Borderline personality disorder and self-directed violence in a sample of suicidal Army soldiers

Abstract

There has been a sharp increase in the military suicide rates in 2004. While, borderline personality disorder (BPD) has a stronger association with suicide attempts than any other mental health disorder, there is limited evidence concerning the prevalence and scope of BPD symptoms among military personnel. This secondary data analysis compared active duty suicidal soldiers to characterize the presence-absence of BPD diagnostic criteria and lifetime history of self-directed violence in a suicidal military sample. The current study examined data of 137 active Service Members with significant suicidal ideation. Approximately one-quarter of the soldiers in this sample met full diagnostic criteria for BPD. The presence of BPD criteria was generally consistent among participants with BPD who reported past self-directed violence relative to those who did not. The number of BPD criteria was a significant predictor of the odds of reporting any non-suicidal self-injury (NSSI) as well as the amount of NSSI, but was not associated with suicide attempt. This study demonstrated that a non-trivial proportion of suicidal soldiers meet criteria for this condition, which is strongly associated with self-directed violence. It is important to rigorously assess for the presence-absence of BPD criteria among suicidal military personnel and cultivate prevention strategies and treatment options for BPD.

Suicide rates have been steadily increasing since 1999 (Curtin, Warner, & Hedegaard, 2016) to become the 10^th leading cause of death in the United States (Kochanek, Murphy, Xu, & Arias, 2017; Murphy, Xu, Kochanek, Curtin, & Arias, 2017), resulting in over $50 billion in estimated medical bills and lost wages per year (Florence, Simon, Haegerich, Luo, & Zhou, 2015). Since the 1950s, the overall U.S. military suicide rate has been lower than the U.S. civilian rate (Cassimatis & Rothberg, 1997; Nock et al., 2013; Rothberg, Ursano, & Holloway, 1987). However, in 2004 there was a sharp increase in the military rate which continued to rise; those rates more than doubled from 2001 to 2009 (Bachynski et al., 2012; S. A. Black et al., 2011), surpassing the civilian rate in 2008 (Armed Forces Health Surveillance Center [AFHSC], 2012a; S. A. Black, Gallaway, Bell, & Ritchie, 2011). These facts make it imperative to identify malleable points of intervention to detect and reduce suicide among Service Members.

It has been widely documented in the general population literature that there is a strong association between psychiatric diagnosis and death by suicide (Cavanagh, Carson, Sharpe, & Lawrie, 2003) and between psychiatric diagnosis and other suicidal behaviors, regardless of socio-demographic factors and country of residence (Nock et al., 2008a). Some authors even suggest that suicide decedents are exclusively individuals with mental illness (Joiner, 2005; Joiner, Buchman-Schmitt, & Chu, 2017; Nock et al., 2008b; Robins, Murphy, Wilkinson, Gassner, & Kayes, 1959). The same patterns can be found in the military as well. Mental health diagnosis has been shown to have the strongest association with suicidal risk, thoughts, behaviors, and deaths (Bachynski et al., 2012; LeardMann et al., 2013; Shen, Cunha, & Williams, 2016; Stein & Ursano, 2013). In the military, being male (LeardMann et al., 2013), lower rank (E1-E4), younger (Bachynski et al., 2012; S. A. Black et al., 2011) and with a history of self-harm (Shen et al., 2016) all increase the risk for suicide substantially. Besides these risk factors, numerous life stressors have been identified that also increase military suicide risk: relationship problems (AFHSC, 2012b; S. A. Black et al., 2011; Shen et al., 2016), history of deployment (S. A. Black et al., 2011; Shen et al., 2016), separation from the military, and history of law violations (Shen et al., 2016).

In spite of the strong association between psychiatric disorder and the risk for suicide death and suicide attempts (SAs), the epidemiology of psychiatric disorders among U.S. military suicide decedents is not well understood. The magnitude of psychiatric disorder within the U.S. military has been assessed through various methods with varying 12-month prevalence estimates reported by different studies. Gadermann and colleagues (2012) predicted 27% prevalence of any psychiatric disorder within the Army through simulation modeling. A Department of Defense (DoD) sponsored review of Army medical records found 25.4% prevalence of any psychiatric disorder when averaged across years 2005–2015 (Deployment Health Clinical Center [DHCC], 2017). In other studies, Riddle et al. (2007) found 18.3% prevalence of any psychiatric disorder among U.S. military Service Members from the Millennium Cohort for years 2001–2003 while Black and colleagues found 15.5% prevalence of any psychiatric disorder from medical record data in the years 2001–2009 among Army personnel (S. A. Black, Gallaway, Bell, & Ritchie, 2011).

