The Desire to be Dead Among Bhutanese Refugees Resettled in the United States: Assessing Risk

Abstract

Bhutanese refugees are at elevated risk for suicide, but culturally inflexible suicide risk models may hinder accurate risk detection in this population. This cross-sectional study aimed to use a theoretical model based on the interpersonal psychological theory of suicide and the cultural model of suicide to improve suicide risk-assessment among Bhutanese refugees. Participants included 60 Bhutanese refugees (31 males and 29 females), aged 18 to 65, resettled in Vermont. Suicidal ideation (n = 4, 6.7%) was low, but a substantial minority (n = 29, 48.3%) endorsed some desire to be dead. Perceived burdensomeness, but not thwarted belongingness, was significantly associated with both suicidal ideation and the desire to be dead. Neither desire to be dead nor suicidal ideation was significantly related to suicide attempt history. Results highlight that including assessments of desire to be dead in addition to assessments of suicidal ideation may address the problem of under-identification of Bhutanese refugees at risk for suicidal behavior, particularly those who do not present with suicidal ideation.

Public Policy Relevance

Suicide risk assessments that exclusively rely on the presence of suicidal ideation are likely to underestimate true levels of risk among Bhutanese refugees. Incorporating the abstract desire to be dead into risk assessments with Bhutanese refugees may provide a more culturally-responsive alternative approach to suicide risk assessment.

Resettled Bhutanese refugees are twice as likely to die by suicide than the general U.S. population, and the age-adjusted suicide mortality rate of 24.4/100,000 individuals signals a significant public health crisis (Ao et al., 2012). Since 2008, over 95,000 Bhutanese refugees have been resettled in the United states [Worldwide Refugee Admissions Processing System (WRAPS), 2018)]. Effective proximal suicide risk assessments are needed to address this crisis. However, given that suicidal ideation is less common in certain non-Western-European populations (Reviewed in Chu, Goldblum, Floyd, & Bongar, 2010), assessing suicide risk among refugee populations can prove difficult. Contributing to this phenomenon of “hidden ideation,” a term used to refer to individuals who exhibit suicide risk, but do not present with concrete suicidal thoughts (Morrison & Downey, 2000), suicidal ideation is heavily stigmatized among Bhutanese refugees (Adhikari, Yotebieng, Acharya, & Kirsch, 2015; Ao et al., 2012; Brown et al., 2019). Psychological autopsies on 14 of 16 resettled Bhutanese refugees who died by suicide between 2009 and 2012 revealed that 46% of respondents (friends and/or families of the deceased) were unable to recall any signs that suicidal behavior was imminent, suggesting that a substantial majority of those who died by suicide displayed no warning signs even to those who knew them well (Schininà, Sharma, Gorbacheva, & Mishra, 2011). Moreover, prevalence estimates of suicidal ideation within the Bhutanese refugee population are dramatically lower (e.g., 3–6%; Adhikari et al., 2015; Ao et al., 2012) than might be expected given the extremely high rate of suicide. The alarming juxtaposition of a suicide mortality rate that is nearly twice that of the general U.S. population (National Center for Health Statistics, 2019), and a prevalence rate of suicidal ideation roughly one half or less that of the general U.S. population (Kessler, Borges, & Walters, 1999; Nock et al., 2008) underscores the need to improve suicide risk detection in Bhutanese refugees.

In recent years, the field has moved towards generating culturally-responsive models of suicide, with the goal of improving risk assessment, upstream and downstream intervention development, and to better understand the course and progression of suicide in and across individual cultural contexts. Joyce Chu and colleagues (2010) developed the first and only – to the authors’ knowledge – novel theoretical model of suicide that was specifically designed to incorporate culturally-relevant contributors to suicide risk. Their model, the cultural model of suicide, advanced our understanding of the factors contributing to suicide in minority communities (Chu et al., 2017; Chu, Lin, Akutsu, Joshi, & Yang, 2018; Yang et al., 2018). Chu and colleagues’ model incorporates four key contributors to suicide risk: (a) cultural sanctions, (b) idioms of distress, (c) minority stress, and (c) social discord. According to the model, culture moderates the presentation and expression of suicide related thoughts and behaviors and determines which stressors, risk factors, and protective factors are salient for an individual. Finally, cultural factors determine the meaning individuals extract from relevant stressors. According to Chu et al. (2010), as risk and protective factors accumulate and are filtered through a cultural lens, individuals either approach or recede from a tolerance threshold below which they can tolerate stress, and above which stressors become intolerable. If individuals’ stressors exceed the tolerance threshold, they may engage in suicidal behavior.

