpatient hospital were screened for a history of trauma using the LEC-5 within 24 hours of admission. Within 5 days of admission, study staff approached individuals who endorsed at least one traumatic event on the LEC-5 for potential participation. Individuals who were willing to participate provided informed verbal and written consent. Participants were first administered the MMSE. Participants then completed a self-report packet, which included the questionnaires used for the current study. Participation was completely voluntary; no financial compensation was provided. The study was approved by all relevant institutional review boards.

Data Analytic Plan First, descriptive statistics and bivariate correlations were calculated for all study variables. Second, data were examined for multivariate outliers and normality. Two variables (number of prior suicide attempts and percentage of days of suicidality during current hospitalization) showed highly nonnormal distributions (skewness > |2|, kurtosis > |7|; Curran et al., 1996; Kline, 2005). Therefore, bootstrapped confidence intervals (95% CI with 5000 resamples) were calculated for all correlation and regression coefficients. This nonparametric estimation method makes no assumptions about the shape of the distributions and produces robust estimates (Davison and Hinkley, 1997; Preacher et al., 2007). © 2017 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.


Third, regression coefficients for each hypothesized path were evaluated, and mediation analyses were conducted using bootstrapping techniques through PROCESS Macro for SPSS (Hayes and Preacher, 2014), a computational tool for observed variable analysis. Bootstrapping as a nonparametric method estimates the sampling distribution of an estimator based on resampling with replacement. The indirect effect was computed for each of the resamples, which results in an empirically generated sampling distribution (Hayes and Preacher, 2014). PTSD symptom severity (PCL-5) was the predictor, with distress tolerance (DTS) as the proposed explanatory (mediator) variable in all models; separate analyses were conducted for each outcome variable: a) suicidality as a basis for current hospital admission, b) self-reported suicidal ideation (BSS-5), and c) percentage of days with suicidality during current hospitalization (see Fig. 1). Covariates included a) sex, b) number of traumatic event types experienced, c) number of DSM-IV axis I diagnoses (per medical records), d) substance use (i.e., positive urine drug screen, per medical records), and e) number of prior suicide attempts (per medical records). Five thousand bootstrap resamplings were conducted to detect the indirect effects of the proposed predictor (PCL-5) on outcome variables via DTS. The indirect effect is calculated as the product of the beta coefficients of the “a” and “b” paths. A bootstrap confidence interval that does not include zero provides evidence of a significant indirect effect (Preacher and Hayes, 2008). Effect sizes (K2) were calculated for each indirect effect (small, 0.01; medium, 0.09; large, 0.25; Preacher and Kelley, 2011). Fourth, to test the specificity of the theoretical models and per recommendations for cross-sectional designs, two alternative models were also tested for each outcome variable. Specifically, in the alternative models, first the predictor and explanatory variables were reversed (alternative model 1), and then the dependent and explanatory variable were reversed (alternative model 2) (Kraemer et al., 2008; Preacher and Hayes, 2008). Finally, post hoc exploratory analyses, identical to those described for the main analyses, were conducted to examine the role of each of four PTSD symptom clusters (intrusions, avoidance, negative alterations in cognitions and mood, and arousal), per the PCL-5, as predictors in separate models.

Posttraumatic Stress and Suicidality

RESULTS Descriptive Statistics and Correlational Data Please see Table 1 for a summary of participant characteristics, including psychiatric diagnoses, sociodemographic characteristics, and trauma exposure types and substance use characteristics. Please see Table 2 for a summary of correlational and descriptive data for all study variables. The average number of traumatic event types experienced by this sample was 5.3 (SD, 3.1; range, 1–17), with physical assault (62.9%), transportation accident (61.2%), and natural disaster (59.0%) as the most commonly endorsed types. The mean PCL-5 score for the sample is 40.2 (SD, 21.5), indicating high levels of PTSD symptomatology. Approximately 64.2% of the sample met criteria for PTSD per the recommended PCL-5 diagnostic cutoff of 33 (Bovin et al., 2015). With regard to suicidality, participants reported an average of 1.11 (SD, 1.48; range, 0–8) suicide attempts before current admission, and 52.3% were currently hospitalized due to suicidality. Furthermore, the average number of DSM-IV (American Psychiatric Association, 1994) axis I psychiatric diagnoses per participant was 1.97 (SD = 1.13). With regard to correlational data, PTSD symptom severity (PCL-5 total score) was significantly negatively associated with distress tolerance (DTS total score). PTSD symptom severity and distress tolerance (PCL-5 and DTS total scores, respectively) were both significantly associated with suicidality as a basis for current hospital admission and severity of suicidal desire (BSS-5). Suicidality as a basis for current hospital admission was significantly positively associated with severity of suicidal desire (BSS-5) and number of past suicide attempts. Number of past suicide attempts was positively associated with severity of suicidal desire. Percentage of days of suicidality during hospitalization was not significantly correlated with any study variables.

Mediation Analyses Please see Table 3 for a summary of mediation analyses. The models also were run with an alternative substance use covariate, composed of self-reported past-month alcohol and substance use, and the pattern of results remained consistent. Notably, the mediation models

FIGURE 1. This figure provides a depiction of the theoretical model tested, whereby PTSD symptom severity is associated with suicidality indirectly via distress tolerance. Covariates include sex, number of traumatic event types, number of psychiatric diagnoses, substance use (positive urine drug screen), and number of past suicide attempts. Note: a, effect of X on M; b, effect of M on Y; c, total effect of X on Y; c′, direct effect of X on Y controlling for M; a*b, indirect effect of M.

© 2017 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.

www.jonmd.com

535


Vujanovic et al.

TABLE 3. Standardized Regression Coefficients: Mediation Models Y

Modela

Model R2

β

p

1

PCL-5 ➔ DTS (a) DTS ➔ suicide-admission (b) PCL-5 ➔ suicide-admission (c) PCL-5 ➔ suicide-admission (c′) PCL-5 ➔ DTS ➔ suicide-admission (a*b) DTS ➔ BSS-5 (b) PCL-5 ➔ BSS-5 (c) PCL-5 ➔ BSS-5 (c′) PCL-5 ➔ DTS➔ BSS-5 (a*b) DTS ➔ %suicide-hospitalization (b) PCL-5 ➔ %suicide-hospitalization (c) PCL-5 ➔ %suicide-hospitalization (c′) PCL-5 ➔ DTS ➔ %suicide-hospitalization (a*b)

0.27* 0.27* 0.21* 0.27*

−0.54 −0.86 0.59 0.25 0.43 −0.96 0.97 0.56 0.41 −0.002 −0.009 −0.01 0.003