There are indications that the prevalence of psychiatric morbidity in the Army increased substantially in the early 2000s evidenced by a near doubling of psychiatric ambulatory visits from 2000 to 2008 (Bachynski et al., 2012). This was consistent with findings by Hyman, Ireland, Frost, and Cottrell (2012) that prevalence of psychiatric diagnoses increased across the military in the period 2005–2007 along with increases in psychiatric medication prescriptions and mental health visits. Further, studies that sought to determine age of onset for psychiatric disorder among Army soldiers found a high proportion with onset prior to age of enlistment indicating the majority of soldiers with psychiatric disorders likely enter the military with an existing disorder. Based on national data weighted to reflect the sociodemographic profile of the Army, Gaderman et al. (2012) estimate that age of onset for psychiatric disorder precedes age of enlistment for more than 90% of soldiers with lifetime prevalence of psychiatric disorders. Similarly, Kessler et al. (2014) found nearly 80% of soldiers with current psychiatric disorders reported onset before enlistment. Taken together, these data suggest there is a significant and growing proportion of soldiers who enter the military with psychiatric disorder, increasing the risk over time for suicide behaviors within the Army.

Given these prevalence data, better insight into the distribution of mental health conditions among suicidal soldiers and soldiers who attempt or die by suicide could inform military suicide prevention efforts. Gibson, Corrigan, Kateley, Watkins, and Pecko (2017) found the yearly prevalence for any psychiatric diagnosis before the event was 59.3% for suicide deaths and 74.3% for SAs or non-fatal self-injury when averaged across years 2014–2016. Gibson et al. (2017) also reported that the most frequent DSM-IV diagnoses among soldiers who died by suicide during this time were for disorders of adjustment (40.3%), mood (34.3%), substance use (27.7%), anxiety-except PTSD (23%), PTSD (16%), personality (5%), and psychosis (2.3%). The most frequent diagnoses among soldiers who attempted suicide during this time were for disorders of adjustment (58.6%), mood (47.3%), anxiety-except PTSD (33%), substance use (26.7%), PTSD (19%), personality (7.3%), and psychosis (2%).

Considering mental health disorders as a strong predictor for death by suicide, previous research has identified a robust association between personality disorders and self-directed violence (American Psychiatric Association [APA], 2003). It has been estimated that about 30–40% of individuals in the general population who die by suicide and 40% of individuals who make a SA were diagnosed with personality disorder (APA, 2003; Duberstein & Conwell, 1997). Of those personality disorders, borderline personality disorder (BPD) presents one of the highest risks for suicide (Leichsenring, Leibing, Kruse, New, & Leweke, 2011; Oldham, 2006) – approximately 50 times higher than in general population (APA, 2001). In fact, it is one of only two disorders that explicitly includes suicidal behavior as a part of the diagnostic criterion (APA, 2013; D. W. Black, Blum, Pfohl, & Hale, 2004).

BPD is estimated to occur in 2% of the general population and most individuals who meet criteria for BPD are women (APA, 2001). BPD is mainly characterized by dysfunction of the emotion regulation system (Linehan, 1993) which is expressed by some key diagnostic criteria, such as a pervasive pattern of unstable interpersonal relationships, affect, and self-image, in addition to marked impulsivity (APA, 2001). The severity of these symptoms and especially impulsivity has been found to pose an increased risk for suicide and SAs (APA, 2003). Most individuals with borderline personality disorder (BPD) will attempt suicide (APA, 2001) and about 10% will eventually die by suicide (D. W. Black, Blum, Pfohl, & Hale, 2004). Additionally, results of meta-analysis focusing on self-injuries and SAs showed that BPD has a stronger association with SA than any other mental health disorder (Victor & Klonsky, 2014). The distinction between suicidal patients with BPD and without the diagnosis could be explained by negative affective states. Mou and colleagues (2018) found that these negative affective states, such as feelings of abandonment, desperation, guilt, hopelessness, loneliness, rage, self-hatred, and upset, were more strongly associated with suicidal ideation among suicidal patients diagnosed with BPD than among suicidal patients without BPD. It appears that some of the key characteristics of BPD diagnosis exacerbate patient’s suicidality. As BPD poses a high risk for suicidality independently, the presence of psychiatric comorbidity elevates this risk even more (D. W. Black et al., 2004). The suicide risk is particularly high with comorbid disorders such as depression, post-traumatic stress disorder (PTSD), and substance use disorders (APA, 2003; D. W. Black et al., 2004).

The diagnosis of BPD (and other personality disorders) does not result from a unitary set of symptoms. Some studies have identified meaningful subgroups or subtypes to be discerned among BPD (Hallquist & Pilkonis, 2012; Smits et al., 2017) and Hallquist and Pilkonis (2012) also showed that these subgroups can differ in severity of the diagnosis. The current classification system for personality disorders was introduced with the third edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III; APA, 1980) and has been the subject of considerable debate. Classification within the DSM is categorical; disorders are present or absent dependent on whether diagnostic criteria exceed a threshold number. Classification is also polythetic such that a subset of total criteria can be sufficient to meet diagnostic requirements. BPD is defined by nine diagnostic criteria of which any five or more may form the diagnosis among 256 possible combinations, while any two individuals diagnosed as BPD might share only one diagnostic criterion (Smits et al., 2017). In addition to heterogeneity of criteria within diagnostic categories, uncertain validity of criteria thresholds poses a problem for polythetic-categorical diagnoses (Krueger & Bezdjian, 2009; Morey & Skodol, 2013). Both issues were evident in a large clinical sample studied by Johansen, Karterud, Pedersen, Gude, and Falkum (2004). They found 136 combinations of five or more BPD criteria among 252 BPD patients, and among the total sample of 930 personality disorder patients there was no distinct threshold for number of BPD criteria between those with or without BPD.