Thomas Joiner’s interpersonal psychological theory of suicide (IPTS; Joiner, 2005) is another model of suicide that has shown cross-cultural relevance. Joiner’s model integrates three primary components: (a) thwarted belongingness, (b) perceived burdensomeness, and (c) acquired capability of suicide. According to Joiner’s model, perceived burdensomeness and thwarted belongingness both contribute to the development of a concrete or abstract desire to be dead. When this desire for death is combined with the acquired capability of suicide, the IPTS posits that a suicide attempt is likely. Previously, IPTS constructs such as perceived burdensomeness and thwarted belongingness were shown to be related to the presence of suicide-related thoughts among Bhutanese refugees (Ellis et al., 2015). Bhutanese refugees who endorsed concrete suicidal ideation, expressed higher levels of perceived burdensomeness, and thwarted belongingness than did Bhutanese refugees who did not express concrete suicidal thoughts (Ellis et al., 2015).

By leveraging the lessons learned from these two robust models of suicide, as well as the vast existing literature centered on Bhutanese refugees (reviewed in Meyerhoff, Rohan, & Fondacaro, 2018), we can improve risk detection among Bhutanese refugees. We hypothesized an integration of the cultural model of suicide and the IPTS for Bhutanese refugees (Meyerhoff et al., 2018). Building on the success of the IPTS model among resettled Bhutanese refugees (Ellis et al., 2015), we identified a construct initially proposed by Joiner (2005): The wish to be dead or suicidal desire, defined as the abstract desire to be dead or to not exist. In prior research, however, this construct was overlooked and interpreted narrowly as suicidal ideation, or specific thoughts about death or suicide. However, the expansive definition of suicidal desire remains true to Joiner’s model, which has already shown its utility within the Bhutanese refugee community. Prior to settling on the desire of death as our construct of interest, we collaborated with community leaders to gain their insights into the problem of suicide in their own community and ensure the proposed constructs were face valid and matched the community’s identified goals. From the project’s initial conceptualization through the completion of data collection, we collaborated and partnered with community members in both formal (via a community advisory board) and informal (via existing relationships with community leaders and organizations) to ensure our project was culturally-responsive and addressed the specific needs and concerns of the local Bhutanese refugee community. Our model explicitly proposed that, within the Bhutanese refugee population, including the abstract desire to be dead in addition to a suicidal ideation would generate more robust risk detection than suicidal ideation alone (Meyerhoff et al., 2018). The present study was designed to use a model-based framework to improve risk-assessment among a sample of Bhutanese refugees. This pilot study represents the first step towards developing more responsive suicide risk assessments for this population.

Methods

Participants and Procedures

A total of 60 adult Bhutanese refugees were recruited from the greater Burlington, Vermont region. Participants were eligible for inclusion if they held refugee status from Bhutan, were adults between the age of 18 and 65, and had the ability to provide informed consent to all study procedures.

In executing this study, we used Community-Based Participatory Research (CBPR) principles and involved key stakeholders in the design and implementation of the research protocol. All study procedures were reviewed and approved by the University of Vermont’s Institutional Review Board (IRB). All participants who indicated interest in the study and met inclusion criteria were asked to provide written and/or verbal informed consent. Research staff ensured that all study procedures were understood, and informed consent was obtained from all participants prior to participation. Study interviews took place at a local community center or at a local community mental health clinic, both of which serve refugees. Following the interviews, all participants were provided with brief psychoeducation about suicide prevention resources. Each of the 30 participants who endorsed any suicide-related construct (i.e., non-zero scores on either the Beck Scale for Suicidal Ideation or the Wish to be Dead Scale, or other suicide-related measure) was given a brief overview of strategies for reducing acute distress (e.g., grounding exercises) and a personalized safety plan that incorporated distress tolerance skills and behavioral activation. All participants also received a $20 gift card to a local grocery store/restaurant.

Translation of Measures

Whenever possible, we relied on measures that had been previously translated and used with this population. For unavailable measures, we translated and modified items based on procedures used in other work with refugee populations (Beaton, Bombardier, Guillemin, & Ferraz, 2000; Silva et al., 2018). Colloquialisms were removed from English measures by the research staff. A professional interpreter fluent in both English and Nepali then conducted forward translations from English to Nepali. Two community members fluent in both English and Nepali individually back translated measures. Then, two translators met with research staff and discussed each item and any discrepancies until a consensus was reached. After a consensus was reached, a translator made grammatical corrections to the Nepali text and one back translator verified that the translated text matched the original intent of the English question. We did not encounter any systematic issues in the translation process, the forward and backwards translation and subsequent consensus and correction approach, outlined above, was used for each of the measures for which a Nepali translation did not previously exist. In the case of questions about suicide and the wish to be dead, we also did not encounter any systematic challenges in translating these constructs. Although a full measurement invariance study is warranted in the future, these constructs were face valid to our consensus translators, community advisory board, and community partners. This process ensured newly translated measures contained culturally face valid item phrasing (e.g., idiomatic phrases such as “feeling blue” were reworded to “feeling sad” in order to access the intended construct in populations less familiar with culturally-bound idioms and phrases).

