Sexual Identity and Race/Ethnicity as Predictors of Treatment Outcome and Retention in Dialectical Behavior Therapy

Abstract

Objective:

There is inconclusive evidence regarding sexual identity and race/ethnicity differences in outcomes in evidence-based psychological treatments. Although Dialectical Behavior Therapy (DBT) is well-studied, little is known about the extent to which its efficacy generalizes to sexual minority and racial/ethnic minority people. This study examined sexual identity, race/ethnicity, and their interaction as predictors of treatment outcome and retention in DBT.

Method:

Data were from five clinical trials conducted in research and community settings with a variety of adult populations (N = 269) engaged in standard DBT, augmented DBT, or DBT components. Longitudinal mixed-effects models evaluated sexual identity and racial/ethnic differences in clinical outcomes (suicide attempts, non-suicidal self-injury (NSSI), global functioning, psychiatric hospitalizations) and retention.

Results:

Sexual identity, race/ethnicity, and their interaction did not predict the average severity or the rate of change in any clinical outcome over time. Sexual minority identity was associated with decreased risk of treatment dropout (OR = .44, p < .001). However, this effect was moderated by race/ethnicity, such that Non-Hispanic White sexual minority participants had the lowest rates of dropout. Exploratory analyses suggested potential differences related to NSSI for certain sexual and racial/ethnic minority subgroups.

Conclusions:

Findings suggest that sexual minority and racial/ethnic minority people may benefit from DBT as much as heterosexual and Non-Hispanic White people, respectively. Sexual minority identity interacted with race/ethnicity to predict dropout, such that sexual minority people were more likely to complete DBT if they were Non-Hispanic White. Further research is needed to clarify potential subgroup and intersectional differences.

In recent years, there has been growing attention to evaluating the effectiveness of psychotherapy for individuals from marginalized populations. Research using large datasets and meta-analytic methods has not found evidence of consistent differences in outcomes across treatments based on sexual identity (Lefevor et al., 2017; Rimes et al., 2019) or race/ethnicity (Cougle & Grubaugh, 2022), though there is some variability in results. This heterogeneity in findings may be due in part to evaluating outcomes across diverse types of psychotherapies. In this study, we compare clinical outcomes and retention between sexual identity and racial/ethnic groups in Dialectical Behavior Therapy (DBT) specifically.

DBT is an evidence-based psychotherapy (EBP) originally designed to treat individuals at high risk of suicide (Linehan, 1993) that has gained robust empirical support for treating borderline personality disorder (BPD; Storebø et al., 2020). Theoretical work suggests that the principles and structure of DBT may be well-suited to minoritized populations given its emphasis on the impact of invalidating environments on mental health and core dialectic between acceptance and change (Chang & Cohen, 2022; Cohen et al., 2021; Pantalone et al., 2019; Pierson et al., 2021; Skerven et al., 2019; Sloan et al., 2017). Several studies have found that sexual minority people improved comparably to heterosexual people in various psychological and functional outcomes when participating in DBT for adolescents (Poon et al., 2022) and a CBT and DBT skills-based partial hospital program (Beard et al., 2017). However, bisexual people may have worse outcomes on suicidal and self-injurious thoughts compared to other sexual identity groups (Beard et al., 2017). Similarly, studies have found that race/ethnicity did not predict outcomes in DBT for binge eating disorder (Safer & Jo, 2010), nor did it predict dropout for patients with BPD receiving outpatient DBT (Landes et al., 2016). One study found that Hispanic youth had greater reductions in suicide attempts compared to Non-Hispanic youth in DBT (Adrian et al., 2019). More rigorously designed studies with larger samples are needed to expand these findings.

Using data from five clinical trials of DBT, the current study examined whether sexual minority and racial/ethnic minority participants differed from heterosexual and Non-Hispanic White participants, respectively, on the average severity and rate of change in suicide attempts, non-suicidal self-injury (NSSI), psychiatric hospitalizations, global functioning, and treatment dropout during DBT. Given the importance of intersectionality, which states that identities are dynamic and mutually influential (Crenshaw, 2018), we also investigated whether sexual minority identity interacted with race/ethnicity in predicting these outcomes. We also conducted exploratory analyses to probe subgroup differences within sexual identity and race/ethnicity.