Likewise, Karukivi, Vahlberg, Horjamo, Nevalainen, and Korkeila (2017) found that personality difficulties just below diagnostic thresholds in a clinical sample were significantly associated with marked psychiatric symptoms and negative effects on subjective well-being with impairments equal to actually diagnosed single personality disorders. These findings are supportive of a more dimensional view of personality disorders expressed in the alternative model for personality disorders appearing in the fifth edition of the DSM (DSM-5; APA, 2013; Oldham, 2015). Additionally, the count of criteria met seems to play a role in regards to suicidality. In his study of BPD inpatients who were followed from 10 to 23 years after discharge, Stone (1989) reported that patients with eight DSM-III criteria had a suicide rate of 36% as opposed to 7% for those who met five to seven criteria. Similar analyses have not been conducted in clinical samples of suicidal military personnel. Such data could illuminate the frequency of BPD diagnosis and distribution of diagnostic criteria among suicidal soldiers, as well as contribute insight into clinical significance of salient, albeit sub-threshold BPD criteria among suicidal personnel.

As BPD has not been widely studied among Service Members, little is known about the prevalence of BPD disorder or traits among military community and clinical samples. Individuals are required to undergo a review of medical screening forms and a general psychiatric evaluation (Cardona & Ritchie, 2007) which are to identify individuals with certain conditions that could be disqualifying for military accession (DoD, 2016). Current or history of BPD and history of suicide, including SAs, suicidal gestures, and self-mutilating behaviors, are exclusion criteria during the recruitment of military personnel (DoD, 2011). The crucial point here to note is that the screening solely relies on disclosures by the applicant to questions embedded in a series of medical history questionnaires that query for history of symptoms and treatment for psychiatric disorder. Self-disclosures of suicidal thoughts, behaviors, and attempts have been historically underreported, as individuals tend to conceal these facts due to fear of stigmatization and/or to avoid involuntary hospitalization (Chiurliza et al., 2018; Nock et al., 2013; Sudak, Maxim, & Carpenter, 2008). We can then speculate whether military applicants might be motivated to conceal their mental health history due to fear of being declined from the military. Few data are publicly available to evaluate the effectiveness of military pre-accession screening procedures. However, a review by the Institute of Medicine (Denning, Meisnere, & Warner, 2014) noted that non-validated procedures in use for pre-accession screening had limited effectiveness and various studies have found psychiatric disorder to be a leading cause of early attrition from military service including discharge during basic training (Hoge et al., 2002; Hoge et al., 2003; Niebuhr, Powers, Krauss, Cuda, & Johnson, 2006). These findings occur within a context that most psychiatric disorders among military members have an onset that predates their enlistment (Gaderman et al., 2012; Kessler et al., 2014), suggesting that service members diagnosed with BPD during military service likely entered the military with BPD or a co-occurring psychiatric disorder. Thus, evidence concerning the prevalence and scope of BPD symptoms among Service Members, including limited information concerning BPD as a risk factor for suicide among military personnel, is limited by these factors.

To date, there are two peer-reviewed articles examining BPD in active duty military. In his psychoanalytic perspective, Bird (1980) presents two case studies of army officers who experienced a “breakdown” one of which resulted in a SA. Despite Bird’s accurate descriptions of BPD symptomology in these officers, his conclusions that the Army can remedy those symptoms by standing in as a parent figure and as their only security have not been empirically investigated. Another study (Chu, Buchman-Schmitt, Joiner, & Rudd, 2017) examined personality disorder symptoms in active duty high-risk military members and found a puzzling subgroup of soldiers who present low desire and yet high plans for suicide. However, this subgroup was not found to have significantly greater symptoms of BPD, as the authors hypothesized. This study did not further explore the diagnosis of BPD in this active military sample. Thus, there is much to learn regarding the presence of BPD and its consequences in the military.

The aim of this paper is to bridge this gap in the literature. As part of a larger study concerning suicidal ideation and self-directed violence in the US Army, this secondary data analysis compared and contrasted active duty suicidal soldiers who did and did not meet DSM-IV criteria for BPD. This analysis sought to characterize the presence-absence of each BPD diagnostic criterion and lifetime history of self-directed violence in a suicidal military sample that included soldiers who did and did not meet BPD criteria. It was hypothesized that suicidal soldiers with more BPD criteria would be more likely to have engaged in self-directed violence, including lifetime counts of non-suicidal self-injury (NSSI) and attempted suicide, compared to suicidal peers with fewer BPD criteria.

Methods

This study is a secondary analysis of a randomized controlled trial conducted within the Division of Behavioral Health outpatient clinics at an Army installation in the southeastern United States. The larger study compared the Collaborative Assessment and Management of Suicidality (CAMS) with Treatment as Usual to reduce suicidal ideation and SAs over 12 months. The trial’s procedures and results are described elsewhere (Jobes et al., 2017). The current study examined only baseline data reflecting the period prior to randomization.