Use of Interpreters

Given that the majority of participants were not fluent in English, and unless English fluency had been previously established, an interpreter fluent in both Nepali and English was present at each study session to aid in the administration of translated measures. Interpreters for this project worked within medical and psychological settings and completed Health Information Portability and Accountability Act (HIPAA) and IRB human subjects training. Additional details about the use of interpreters can be found in previously published reports (Meyerhoff et al., 2018).

Measures

Demographics.

Participants provided demographic information such as marital status, income, primary language, refugee status, number of children, and number of people in their household.

Desire to be dead.

To assess the desire to be dead, we administered the Wish to be Dead Scale (WDS; Lester, 2013). The WDS is a 10-item measure with a Likert-scale or dichotomous True/False format. Both perform similarly (Lester, 2013). In this study, we elected to use the dichotomous format. Participants were asked to respond to questions on this measure while thinking about how they felt currently. The measure has been shown to have good construct validity and is associated with, but distinct from, suicide ideation (Dadfar, Lester, & Vahid, 2016; Lester, 2013). The WDS has been used in multicultural contexts and retains adequate to good reliability and internal consistency. In the current study, internal consistency on the WDS was good (α=.858).

Suicidal ideation.

To assess suicidal ideation, we administered the Beck Scale for Suicidal ideation (BSS; Beck & Steer, 1991) and extracted a 4-item subscale referred to as the ideation intensity subscale (Beck & Steer, 1991). Despite extracting a single subscale from this measure, the full 19 core-item scale was administered in this study. Each of the items asks a participant to choose one of three responses that best describes how they have been feeling in a given domain over the past week (including today). Possible descriptors are presented in ascending order of severity: 0 describes a complete absence of the suicide-related symptom, 1 describes the presence of a suicide-related symptom at a moderate level, and 2 describes the presence of a suicide-related symptom at a strong level. In a traditional administration of this measure, 5 screening items are presented to assess for the presence of broad suicidal ideation, and if responses to each of these 5 items is 0, administration typically stops, as it is assumed that no suicidal thoughts or behavior are present. However, explicit endorsement of suicidal ideation is often underestimated among Bhutanese refugees (Adhikari et al., 2015). Thus, in order to capture the breadth of base-rate suicide-related symptoms in this study, all 19 core items were administered to participants. Internal consistency for the full BSS scale was adequate (α=.785), while internal consistency for the Ideation Intensity subscale was unacceptable (α=.376), likely the result of low rates of response.

Suicide attempt history.

We collected suicide attempt history data on 53 of 60 individuals. Two questions focused on (a) the number of prior attempts and (b) the intent to die during the last attempt (BSS items 20 and 21) were initially left out of our early questionnaire administration sessions but were added to our protocol for the final 24 participants. Previous participants were re-contacted by phone in an effort to obtain this data. We were unable to reach or obtain consent to ask these two questions from seven individuals.

Perceived burdensomeness and thwarted belongingness.

To capture constructs of perceived burdensomeness and thwarted belongingness, we asked participants all items from the 15-item Interpersonal Needs Questionnaire (INQ; Van Orden, Cukrowicz, Witte, & Joiner, 2012). The INQ consists of 9 statements that measure perceived burdensomeness and 6 statements that measure thwarted belongingness. Statements are rated on a 7-point Likert scale with anchors at 0 (not at all true for me) to 6 (very true for me). The 15-item version of the INQ has demonstrated content, convergent, and divergent validity on the perceived burdensomeness subscale, and content and convergent validity were demonstrated for the thwarted belongingness subscale in clinical, analogous, and community samples (Van Orden et al., 2012). A brief 12-item version of this measure has been used to assess these constructs in Bhutanese refugees resettled in the United States (Adhikari et al., 2015; Ao et al., 2012), with good validity and reliability (Ellis et al., 2015). In the present study, internal consistency for the perceived burdensomeness (α=.702) and thwarted belongingness (α=.728) subscales was acceptable.

Data Analytic Plan

All analyses were conducted using R version 3.4.3 (R Core Team, 2017) with the following packages: car (Fox & Weisberg, 2017), countreg (Zeileis & Kleiber, 2018), ggplot2 (Wickham & Chang, 2016), MASS (Ripley, 2017), plyr (Wickham, 2016), pscl (Jackman, Tahk, Zeileis, Maimone, & Fearon, 2017), and psych (Revelle, 2018). Hierarchical logistic regressions were conducted using IBM/SPSS Statistics version 24 (IBM Corp., 2016).