Method

Data Sources

The current study merged data from five clinical trials of DBT (see Table 1) involving adult participants who received standard DBT, augmented DBT (DBT + DBT Prolonged Exposure, DBT + buprenorphine), or DBT components (DBT individual therapy only or DBT group therapy only) and was approved by the VA Puget Sound IRB.

Table 1.

Summary of datasets included in this study

Study	Source	Study Type (Setting)	DBT Interventions	Patient Population	Sample Size	Treatment Length	Outcome Assessments (months)	Data Collection Period
1	Linehan et al., 2015	Efficacy (US university clinic)	DBT; DBT skills training w/case management; DBT individual therapy + activities group	Women with BPD + recent SIB	99	1 year	0, 4, 8, 12	April 2004 – September 2008
2	Harned et al., 2014	Efficacy (US university clinic)	DBT; DBT + DBT PE	Women with BPD + PTSD + recent SIB	26	1 year	0, 4, 8, 12	August 2009 – May 2013
3	Harned et al., 2021	Effectiveness (US public mental health agencies) a	DBT; DBT + DBT PE	Adults with PTSD	30	Variable	0, 4, 8, 12	February 2016 – March 2019
4	Linehan et al., 2006	Efficacy (US university clinic)	DBT	Women with BPD + recent SIB	52	1 year	0, 4, 8, 12	Not recorded
5	Linehan et al., 2009	Efficacy (US university clinics)a	DBT + buprenorphine	Adults with BPD + opioid dependence	62	1 year	0, 4, 8, 12	June 2004 – October 2008

Note: DBT = Dialectical Behavior Therapy. DBT PE = Dialectical Behavior Therapy Prolonged Exposure protocol. BPD = borderline personality disorder. SIB = self-injurious behavior. PTSD = posttraumatic stress disorder. US = United States.

^aMulti-site trial

Participants

A total of 269 adult participants were assigned to DBT conditions (inclusion criteria in Table 1 and demographics in Table 2).

Table 2.

Demographics across participants from all 5 DBT clinical trials, by sexual orientation and by race and ethnicity

Variable	Total sample	By sexual orientation		By race/ethnicity
	(n = 269)	Heterosexual (n = 199)	Sexual minority (n = 70)	Non-Hispanic White (n = 191)	Racial/ethnic minority (n = 75)
	n(%) or M(SD)	n(%) or M(SD)	n(%) or M(SD)	n(%) or M(SD)	n(%) or M(SD)
Age	29.4 (11.7)	29.9 (11.9)	27.9 (11.3)	29.2 (12.3)	29.5 (10.1)
Sex/gender
Male	32 (12%)	28 (14%)	4 (6%)	21 (11%)	11 (15%)
Female	236 (88%)	170 (85%)	66 (94%)	170 (89%)	64 (85%)
Transgender	1 (0%)	1 (0%)	0 (0%)	0 (0%)	0 (0%)
Sexual orientation		-	-
Heterosexual	199 (74%)	-	-	143 (75%)	53 (71%)
Gay/lesbian	15 (6%))	-	-	6 (3%)	9 (12%)
Bisexual	37 (14%)	-	-	28 (15%)	9 (12%)
Uncertain	16 (6%)	-	-	13 (7%)	3 (4%)
Other	2 (1%)	-	-	1 (1%)	1 (1%)
Race/ethnicity				-	-
Non-Hispanic White	191 (71%)	143 (72%)	48 (69%)	-	-
Native American/Alaska Native	1 (0%)	1 (0.5%)	0 (0%)	-	-
Black/African American	18 (7%)	12 (6%)	6 (9%)	-	-
Asian/Asian American	7 (3%)	5 (3%)	2 (3%)	-	-
Multiracial	29 (11%)	20 (10%)	9 (13%)	-	-
Hispanic a	20 (7%)	15 (8%)	5 (7%)	-	-
Treatment Type
Standard DBT	113 (42%)	72 (36%)	41 (59%)	84 (44%)	28 (37%)
Augmented DBT	90 (33%)	75 (38%)	15 (21%)	63 (33%)	26 (35%)
DBT components	66 (25%)	52 (26%)	14 (20%)	44 (23%)	21 (28%)
Education
Less than high school	24 (9%)	19 (10%)	5 (7%)	13 (7%)	11 (15%)
High school graduate or certificate of GED	51 (19%)	41 (21%)	10 (14%)	35 (18%)	16 (21%)
Some college or technical school	139 (52%)	96 (48%)	43 (61%)	100 (52%)	37 (49%)
College graduate	53 (20%)	42 (21%)	11 (16%)	42 (22%)	11 (15%)
Income
< $15,000	180 (67%)	127 (64%)	53 (76%)	127 (67%)	52 (69%)
$15,000–29,999	56 (21%)	46 (23%)	10 (14%)	41 (22%)	15 (20%)
>= $30,000	23 (9%)	19 (10%)	4 (6%)	18 (9%)	4 (5%)
Marital Status
Single, divorced, or separated	234 (87%)	171 (86%)	63 (90%)	147 (87%)	65 (87%)
Married or widowed	34 (13%)	27 (14%)	7 (10%)	24 (13%)	10 (13%)