Participants

The full sample included 148 active duty soldiers, who met inclusion for the study if they were English speaking, at least 18 years of age, and had significant suicidal ideation; defined as a score of 13 or higher on the Scale for Suicidal Ideation-Current as per scale developer training and consultation of the instrument’s use (personal communication, Greg Brown, 2007). In line with effectiveness-oriented research, this study attempted to limit exclusion to those for whom participation in a voluntary research study was not possible or study treatment was not appropriate. Specifically, exclusion criteria were: (a) significant psychosis, cognitive, or other problems such that a participant could not understand the study procedures, could not provide informed consent, or for whom psychosocial therapeutic care was contra-indicated, (b) a judicial order to treatment such that participation would not be truly voluntary, or (c) separation, change of station, or deployment expected in the next twelve weeks (the projected duration of the treatment intervention). In addition, (d) soldiers in the Warriors in Transition unit and (e) pregnant women were also excluded, at the request of our military partners or military IRB. Given the focus of this study on soldiers who met criteria for BPD, soldiers who did not complete a structured diagnostic assessment for BPD (n = 11), and thus could not be categorized as having BPD or not, were excluded from the present analyses. Thus, this study’s sample includes 137 participants.

Measures

This study investigated the lifetime history of self-directed violence which was assessed by the Suicide Attempt Self-Injury Count. Additionally, we also examined the presence-absence of BPD diagnostic criteria via the Structured Clinical Interview for DSM-IV Axis II Borderline Personality Disorder.

The Scale for Suicide Ideation-Current (SSI-C; Beck, Brown, & Steer, 1997) is an interviewer-administered scale that measures a patient’s suicidal ideation at its worst point in the past two weeks. This measure has been found to be a valid and reliable measure of suicidal ideation for use with psychiatric patients (Cronbach’s alpha=.89; Beck, Brown, & Steer, 1997) and in the current study (Cronbach’s alpha=.74). It was used at screening to determine level of suicidal ideation criterion for inclusion in the study and as a baseline assessment.

The Suicide Attempt Self-Injury Count (SASI-Count) (Linehan & Comtois, 1996; Linehan, Comtois, Brown, Heard, & Wagner, 2006) is a brief interview covering past self-inflicted injuries categorizing them into SAs and NSSI. The tool also creates counts of self-inflicted injuries by method, medical risk severity, and lethality. It has a Lifetime form and a Recent form which covers a specific assessment period. The Lifetime and Recent version (for past year) were conducted at baseline. Follow-up assessments conducted the Recent version (for the period since previous assessment). Interviewer ratings are on the SASI-Count the same in the Suicide Attempt Self-Injury Interview which has shown strong reliability and validity (Linehan et al., 2006).

The Structured Clinical Interview for DSM-IV Axis II Borderline Personality Disorder (SCID-BPD; First, 1997) was used to identify patients with borderline personality disorder given the suicide risk associated with this disorder. The study research coordinator (RC) was trained to conduct the SCID-II via standardized SCID training videos and materials as well as through supervision by the last author (KAC) who was trained by an approved SCID trainer. Both the presence-absence of BPD diagnosis, and the number of symptoms were derived from the SCID and included in subsequent analyses.

The University of Washington Risk Assessment Protocol (UWRAP; Linehan, Comtois, & Ward-Ciesielski, 2012; Pearson, Stanley, King, & Fisher, 2001; Reynolds, Lindenboim, Comtois, Murray, & Linehan, 2006) was used to determine the participant’s stress level and suicidal intent before and after the research interview and suicide risk. The protocol guides the assessor to appropriate responses for different levels of suicide risk.

Procedures

Army behavioral health clinicians, clinic chiefs, and other medical staff identified soldiers who reported suicide ideation or were thought to be at suicide risk. An RC met with potential participants to conduct informed consent to the clinical trial and conduct the screening to determine willingness and criteria to participate. This was followed by a baseline assessment prior to randomization and a diagnostic interview that was conducted within one month of randomization and referred to 30 days prior to baseline.

Assessments were conducted in the RC’s office in the Army behavioral health (BH) clinic or on the military treatment facility inpatient psychiatric unit. By regulation, it was not possible to pay active duty Army personnel for their participation in the research assessments. All procedures were reviewed and approved by Department of Defense IRB and Human Research Protection Office (HRPO) as well as University of Washington and Catholic University of America institutional review boards.

Each of the instruments mentioned above was administered by a trained RC at the first timepoint in the study. This coordinator was given extensive training by a clinical psychologist, had interviews modeled for her by the clinical psychologist, and consulted extensively throughout the study with regard to differential diagnosis with the clinical psychologist and/or licensed social workers in the BH clinic in which the study was based. Shankman and colleagues (2017) asserted that training in SCID only is adequate for research labs that are seeking to assess psychopathology and other studies have shown inter-rater validity and reliability of SCID administration by non-clinicians (e.g., Birmaher et al., 2009; Goldstein et al., 2010; Hidalgo-Mazzei, Walsh, Rosenstein, & Zimmerman, 2015; Zimmerman, Ellison, Young, Chelminski, & Dalrymple, 2015). Prior to analysis the diagnostic data were further checked and cleaned by a research scientist who is also a licensed social worker.

Statistical Analyses

Descriptive statistics were calculated to characterize the socio-demographic and military career characteristics of participants, as well as participants’ history of self-directed violence, stratified by BPD diagnostic status. In addition, chi-squared tests of independence were conducted to test the association between BPD diagnostic status and participants’ socio-demographic, psychiatric, and other characteristics.