Because our data reflected typical patterns in the rarity of suicidal ideation and suicidal desire, resulting in positively skewed distributions with an over-representation of zero ideation, suicidal behavior, and/or desire to be dead we employed a series of hurdle models as our primary data analytic approach. Hurdle models are two-part models used most frequently with count data and are useful in situations where data is positively skewed, has a lower bound of zero, and a significant over-representation of zeros (Atkins, Baldwin, Zheng, Gallop, & Neighbors, 2013). The principle assumption underlying hurdle models is that zero and non-zero data are generated from two different processes. In this case, individuals who have experienced no suicidal ideation, desire, or behavior are qualitatively different from individuals who report non-zero levels of suicidal ideation, desire, and behavior. This captures an underlying assumption of our data, namely that individuals who are at risk for suicide exhibit at least some non-zero suicidal ideation, desire, or behavior. When running a hurdle model, data are automatically split into two forms to reflect the two different processes by which data are assumed to be generated. The first form is a dichotomized dataset in which the dependent variable is collapsed into either zeros (absent) and ones (present) so that the first of the analyses comprising the hurdle model approach can be performed: a logistic regression. The second form is a zero-truncated dataset consisting only of participants with the dependent variable present at a non-zero level, which spans the full range of non-zero values. A regression analysis, is then performed on this zero-truncated dataset to determine the extent to which degree of ideation, desire, and behavior is associated with any given predictor or outcome. Hurdle models that leverage logistic and linear regression models do not account for specific measurement error and are useful analytic tools for preliminary cross-sectional studies, such as this one, to examine associations among measured constructs.

We followed the data analytic methodology used in previous studies of suicide-related data (Jobes et al., 2017) and used hurdle models to test the following hypotheses: (a) that higher perceived burdensomeness and thwarted belongingness will each independently predict desire to be dead and suicide ideation (zero vs. non-zero and zero-truncated levels); (b) that the interaction of perceived burdensomeness and thwarted belongingness would predict both desire to be dead and suicidal ideation (zero vs. non-zero and zero-truncated levels).

Results

Sample Characteristics

Our full sample consisted of 60 eligible participants. We present basic demographic characteristics of our sample in Table 1, basic descriptive statistics (i.e., Mean, SD, and SE) of all measures used in Table 2, and Pearson correlations coefficients of all continuous- and count-based measures in Table 3. The gender distribution of our sample mirrors that of prior samples of Bhutanese refugees resettled in the United States. In prior studies, women made up 40–47% of recruited samples (Adhikari et al., 2015; Ao et al., 2012). Our sample comprises 31 (52%) men and 29 (47%) women. Mean age was 43.7 (SD=10.5) for men and 38.2 (SD=10.8) for women.

Table 1:

Sample Characteristics

	N	Overall % (n=60)
Gender
Man	31	52
Woman	29	48
Marital Status
Single	3	5
Married	45	75
Separated	7	12
Divorced	5	8
Literacy
English	31	52
Nepali	45	75
Religion
Hinduism	39	65
Buddhism	10	17
Christianity	8	13
Kirat	3	5
Caste/Ethnicity
Bahun	12	20
Chhetri	22	37
Janajati	3	5
Other	22	37
Employment Status
Employed	36	60
Unable to work	15	25
Sole Financial Provider for Family	10	17
Parent (has children)	57	95
Highest Level of Education
No Schooling	17	28
Primary School	19	32
Secondary School	17	28
Some University	5	8
Graduate Degree	2	3
Satisfaction with Community Support
Not at all	1	2
A little	4	7
Somewhat	20	33
Very	35	58
Involved with Community Organizations	50	83
Self-Perception of Overall Physical Health
Excellent	5	8
Very Good	15	25
Good	10	17
Fair	23	38
Poor	7	12
In Treatment for Mental Health issue	12	20
Prior Suicide Attempt*	4	8
Knew Someone who Engaged in Non-fatal Suicide
Behavior	15	25
Knew Someone who Died from Suicide	29	48
Any Substance Use	35	58
Cigarettes	9	15
Chewing Tobacco	12	20
Betel Nut or Paan	18	30
Alcohol	12	20
Domestic Violence Victim in Past Year
Physical	2	3
Verbal	2	3
Sexual	0	0

Note: Percentages may not always add up to 100%

^*Total sample size=53

Table 2:

Means and Descriptive Statistics

	n	M	SD	SE	Subscale/ Measure Item Numbers	Range of Possible Scores	Chronbach’s Alpha
Desire to be Dead (WDS)	60	1.62	2.5	0.32	1–11	0–11	0.858
BSS Total	60	1.02	2.45	0.32	1–19	0–38	0.785
Intensity of Ideation	60	0.12	0.45	0.06	6,7,9,15	0–8	0.376
Active Desire	60	0.32	0.95	0.12	1,2,3,4,8	0–10	0.697
Planning	60	0.02	0.13	0.02	12,16,17,18	0–8	0
Passive Desire	60	0.48	1.08	0.14	5,11,12,13,14	0–10	0.461
Concealment	60	0.02	0.13	0.02	10,19	0–4	0
INQ							--
Perceived Burdensomeness (INQ)	60	1.17	0.46	0.06	1–6	1–7	.702
Thwarted Belongingness (INQ)	60	1.74	0.82	0.11	7–15	1–7	.728

Note: Range indicates possible range of scores, not actual range of scores.