Note: For studies 1, 2, 4, and 5, participants were asked, “What is your sex?” with response options: (1) female and (2) male. For study 3, they were asked “What is your gender?” with the following response options: (1) female, (2) male, and (3) transgender. For studies 1, 2, 4, and 5, participants were asked “What is your sexual orientation?” with response options: (1) uncertain, (2) heterosexual, (3) homosexual/lesbian/gay, (4) bisexual, and (5) other. For study 3, participants had the following response options: (1) heterosexual, (2) homosexual, (3) bisexual, and (4) other. For ethnic background, all participants were asked “Is your ethnic background Hispanic or Latino?” with answer options: (1) yes or (2) not Hispanic or Latino. For racial background, participants in studies 1, 2, 4, and 5 were asked “What is your racial background?” with response options: (1) White/Caucasian, (2) Native American, American Indian, or Alaska Native, (3) Black or African American, (4) Chinese or Chinese American, (5) Japanese or Japanese American, (6) Korean or Korean American, (7) other Asian American, (8) East Indian, (9) Middle Eastern/Arab, (10) Native Hawaiian or other Pacific Islander, and (11) other. Participants in study 3 had the same response options except there was a single “Asian or Asian American” option for all Asian ethnicities. Participants who were Hispanic were included in the Hispanic category regardless of their race.

Procedures

Independent evaluators who were masked to treatment condition conducted participant assessments at baseline, 4-, 8-, and 12-months (see Table 1).

Measures

Demographics.

Sex/gender, age, ethnic background, racial background, sexual identity, education, income, and marital status were assessed.

Suicide attempts and NSSI.

The Suicide Attempt Self-Injury Interview (SASII; Linehan et al., 2006) assessed frequency of suicide attempts and NSSI in the year prior to baseline (Studies 1–4) or lifetime (Study 5) and for outcomes since the previous assessment. This measure has very good interrater reliability and adequate validity (Linehan et al., 2006). Although these behaviors may not have been evenly distributed over the year prior to baseline, to create comparable timeframes, we divided baseline counts by three.

Hospitalizations.

The Treatment History Interview (Linehan & Heard, 1987) was used to assess the number of admissions to inpatient psychiatric units in the past year (baseline) and since the previous assessment (outcomes), with baseline counts divided by three.

Global functioning.

Global functioning was assessed using the Global Assessment Scale (GAS; Weissman, 1976; Studies 1, 2, 5) and the Global Assessment of Functioning Scale (GAF; Study 4). Both the GAS and GAF use a 0–100 rating scale to rate the overall severity of illness with higher scores indicating better functioning. The GAS has demonstrated inter-rater liability between 0.61–0.91 (Endicott, 1976).

Treatment dropout.

Participants were considered to have prematurely dropped out from DBT if they missed four consecutive sessions of either DBT individual or group therapy (Studies 1, 2, 4, 5) or if they dropped out of treatment prior to the expected program discharge (Study 3).

Analytic Strategy

All analyses were conducted using R Studio (RStudio Team, 2022). Primary analyses consisted of mixed-effects models. Preliminary analyses examined the distribution of outcome measures to identify which models best fit the data using Hu and Bentler’s (1998) guidelines (e.g., Poisson models, negative binomial models, and zero-inflated and hurdle models). Outcomes were modeled using Hierarchical Linear Models (HLM) for continuous outcomes (global functioning) and Hierarchical Generalized Linear Models (HGLM) for binary (hospitalization, dropout) and count outcomes (suicide attempts, NSSI acts). The lme4 package (Bates et al., 2015) was used for HLM analyses and the glmmTMB R package (Brooks et al., 2017) was used for HGLM analyses. Missing data ranged from 0–14.5% across outcomes (see Table 3).