The association between BPD diagnosis and counts of NSSI was evaluated via a hurdle model, a form of count regression. Hurdle models include both a logistic regression portion that represents the association between predictors and the odds of any counts of the outcome, and a counts portion that represents the association between predictors and the number of counts among participants with one or more (i.e., any) counts of the dependent variable (Atkins, Baldwin, Zheng, Gallop, & Neighbors, 2013). Given the limited variability and extreme values of counts for attempted suicide in this sample, it was not possible to calculate a hurdle model or other count regression, and thus binary logistic regression was used to assess the association between BPD diagnosis and attempted suicide. Age and gender were included as planned covariates in all analyses, as both age and gender have been found to be associated with suicide-related outcomes (Nock et al., 2013). Alpha was set at 0.05 for all analyses.

To visualize patterns of BPD symptoms for subgroups of participants with and without past history of self-directed violence, a quilt plot (Wand, Iversen, Law, & Maher, 2014) was created using conditional formatting in Microsoft Excel. Quilt plots are simple graphical visualization that can represent categorical data in a way that may be more readily interpretable than conventional reporting of frequency distributions or other numerical data, but have not been widely implemented outside of a few specific disciplines (Wand et al., 2014). As noted by Wand and colleagues (2014), quilt plots “are used for data presentation… and p-values are not generated,” and thus the information in our quilt plot is offered to provide rich description of participants’ clinical characteristics, and no inferential statistics were calculated.

Results

Approximately one-quarter (n = 38, 27.7%, 95% CI = 20.2% to 35.2%) of the suicidal soldiers in this active duty sample met full diagnostic criteria for BPD. As is reported in Table 1, participants with BPD were on average younger than those without BPD, and less likely to have ever been married. Participants with BPD also reported lower military rank compared to those without BPD, which was consistent with the BPD subsample being younger. All psychiatric comorbidities for participants in this sample are reported in Table 2. Participants’ BPD diagnostic status (i.e., presence-absence) was associated with any current (i.e., past month) psychiatric comorbidities (χ²(1)=10.92, p=.001), such that participants with BPD were more likely to meet criteria for more than one disorder. However, there was not a statistically significant association between BPD diagnosis and lifetime psychiatric comorbidity (χ²(1)=3.73, p=.053), although this association approached significance, and indicated a similar pattern to current comorbidity status.

Table 1.

Baseline Sociodemographic Characteristics: Overall and by BPD Diagnosis

	BPD	No-BPD	Overall
Sociodemographics	n(%)	n(%)	N(%)	pd
Age in years, M(SD)	24.63(5.0)	27.6(6.0)	26.8(5.9)	.007
Male gender	29(76.3)	85(85.9)	114(83.2)	.205
Ethnicity a
White/Caucasian	23(60.5)	47(47.5)	70(51.1)
Black or African American	9(23.7)	23(23.2)	32(23.4)
Puerto Rican	1(2.6)	8(8.1)	9(6.6)
Other	5(13.1)	16(16.1)	21(15.4)	.532
Sexual Orientation b
Heterosexual	29(76.3)	84(84.8)	113(82.5)
Bisexual	7(18.4)	9(9.0)	16(11.6)
Homosexual	1(2.6)	2(2.0)	3(2.2)	.314
Marital Status c
Married	13(34.2)	53(53.5)	66(48.2)
Single, never married	19(50.0)	18(18.2)	37(27.0)
Separated	4(10.5)	20(20.2)	24(17.5)
Divorced	0	7(7.1)	7(5.1)
Widowed	0	1(1.0)	1(7)	<.001e
Number of children, M(SD)	.77(1.2)	17(15)	1.45(1.5)	.001
Education
High school or GED	20(52.6)	37(37.4)	57(41.5)
Some college, associate’s degree, or technical training	15(39.5)	52(52.5)	67(48.9)
Bachelor’s or graduate degree	2(5.2)	9(9.1)	11(8.0)	.222
Rank
Junior enlisted (E1-E4)	36(94.6)	62(62.5)	98(71.5)	<.001f
Noncommissioned officer (E5–E9)	2(5.3)	33(33.3)	35(25.6)
Officer (W2–O3)	0	4(4.0)	4(2.9)
Number of combat deployments
0	25(65.8)	33(33.3)	58(42.3)
1	9(23.7)	24(24.2)	33(24.1)
2	3(7.9)	23(23.2)	26(19.0)
3 or more	1(2.6)	18(18.l)	19(13.9)	.002

Note. 148 participants completed baseline assessments but only 137 have completed SCID-BPD to be assessed for the BPD diagnosis.

^aFive participants in the no-BPD diagnosis group refused to answer this question.

^bOne participant in the BPD diagnosis group answered “I don’t know”; four participants in the no-BPD group refused to answer this question.

^cTwo participants in the BPD diagnosis group refused to answer this question.

^dParticipant characteristics were compared via independent samples t-tests for continuous variables, and chi-squared tests of independence for nominal and categorical variables.

^eDivorced and widowed participants were combined with separated participants in the No-BPD category given the counts of zero for these categories among participants with BPD.