Table 3.

Pearson Correlations of Count and Continuous Variables

		1.	2.	3.	4.	5.	6.	7.	8.
1.	INQ Perceived Burdensomeness
2.	INQ Thwarted Belongingness	0.453**
3.	BSS Total	0.722**	0.537**
4.	Perceived BurdensomenessBSS-Ideation Intensity	0.737**	0.435**	0.715**
5.	Perceived BurdensomenessBSS-Active Desire	0.609**	0.541**	0.890**	0.582**
6.	Perceived BurdensomenessBSS-Planning	−0.050	−0.118	0.160	−0.034	−0.044
7.	Perceived BurdensomenessBSS-Passive Desire	0.593**	0.418**	0.894**	0.470**	0.659**	0.306*
8.	Perceived BurdensomenessBSS-Concealment	0.520**	0.471**	0.804**	0.544**	0.787**	−0.017	0.670**
9.	WDS Total	0.641**	0.433**	0.600**	0.533**	0.632**	−0.085	0.421**	0.388**

^**= p<.01

^*= p<.05

Suicidal Ideation and Desire for Death

Bhutanese refugees rarely endorse suicidal ideation, generally at a rate of 3–6% (Adhikari et al., 2015; Ao et al., 2012). We found a similar prevalence of suicidal ideation (specific thoughts about suicide) in our sample, with four individuals (7%) endorsing some non-zero level of suicide ideation (BSS: Ideation Intensity subscale). However, nearly half of our sample (29/60, 48%) reported experiencing at least some (non-zero) desire to be dead. Moreover, all of those who reported non-zero suicidal ideation also endorsed non-zero desire to be dead. This suggests that desire to be dead may be a helpful construct to explore as a culturally acceptable risk assessment tool.

A total of four participants (8%; n=53) had previous suicide attempts. Of these individuals, two indicated that they had a low wish to die during their last attempt and two indicated a high wish to die during their last attempt. The recency of these attempts was not assessed.

A significant minority of our sample personally knew someone who had engaged in a non-fatal suicide attempt (15/60, 20%), and nearly half our sample (29/60, 48%) personally knew someone who died by suicide. In response to several questions about the acceptability of suicide, the vast majority of our sample found suicide to be unacceptable even if someone is (a) tired of living or (b) doesn’t see any reason to continue living (53/60, 88%); however, 7/60 participants (12%) found suicide to be either somewhat or very acceptable due to either of these reasons.

Perceived Burdensomeness, Thwarted Belongingness, and the Desire to be Dead

For the first part of the hurdle approach, we utilized a binomial logistic regression to test whether perceived burdensomeness and/or thwarted belongingness predicted the presence of any (non-zero) desire to be dead (WDS) as compared to the absence of any desire to be dead. We found that neither perceived burdensomeness nor thwarted belongingness was predictive of the presence of the desire to be dead. However, we then tested the predictive power of perceived burdensomeness and thwarted belongingness to predict the intensity of the desire to be dead within individuals who endorsed non-zero levels of desire to be dead. We used a negative binomial distribution with a log link due to a marginally better fit than the same model using a Poisson distribution with a log link (Neg. Bin.: AIC=195.55, BIC=210.21; Poisson: AIC=198.21, BIC=210.78). Rootograms of both a negative binomial model and Poisson model visually confirmed the negative binomial model to have slightly better fit to the data than the Poisson model. Our zero-truncated model shows that perceived burdensomeness was predictive of WDS score (Table 4). Specifically, while maintaining thwarted belongingness at a constant, for every 1-point increase in perceived burdensomeness, we would expect to see a 209% increase in WDS score, corresponding to a 2-point increase in WDS score.