Table 3.

Descriptives for main variables of interest by sexual orientation and race/ethnicity

Outcome	Pre-treatment		Post-treatment
	Heterosexual	Sexual Minority	Heterosexual	Sexual Minority
Suicide attempts, M(SD)	1.05 (3.17)	1.90 (6.88)	0.18 (0.82)	0.09 (0.43)
NSSI episodes, Median (SD)	1 (16.92)	3 (19.89)	0 (4.85)	0 (4.23)
Hospitalizations, M(SD)	0.50 (1.36)	0.58 (0.95)	0.03 (0.20)	0.07 (0.26)
Global functioning, M(SD)	40.16 (6.50)	42.07 (7.46)	51.90 (10.09)	51.64 (9.76)
Dropout, n (%)	-	-	78 (39.2)	19 (27.1)
	Non-Hispanic White	Racial/Ethnic Minority	Non-Hispanic White	Racial/Ethnic Minority
Suicide attempts, M(SD)	1.44 (5.16)	0.88 (1.73)	0.16 (0.83)	0.13 (0.38)
NSSI episodes, Median (SD)	1 (16.35)	1 (21.20)	0 (5.03)	0 (3.68)
Hospitalizations, M(SD)	0.53 (1.25)	0.52 (1.35)	0.03 (0.20)	0.07 (0.26)
Global functioning, M(SD)	40.39 (6.10)	41.05 (8.23)	52.25 (9.81)	50.17 (10.65)
Dropout, n (%)	-	-	67 (35.1)	29 (38.7)

Note: Missing data were as follows: 0% for treatment dropout, 12.2% for suicide attempts, 12.2% for NSSI episodes, 13.6% for hospitalizations, and 14.5% on global functioning.

In HLM/HGLM models, predictors in the model were a linear effect of Time (all assessment points from baseline to 12 months), a quadratic effect of Time, sexual minority (0 = heterosexual, 1 = sexual minority), racial/ethnic minority (0 = Non-Hispanic White, 1 = racial/ethnic minority); all two-way interactions, and the three-way interaction. Minimally adjusted models controlled for study, sex/gender, and age. DBT intervention type was examined as a potential moderator but did not significantly interact with either identity factor in predicting any outcome and was excluded from the final models.

Secondary exploratory analyses were conducted to probe differences among sexual identity as well as race/ethnicity subgroups. For sexual identity, categories included heterosexual (n = 198), lesbian/gay (n = 16), bisexual (n = 37), and uncertain (n = 16). Two participants who selected “other” as their sexual identity were excluded from these analyses. For race/ethnicity, categories included Non-Hispanic White (n = 191), Non-Hispanic Black/African American (n = 18), Non-Hispanic Asian/Asian American (n = 7), Hispanic (n = 20), and Non-Hispanic Multiracial (n = 29). One participant who was Native American/Alaskan Native was excluded from these analyses. Subgroup analyses were conducted within the HLM/HGLM models by changing the reference group to examine pairwise comparisons. Given the exploratory nature, corrections for multiple comparisons were not applied; however, we encourage due caution in interpreting these exploratory analyses.

The data and related study materials can be requested from the Principal Investigators of the relevant clinical trials. The analysis code can be accessed from the first author.

Results

Main Analyses

For all outcomes, descriptive data at pre- and post-treatment are in Table 3, separated by sexual identity and race/ethnicity. Results of the mixed effects models are in Table 4.

Table 4.