^fOfficers in the No-BPD category were excluded from this analysis given the absence of any participants with BPD in this category.

Table 2.

Baseline Lifetime and Current Prevalence of Comorbid Diagnoses: Overall and by PBD Diagnosis

Comorbid Diagnosis	BPD		No-BPD		Overall
	n(%)		n(%)		N(%)
	Lifetime	Current	Lifetime	Current	Lifetime	Current
Bipolar I disorder	2(5.3)	1(2.6)	1(10)	1(10)	3(2.2)	2(1.5)
Bipolar II disorder	1(2.6)	1(2.6)	2(2.0)	2(2.0)	3(2.2)	3(2.2)
MDD	28(73.7)	20(52.6)	82(82.8)	62(62.6)	110(80.3)	82(69.9)
Dysthymic disorder	3(7.9)	-	4(4.0)	-	7(5.1)	-
Schizophrenia	0	0	0	0	0	0
Alcohol disorder	22(57.9)	7(18.4)	46(46.5)	12(12.1)	68(49.6)	19(13.9)
Alcohol abuse	6(15.8)	2(5.3)	9(9.1)	2(2.0)	15(10.9)	4(2.9)
Drug dependence	10(26.3)	2(5.3)	16(16.2)	3(3.0)	26(19.0)	5(3.6)
Drug abuse	8(21.1)	0	23(23.2)	1(10)	31(22.6)	1(7)
Panic disorder	20(52.6)	12(31.6)	41(41.4)	27(27.3)	61(44.5)	39(28.5)
Social phobia	7(18.4)	6(15.8)	13(13.1)	12(12.1)	20(14.6)	18(13.1)
OCD	2(5.3)	2(5.3)	4(4.0)	4(4.0)	6(4.4)	6(4.4)
PTSD	22(57.9)	17(44.7)	62(62.6)	52(52.5)	84(61.3)	69(50.4)
GAD	-	11(28.9)	-	15(15.2)	-	26(19.0)
Acute stress disorder	-	1(2.6)	-	0	-	1(7)
Adjustment disorder	-	3(7.9)	-	14(14.1)	-	17(12.4)
Any comorbidity	37(97.4)	35(92.1)	85(85.9)	63(63.6)	122(89.1)	98(71.5)

Note. Current prevalence characterized by meeting the criteria in the past month.

Detailed information concerning participants’ past history of self-directed violence, stratified by diagnostic status, is reported in Table 3. Half of the participants reported at least one prior SA, and forty-one percent (n = 57) reported at least one episode of NSSI.

Table 3.

Baseline History of Lifetime Self-Harming Behaviors: Overall and by BPD Diagnosis

	BPD		No-BPD		Overall
	n(%)		n(%)		N(%)
Lifetime history of
Any self-inflicted injury	28(73.7)		56(56.6)		84(61.3)
NSSI	25(65.8)		32(32.3)		57(41.6)
Attempted suicide	23(60.5)		45(45.5)		68(49.6)
Method
Cutting	20(52.6)		24(24.2)		44(32.1)
Overdose	12(31.6)		12(12.1)		24(17.5)
Burning	8(21.1)		13(13.1)		21(15.3)
Hanging or strangling	10(26.3)		10(10.1)		20(14.6)
Jumping from a high place	6(15.8)		4(4.0)		10(7.3)
Firearm	0		1(10)		1(.7)
Poisoning	3(7.9)		0		3(2.2)
Asphyxiation	4(10.4)		5(5.1)		9(6.6)
Drowning	4(10.5)		2(2.0)		6(4.4)
Stabbing	8(21.1)		3(3.0)		11(8.0)
Head banging	12(31.6)		18(18.2)		30(21.9)
Other	14(36.8)		22(22.2)		36(26.3)
	M(SD) Mdn(IQR)		M(SD) Mdn(IQR)		M(SD) Mdn(IQR)
Lifetime count of
Self-inflicted injury acts	191.7(444.0)	12.5(81.3)	26.7(77.8)	1.0(4.0)	72.8(252.8)	1.0(16.0)
NSSI episodes	184.9(435.5)	6.0(81.0)	22.0(67.1)	.0(2.3)	67.5(246.2)	.0(11.5)
Suicide attempts	6.8(16.8)	1.0(33)	4.7(26.5)	.0(1.0)	5.3(24.1)	.0(2.0)
Age at
First self-inflicted injury	13.9(5.9)	13.0(4.0)	18.9(7.5)	19.0(11.0)	17.3(7.4)	14.5(9.0)
Most severe self-inflicted injury	17.8(5.7)	16(6.0)	21.6(6.6)	21.0(9.0)	20.3(6.5)	19.5(9.0)
Most recent self-inflicted injury	21.3(4.8)	20.0(5.0)	23.5(6.7)	22.0(9.0)	22.8(6.2)	22.0(7.0)
Lifetime lethality of self-inflicted injury a
Highest	4.0(1.5)	4.0(2.8)	3.6(1.6)	3.0(3.0)	3.7(1.6)	3.0(3.0)
Lowest	2.1(1.0)	2.0(.8)	2.6(1.6)	2.0(2.5)	2.4(1.4)	2.0(2.0)

^aNote. Lethality was scored on 1 (very low) to 6 (severe) scale.