Table 4:

2-Part Hurdle Model Results

Zero Truncated Count Model (negative binomial distribution with log link)
	IRR	z	p	95% CI
				Lower	Upper
Desire to be Dead (WDS)
Intercept Thwarted	0.830	−0.352	0.7252	0.293	2.349
Belongingness Perceived	1.048	0.234	0.8146	0.709	1.549
Burdensomeness	2.093	2.314	0.0206	1.120	3.912
Zero Hurdle Logistic Model (binomial with logit link)
	OR	z	p	95% CI
				Lower	Upper
Intercept Thwarted	0.010	−1.803	0.071	<.0001	1.505
Belongingness Perceived	2.051	1.671	0.094	0.892	4.852
Burdensomeness	23.465	1.298	0.194	0.197	2747.945
Zero Truncated Count Model (Poisson distribution with log link)
	IRR	z	p	95% CI
				Lower	Upper
Suicide Ideation (BSS Ideation Intensity)
Intercept Thwarted	0.098	−0.655	0.512	<.0001	101.701
Belongingness Perceived	0.913	−0.174	0.862	0.325	2.563
Burdensomeness	3.014	0.776	0.438	0.186	48.889
Zero Hurdle Logistic Model (binomial with logit link)
	OR	z	p	95% CI
				Lower	Upper
Intercept Thwarted	0.0015977	−2.942	0.00326	<.0001	0.054
Belongingness Perceived	2.960	1.153	0.249	0.468	18.724
Burdensomeness	11.317	2.497	0.01254	1.685	76.027

Note: IRR=Incidence Rate Ratio;

OR=Odds Ratio

Perceived Burdensomeness, Thwarted Belongingness, and Suicidal Ideation

We then conducted the same set of analyses to examine whether perceived burdensomeness and/or thwarted belongingness was predictive of the presence or absence of suicidal ideation (BSS: Ideation Intensity subscale) as well as whether perceived burdensomeness and/or thwarted belongingness was predictive of suicidal ideation. Despite our initial hypothesis that both perceived burdensomeness and thwarted belongingness would be predictive of non-zero levels of suicidal ideation, as well as intensity of suicidal ideation, only four individuals reported any suicidal ideation. Thus, results are limited in their implications. However, we found that perceived burdensomeness was predictive of non-zero suicidal ideation. Specifically, while holding thwarted belongingness constant, each 1-point increase in perceived burdensomeness yielded an approximately 11-times greater likelihood of an individual experiencing some non-zero suicidal ideation (see Table 4). For zero-truncated count data, we used a Poisson distribution with a log link due to a slightly better fit than a negative binomial model (Poisson: AIC= 33.95, BIC= 46.51; Neg. Bin.: AIC= 35.95, BIC= 50.61). Rootograms of both Poisson and negative binomial models appeared identical and did not provide any indication as to better model fit. Neither perceived burdensomeness nor thwarted belongingness was predictive of the intensity of suicide ideation.

Interaction of Perceived Burdensomeness and Thwarted Belongingness

The interaction of thwarted belongingness and perceived burdensomeness was not significant in any step of any of our models predicting (a) the desire to be dead or (b) suicidal ideation.

Suicide Attempt History, Wish to be Dead, and Suicidal Ideation

Fisher’s Exact Tests revealed that within our small sample, suicide attempt history was not associated with either the presence of current desire to be dead (p=.305) or current suicidal ideation (p=1; see Table 5).

Table 5:

Fisher’s Exact Test Results

	Suicidal Ideation (present)	Suicidal Ideation (absent)
Suicide Attempt (present)	0	4
Suicide Attempt (absent)	4	45
	Desire to be Dead (present)	Desire to be Dead (absent)
Suicide Attempt (present)	3	1
Suicide Attempt (absent)	20	29

Note: values are numbers of participants

Discussion

The primary aim of this study was to employ a model-based framework (Meyerhoff et al., 2018) to improve suicide risk-assessment within Bhutanese refugees resettled in Vermont. Using Joiner’s IPTS (Joiner, 2005) and Chu’s cultural model of suicide (Chu et al., 2010) as the foundation for our framework (Meyerhoff et al., 2018), we used a model-driven approach to explore the array of risk factors for suicide that should be assessed in Bhutanese refugees, beyond those typically assessed in traditional Western settings (i.e., only suicidal ideation). The overall goal of this work is to inform a more culturally-responsive assessment of suicide risk in Bhutanese refugees.