Results of Mixed Effects Models

	Time	SM	RM	SM X RM	Time x SM	Time x RM	Time x SM x RM
	b [95% CI]	b [95% CI]	b [95% CI]	b [95% CI]	b [95% CI]	b [95% CI]	b [95% CI]
Suicide attemptsa	−3.48 [−5.29, −1.67]	0.16 [−0.46, 0.77]	−0.06 [−0.69, 0.58]	−1.14 [−2.44, 0.16]	−0.91 [−2.75, 0.92]	0.68 [−1.03, 2.38]	−1.75 [−6.56, 3.06]
NSSI episodesa	−3.36 [−4.49, −2.24]	0.55 [−0.15, 1.26]	−0.40 [−1.15, 0.35]	−0.33 [−1.67, 1.01]	−0.03 [−1.05, 1.00]	0.41 [−0.72, 1.54]	0.62 [−1.28, 2.51]
Global functioninga	21.56 [17.95, 25.16]	1.38 [−1.27, 2.38]	0.24 [−1.69, 2.16]	0.53 [−3.23, 4.29]	−1.78 [−5.54, 1.96]	−2.28 [−6.41, 1.84]	−0.46 [−8.61, 7.65]
	OR [95% CI]	OR [95% CI]	OR [95% CI]	OR [95% CI]	OR [95% CI]	OR [95% CI]	OR [95% CI]
Hospitalizationb	0.01 [0.003, 0.02]	1.02 [0.48, 2.18]	0.90 [0.41, 1.97]	1.04 [0.25, 4.22]	-	-	-
Dropoutc	N/A	0.44 [0.30, 0.66]	0.88 [0.63, 1.23]	3.32 [1.70, 6.55]	N/A	N/A	N/A

Notes: Each row is a separate mixed effects model. SM = sexual minority. RM = racial/ethnic minority. Bolded values indicate significant at the p <.05 level. HLM was used for continuous outcomes (global functioning). HGLM with a binary logistic distribution with was used for variables with a dichotomous outcome (hospitalization, dropout), and HGLM with a negative binomial distribution was used for count outcome (suicide attempts, NSSI episodes). Covariates included age, gender, and study.

^aModel includes time, quadratic effect of time.

^bModel includes time and does not include interactions with time due to the model not converging when interaction terms were added.

^cModel does not include time as dropout did not vary by time.

Clinical outcomes.

Significant time effects were found for each outcome indicating that suicide attempts, NSSI episodes, and hospitalizations decreased while global functioning increased over the course of treatment. There were no significant differences on these outcomes on average or on the rate of change over time when comparing sexual minority with heterosexual participants, when comparing racial/ethnic minority participants with Non-Hispanic White participants, and when examining the interaction of sexual identity with race/ethnicity.

Treatment dropout.

Sexual minority identity was significantly associated with decreased risk of treatment dropout (p < .001). Compared to heterosexual participants, sexual minority participants were 56% less likely to drop out (OR = .44), with 27.1% of sexual minority and 39.2% of heterosexual participants dropping out across the five studies. Dropout rates did not differ between racial/ethnic minority (38.7%) and Non-Hispanic White participants (35.1%). However, sexual identity significantly interacted with race/ethnicity in predicting dropout (p < .001). Model-based estimates for each subgroup indicated that dropout rates were 23.9% for Non-Hispanic White sexual minority participants (n = 48), 38.6% for racial/ethnic minority heterosexual participants (n = 53), 41.6% for Non-Hispanic White heterosexual participants (n = 143), and 48.0% for racial/ethnic and sexual minority participants (n = 22).

Exploratory Subgroup Analyses

Results from exploratory analyses to evaluate potential differences between each sexual identity and race/ethnicity subgroup are presented in Supplemental Tables 1 and 2.

Clinical Outcomes.

There were no significant differences by sexual identity or racial/ethnic subgroup on suicide attempts or hospitalizations on average or on the rate of change over time. Compared to heterosexual (p = .042) and bisexual participants (p = .006), participants who reported being uncertain about their sexual identity had a significantly slower decrease in NSSI over the course of treatment. Compared to Non-Hispanic White participants, Non-Hispanic Black participants had a significantly slower decrease in NSSI over the course of treatment (p = .025). Black participants also demonstrated significantly lower rates of NSSI on average across all timepoints compared to both Non-Hispanic White (p = .023) and Non-Hispanic Multiracial participants (p = .0496). See Supplemental Figures 1 and 2 for model-based estimates of NSSI over time by subgroup. Compared to Non-Hispanic White (p = .039) and Black participants (p = .013), Hispanic participants reported significantly greater global functioning on average across timepoints. However, the rate of change in global functioning did not differ between groups.

Treatment Dropout.

Compared to heterosexual participants, gay/lesbian participants were 55% less likely to drop out of treatment (p = .024).