Patterns of BPD criteria presence-absence for different subgroups of participants with and without histories of NSSI and attempted suicide are presented in Figure 1. The quilt plot in Figure 1 demonstrates greater presence of particular criteria with darker shades of blue, which allows one to observe patterns across clinical subgroups, including participants with and without past history of attempted suicide or NSSI. The presence of BPD criteria was generally consistent across subgroups, except for the marked difference in criterion 5, which reflected the nature of the subgroups included. For participants with past history of NSSI or attempted suicide, efforts to avoid abandonment were somewhat more common among those with BPD. The reverse of this pattern was found among those who did not meet BPD criteria: efforts to avoid abandonment were less common among those with past NSSI. There was also a higher proportion of participants who met the impulsivity criterion among those with past history of NSSI or attempted suicide, though chronic feelings of emptiness were less common among participants who previously engaged in self-directed violence. The total proportion of participants who met each of the BPD criteria are also presented, and indicate that soldiers without BPD reported non-trivial experiences of BPD criteria, including near one-third who met the impulsivity and/or paranoid ideation or severe dissociation criteria.

Figure 1.

Quilt plot with proportion of participant subgroups meeting BPD criteria.

To elucidate the patterns of BPD criteria counts among soldiers who did not meet the threshold for a diagnosis of BPD, criteria counts were graphed and are presented in Figure 2. As Figure 2 demonstrates, relatively few particpants were subthreshold with three or four criteria. This suggests that the distinction between those with and without a BPD diagnosis is not confounded by a substantial number of subthreshold or marginal cases of BPD criteria that would be potentially prone to misclassification.

Figure 2.

Distribution of total BPD symptom count, stratified by diagnostic status

In a hurdle model (see Table 4) calculated to test the association between BPD criteria count (i.e., total number of diagnostic criteria) and lifetime counts of NSSI, BPD diagnosis was a significant predictor of the odds of reporting any NSSI (Logistic regression portion, OR=1.23, p=.005), above and beyond the effects of age (0R=0.92, p=.024) and gender (0R=0.61, p=.300). Thus, a one criterion increase in BPD symptoms was associated with a 23 percent increase in the odds of having any lifetime NSSI. The number of BPD criteria also predicted greater lifetime counts of NSSI (Counts portion, 0R=1.36, p=.026), above and beyond age (0R=0.98, p=.834) and gender (0R=85, p=.828). This indicated that there was a 36 percent increase in the number of counts of NSSI for each one criterion increase in BPD symptoms.

Table 4.

Hurdle Model Analyses

Variable	Logistic regression portion			Counts portion
	beta(SE)	OR (95% CI)	p	beta(SE)	RR (95% CI)	p
Dependent variable: lifetime count of NSSI
Age	−.08(.04)	.92 (.86 to .99)	.024	−.02(.08)	.98 (.85 to 1.14)	.834
Gender	−.50(.48)	.61 (.23 to 1.56)	.300	−.16(.74)	.85 (.20 to 3.61)	.828
BPD criteria count	.20(.07)	1.23 (1.06 to 1.42)	.005	.31(14)	1.36 (1.04 to 1.78)	.026
Dependent variable: lifetime count of suicide attemptsa
Age	−0.07(.03)	.93 (.87 to .99)	.029	-	-	-
Gender	−0.13(.47)	.88 (.35 to 2.20)	.781	-	-	-
BPD criteria count	0.10(.07)	1.10 (.96 to 1.26)	.156	-	-	-

Note. SE = standard error or unstandardized beta coefficients; OR = odds ratio; RR = rate ratio.

^aOnly binary logistic regression was employed for this dependent variable.

Binary logistic regression was used to test the association between BPD criteria count and odds of any SA. There was not significant association between BPD criteria and ever attempting suicide (0R=1.10, p=.156), above and beyond the influence of age (0R=0.93, p=.029) and gender (0R=0.88, p=.781).

Discussion

This study demonstrated a large proportion of suicidal soldiers met diagnostic criteria for BPD, which was surprising given the low prevalence of any personality disorder in military medical records of soldiers who attempt or die by suicide. Further, our findings report in-detail the socio-demographic, clinical, and self-directed violence-related characteristics of suicidal soldiers with and without BPD. Consistent with the interpersonal and professional difficulties associated with BPD, suicidal soldiers with BPD were less likely to be married, reported lower military rank, and fewer deployments. These factors may reflect the impact of BPD symptoms, but may also be reflective of the younger age of soldiers with BPD. Also consistent with defined symptoms of BPD, the soldiers’ number of BPD symptoms was associated with increased odds of NSSI, and a higher number of counts of NSSI, if any. There was not a significant effect of BPD symptoms upon the odds of attempting suicide, which may indicate that this association is not as strong as the association between BPD symptoms and NSSI in this population. However, it may also be due to the relatively modest sample size included in this study, and the relatively low base rate of suicidal behavior.