Within our sample of 60 individuals, we found more non-zero endorsement of desire to be dead than of suicidal ideation. Specifically, 29 individuals endorsed some desire to be dead, whereas only four individuals endorsed some suicidal ideation. To our knowledge, this study is the first to assess both desire to be dead and suicidal ideation in a sample of Bhutanese refugees. Our results suggest that desire to be dead (as captured by the WDS) was more readily endorsed than suicidal ideation. While this crude comparison of the number of endorsements does not necessarily indicate the degree to which each measure captures true suicide risk, it suggests that individuals who endorsed suicidal ideation also tended to endorse the desire to be dead. Extrapolating from this finding, and drawing on the extant and emerging literature on suicide among Bhutanese refugees (reviewed in Meyerhoff et al., 2018), it may be the case that within the Bhutanese refugee population, the desire to be dead is a more palatable way of assessing individuals who may be at risk for suicide-related behaviors. Suicidal ideation may capture a different subset of at-risk individuals within a Bhutanese refugee population than it does in a non-minority sample. Additionally, three of the four individuals who endorsed making one or more prior suicide attempts also exhibited some current desire to be dead, while none of the four individuals who endorsed one or more prior suicide attempts endorsed current (past week) suicidal ideation. Importantly, our data are cross-sectional, and we did not collect data on the recency of these suicide attempts. Our data do not speak to the temporal precedence of either desire to be dead or suicidal ideation in relation to the suicide attempt. It remains possible that individuals who attempted suicide may have experienced suicidal ideation and/or weaker or no desire to be dead prior to their suicide attempt(s), but that suicidal ideation subsided and desire to be dead increased in the interim between attempt(s) and study participation. Therefore, it remains an empirical question as to whether desire for death better captures risk of suicide attempt than does suicidal ideation. With that in mind, in many minority cultures, it is common for suicidal ideation to be suppressed, such individuals present as “hidden ideators” (reviewed in Chu et al., 2010). Assessing desire for death, along with suicidal ideation, may better capture the full spectrum of risk and could potentially mitigate against any “hidden ideation” among Bhutanese refugees.

To our knowledge, only one other study to date has examined elements of the IPTS in a sample of Bhutanese refugees (Ellis et al., 2015). Despite our relatively small sample size (N=60) in comparison to that of Ellis et al.’s (2015; N=423), overall mean scores for perceived burdensomeness were comparable to those found in Ellis et al. (M=1.76, SD=0.74); however, Ellis and colleagues (2015) noted a greater overall level of thwarted belongingness (M=2.48, SD=1.13) than was found in our sample. This low thwarted belongingness score may be reflected in the 92% of participants in our sample who felt either “somewhat” or “very” satisfied with their community support. High levels of satisfaction may be the result of Vermont’s strong centralized infrastructure to support refugee resettlement (reviewed in Bose, 2013; see also the list of our community partners in Meyerhoff et al., 2018; the interested reader is also encouraged to explore “Refugee Resettlement in Small Cities,” 2019 [http://spatializingmigration.net/]).

Our results suggest the presence of suicidal ideation (relative to the absence of suicidal ideation) was associated with higher levels of perceived burdensomeness but not thwarted belongingness. However, neither perceived burdensomeness or thwarted belongingness was related to the intensity of suicidal ideation within those participants who endorsed any suicidal ideation. These findings partially mirror those from Ellis et al. (2015), whereby higher thwarted belongingness and higher perceived burdensomeness were each associated with ideation status¹. Although Ellis et al. (2015) did not examine intensity of ideation, their results highlight that elements of the IPTS (i.e., thwarted belongingness and perceived burdensomeness) might serve as early risk factors for suicidal ideation or intervention targets that may reduce overall suicide risk. Our results imply a similar conclusion as it relates to perceived burdensomeness. Although we replicated Ellis et al.’s finding that perceived burdensomeness is positively related to suicidal ideation, large confidence intervals (95% CI [1.7, 76.0]) suggest that the accuracy of our odds ratio (OR=11.3) is tenuous.

This study reports some novel findings when desire to be dead was examined as an outcome variable. Perceived burdensomeness, but not thwarted belongingness, was related to the intensity of the desire for death, but neither thwarted belongingness or perceived burdensomeness was associated with the presence (vs. the absence) of the desire to be dead. This suggests that while perceived burdensomeness is not a useful tool for determining whether or not resettled Bhutanese refugees will endorse any desire to be dead, once some desire to be dead is established, monitoring perceived burdensomeness may be clinically meaningful. Importantly, we did not find any interaction effects between perceived burdensomeness and thwarted belongingness in any of our models (predicting suicidal ideation or desire to be dead). This is an expected outcome given that we did not find any main effects of thwarted belongingness across any of our models. The results of the current study, in combination with previous research (Ellis et al., 2015), suggest that perceived burdensomeness, but not thwarted belongingness, has preliminary support as a candidate risk factor for the development of the desire to be dead and suicidal ideation – and must be further studied longitudinally – within the context of the Bhutanese refugee community. Critically, our results suggest that including the desire to be dead along with direct assessments of suicidal ideation is important for detecting potential risk in this population. Although it is, as of yet, unclear whether the desire to be dead effectively captures the specific risk of suicide attempts or suicidal behavior, it is clear that the desire to be dead is more likely to be endorsed in risk assessments within a community sample and may tap a more culturally-responsive construct.