Discussion

In primary analyses, sexual minority participants did not differ from heterosexual participants in average levels or rates of change in suicide attempts, NSSI, global functioning, or hospitalization during DBT. This finding is consistent with past studies of DBT or DBT-based treatments (Beard et al., 2017; Poon et al., 2022). With regard to therapy more broadly, this finding is similar to a past study showing no differences in rates of change in anxiety and depression between sexual minority and heterosexual clients in college counseling centers (Lefevor et al., 2017), but differed from a past study finding worse outcomes on depression, anxiety, and functional impairment for lesbian and bisexual clients compared to heterosexual clients (Rimes et al., 2019).

Additionally, sexual minority participants in the present sample were less likely to drop out of DBT (27.1%) than heterosexual participants (39.2%) on the whole. This finding may also reflect more general patterns of increased help-seeking among sexual minority people who have been found to use mental health services more frequently than heterosexual people (Filice & Meyer, 2018). It is important to note that the protective effect of sexual minority identity on reducing dropout was moderated by race/ethnicity such that Non-Hispanic White sexual minority people were less likely to drop out, whereas heterosexual individuals and sexual minority people of color demonstrated appreciably higher dropout rates. It is possible that individuals with both marginalized identities experience greater stress due to the cumulative effects of both heterosexism and racism, and clinicians may be less knowledgeable or equipped to provide affirming care that incorporates intersectional considerations.

Regarding race/ethnicity differences, our findings indicate that racial/ethnic minority participants did not differ from Non-Hispanic White participants on the average severity or rate of change over time in any clinical outcome or treatment dropout. These findings are consistent with some past research on race/ethnicity differences in outcomes in DBT (Landes et al., 2016; Safer & Jo, 2010), as well as with research indicating a lack of race/ethnicity differences in psychosocial treatments more broadly (Cougle & Grubaugh, 2022). However, these findings contrast with other research suggesting that Hispanic youth had greater reductions in suicide attempts compared to Non-Hispanic youth participating in DBT (Adrian et al., 2019).

Taken together, these results suggest that sexual minority and racial/ethnic minority people may benefit comparably from DBT as heterosexual and Non-Hispanic White people, respectively. These findings provide preliminary support for the use of existing DBT interventions, including standard DBT, augmented DBT, and DBT components, with these minority groups. However, research with larger samples and greater power to detect potential differences is needed before drawing firm conclusions about the effectiveness of DBT for sexual and racial/ethnic minority people. Research is also needed to determine whether potential adaptations to DBT for marginalized groups (e.g., Cohen et al., 2021; Haft et al., 2022) enhance outcomes even further.

Though underpowered, subgroup analyses suggest potential differences in treatment outcome. Specifically, individuals who described their sexual identity as uncertain may derive somewhat less benefit from DBT compared to heterosexual people, though in the current study the observed outcome differences were not clinically significant. Further, Non-Hispanic Black people were found to have a slower rate of change in NSSI than Non-Hispanic White people, possibly due to NSSI being less frequent in Black people both in this study and in general (Wester & Trepal, 2015). In addition, our findings suggest that Hispanic participants may have greater overall global functioning compared to Non-Hispanic White and Black participants. Further research is needed to elucidate if and why these differences might exist.

Limitations

First, our study was likely underpowered to detect subgroup differences, and some groups (e.g., Native American/American Indian populations) were not represented. Although this study represents the largest sample that has tested sociodemographic differences in DBT outcomes that we could identify, a larger sample size would allow more rigorous testing of subgroup differences. Future work may consider combining DBT datasets with large datasets evaluating other forms of psychotherapies. Second, most of the studies were efficacy trials predominantly consisting of US cisgender women treated in university clinics; results may not generalize to other groups. Third, future studies should also include a more nuanced and comprehensive way of assessing participant gender and include more options for sexual orientation. Fourth, our subgroup analyses were prone to Type I error due to multiple comparisons.

Public Health Significance Statement:

Dialectical Behavior Therapy (DBT) is an evidence-based psychotherapy for borderline personality disorder and self-injurious behaviors. This study found no evidence of differences in DBT treatment outcomes based on sexual identity or race/ethnicity. Treatment dropout was lower among sexual minority than heterosexual participants for those who were Non-Hispanic White.