The high prevalence of BPD found in this study was derived using a broadband clinical assessment of DSM Axis I and II disorder, including comprehensive assessment of each BPD symptom. This practice may not represent the typical approach in military clinics. Since personality disorder diagnosis can be the basis for an involuntary non-medical separation (DoD, 2016), providers may refrain from giving a DSM IV Axis II diagnoses when an Axis I diagnosis is given. This would be consistent with the high co-morbidity of BPD and other diagnoses within this sample. Additional study is needed to understand the extent to which our finding is representative of suicidal Service Members or Service Members in general.

The true prevalence of BPD among suicidal soldiers is relevant for planning effective treatment interventions. BPD is refractory to many psychological treatments which are often ineffective for mitigating suicide risk among BPD patients (e.g., Linehan, 1993) but can benefit from a range of evidence-based practices such as Dialectical Behavior Therapy and Mentalization Based Therapy. The high number of soldiers who attempt or die by suicide after receiving psychological treatment from the Army is an important quality of care issue. There is limited information about diagnostic and clinical assessments by Army providers of suicidal soldiers or their treatment outcomes from psychological interventions that precede SAs or deaths. The association of self-injury and BPD diagnosis in this sample supports the benefit of routine structured and comprehensive assessment for diagnosis of a personality disorder or diagnostic traits among suicidal soldiers who present for medical treatment. Under-diagnosis of BPD or clinically-significant BPD criteria could result in untreated maintaining factors for durable patterns of self-directed violence that are consistent with BPD symptoms. This may be particularly important among soldiers with elevated suicidality who receive BH care.

Further study of the prevalence of covert (i.e., undisclosed and/or untreated) self-directed violence among suicidal soldiers receiving BH treatment and potential missed opportunities for more effective suicide risk mitigation are suggested by these findings. The significant association between BPD diagnosis and NSSI is important to consider as a potential contributor to long-term risk of other forms of self-directed violence (e.g., Klonsky, May, & Glenn, 2013). While these findings did not indicate a significant association between BPD diagnosis and any history of attempted suicide, there could be military unique factors at play. Military pre-accession screening does not directly evaluate applicants for personality disorder, but it does screen for history of SAs. Given that approximately 80% of suicide attempters have one or more psychiatric disorders (Nock, Hwang, Sampson, & Kessler, 2009), applicants who are screened out for SA likely decrease the prevalence of existing or prodromal BPD by some degree and perhaps selects out those with higher severity BPD. Also, there may be military unique protective factors.

In addition to suicidal thoughts and behaviors, hallmarks of BPD include emotional dysregulation and unstable sense of identity. The military provides considerable lifestyle structure that externally organizes daily activities and longer-term planning. The military also has strong markers of identity and interpersonal boundaries in the form of rank, duty position, and unit affiliation that are reinforced by wear of the uniform, social traditions, and military customs and courtesies. For some persons with BPD, aspects of the military such as these may provide stabilizing counterbalance to vulnerabilities inherent to BPD that otherwise could manifest in suicidal ideation and urge for self-harm.

Clinically, military treatment providers may want to more directly assess and treat BPD to reduce suicide risk, improve occupational functioning and reduce subjective distress of soldiers with BPD. This might be done by the implementation of evidence-based BPD-specific interventions such as Dialectical Behavior Therapy, Mentalization Based Therapy or another well-structured long-term psychotherapy which have been shown to be effective in reducing suicide risk (Cristea et al., 2017; DeCou, Comtois, & Landes, 2018). When a full evidence-based model is not feasible, key principles from these therapies can be used to assess, manage, and resolve suicidality (Carmel, Sung, Jeff, & Comtois, Kate, 2018). For instance, the Biosocial Theory of BPD (Linehan, 1993) suggests that BPD is the result of a series of transactions over the individual’s history between a biological vulnerability to emotional sensitivity and reactivity and an environment that is pervasively invalidating and oversimplifies difficulty. As Linehan proposes, treatment providers of BPD patients should focus on providing skills in emotion regulation and distress tolerance to address biological vulnerability, and provide validation of patient’s emotions and thoughts and teach interpersonal effectiveness skills to mitigate the negative impact of invalidating environments.

Limitations

These findings are limited by the cross-sectional nature of the data, which do not support causal inference. This is balanced by the methods undertaken to ensure standardized interview-based assessments using valid and reliable instruments. Further, our findings may not be generalizable to other samples of military personnel who did not experience suicidality, or suicidal personnel who did not present for treatment services. Additionally, although the study assessor went through rigorous training in administering the structured interviews, they were not a licensed clinician, which could have potentially impeded the accuracy of the diagnoses. Given that this was a secondary analysis project, our analyses are not powered to detect attempted suicide, which is a low base rate behavior. Finally, this study did not employ probabilistic sampling, and thus does not offer an estimate as to any true prevalence of BPD among soldiers in general, or suicidal soldiers in particular. Nonetheless, the relatively large proportion of soldiers with BPD noted suggests the need for such studies in the future.

Conclusion

Although BPD and other personality disorders are not widely observed in existing epidemiological data derived from military records, this study demonstrated that a non-trivial proportion of suicidal soldiers meet criteria for this condition, which is strongly associated with self-directed violence and suicide. Given the association between BPD symptoms and suicide-specific outcomes in this sample, it is important to cultivate prevention strategies and treatment options that acknowledge and rigorously assess for the presence-absence of BPD criteria among suicidal military personnel.