The current study has several important limitations. First, while our analytic procedures for hurdle models followed previously published methodologies (Jobes et al., 2017), there remains a question as to whether scores on the WDS and/or BSS truly constitute count data or whether it is more appropriate to treat these scores as continuous variables. Another limitation of the current study is that it relies on a small sample size. Additionally, individuals who endorsed suicidal ideation exhibited low intensity and variability of suicidal ideation. Similarly, thwarted belongingness scores were lower and less variable than in previous studies of a similar population (Ellis et al., 2015). Furthermore, another limitation of the current study is that we relied on cross-sectional rather than longitudinal data. Finally, all measures used in this study were originally developed for use in English speaking populations and are based on Western constructs and cultural formulations. Many of the measures we used were intended to be self-report measures, but due to literacy concerns, all measures were administered in an interview format often through an interpreter.

Future studies of suicide-related thoughts and behaviors and suicide risk among Bhutanese refugees should be longitudinal in nature to capture within-person variation in suicide-related thoughts as a marker of risk. Moreover, longitudinal studies will allow for the testing of causal relationships between proximal risk factors and suicidal behaviors. Candidate models should also be tested against one another in order to refine our theories and maximize our ability to detect suicide risk among Bhutanese refugees. Finally, studying a clinical sample or a community sample with a history of suicide-related thoughts or behaviors is important, given their relative rarity.

Clinically, the results from this study highlight the need to consider an individual’s culture when assessing for suicide risk. Within the Bhutanese refugee sociocultural context, there are numerous factors that may contribute to an increased sense of perceived burdensomeness, thwarted belongingness, and “hidden ideation” status. Specifically, decades long oppression and isolation from a host country and country of origin, postmigration stressors, minority status and stress, a paucity of religious centers, language barriers, increased sense of isolation due to separation from family, lack of access to community resources, and many more contributing factors (Ao et al., 2012; Ellis et al., 2015; reviewed in Meyerhoff et al., 2018). Clinicians, subsequently, should adopt measures that do not rely entirely on the presence of suicidal ideation as the hallmark symptom of suicide risk and should instead utilize measures that take into account a variety of culturally-responsive risk factors and warning signs. Clinicians can make use of measures like the Cultural Assessment of Suicide Risk (CARS; Chu et al., 2013) or it’s screener version (CARS-S; Chu, Hoeflein, Goldblum, Espelage, et al., 2017), and the WDS. If patients exhibit a desire to be dead, clinicians should also consider monitoring the intensity of perceived burdensomeness, as these two risk factors combined may contribute to suicide risk. It is important to acknowledge that while some care providers, researchers, professionals, and lay-leaders believe that assessing for suicide risk may inadvertently and paradoxically increase one’s risk status, numerous studies highlight that screening for suicide risk does not significantly increase suicidal ideation (Crawford et al., 2011; reviewed in Dazzi, Gribble, Wessely, & Fear, 2014). The U.S. Preventative Services Task Force conducted a review that found insufficient evidence to conclude that there is any direct iatrogenic effect (or net benefit) of screening for suicide risk in primary care settings, and concluded that further research is necessary (LeFevre, 2014). However, The Joint Commission, the governing body for hospitals in the United States, implemented new national safety standards (National Patient Safety Goals 15.01.01) for required universal screenings using validated measures for all patients who are being treated for any primary behavioral health issue. Though The Joint Commission, does not make this a requirement for all patients who are being treated for a primary medical issue, they urge healthcare settings to adopt a policy of universal screening for suicide risk (The Joint Commission, 2018). Finally, while assessing for suicide risk is not a neutral act and has the potential to induce temporary discomfort or distress, results from previous research suggest insufficient evidence that it is, in fact, harmful (LeFevre, 2014). It is important that screening be done with sensitivity and care, but the potential benefits of identifying those at-risk for suicide and connecting them to services or interventions that are evidence-based outweigh any currently known drawbacks. Future qualitative research into how suicide risk screenings, broadly, and assessments of the desire to be dead, specifically, are received by Bhutanese refugees is warranted.

To further aid in stemming the suicide crisis in the Bhutanese refugee community, community leaders, or gatekeepers, can be trained to identify culturally-relevant risk factors and warning signs, empowering them to integrate risk screening into community-wide outreach efforts. Local community mental health clinics in areas with high populations of resettled refugees should invest in culturally-informed treatment methods that are also trauma informed. Additionally, in collaboration with local public health offices, community leaders can direct mental health de-stigmatization campaigns that aim to educate community members about culturally-relevant warning signs for suicide that also promote help seeking behavior. Finally, to reduce upstream risk, qualitative research should be conducted to identify factors that contribute to perceived burdensomeness in order to develop policies and programming that reduce the prevalence and intensity of perceived burdensomeness within the Bhutanese refugee community. It is critical to continue examining culturally-responsive risk factors for, and models of, suicide in order to prevent the continuation of the suicide epidemic in the United States, generally, and among resettled Bhutanese refugees, specifically.