Illicit Drug Use, Cognitive Distortions, and Suicidal Ideation among Homeless Youth: Results from a Randomized Controlled Trial

Abstract

Illicit drug use and cognitive distortions confer significant risks to youth suicidal thoughts and behaviors. However, there has been limited evidence regarding the efficacy of suicide prevention interventions with homeless youth, especially studies testing whether such interventions can reduce the risk for suicidal ideation associated with illicit drug use. Suicidal homeless youth (N = 150) between the ages of 18 to 24 years were recruited from a drop-in center. Youth were randomly assigned to Cognitive Therapy for Suicide Prevention (CTSP) + Treatment as Usual (TAU) or TAU alone. Youth reported their illicit drug use, cognitive distortions, and suicidal ideation 4 times over 9 months. A multiple-group multilevel structural equation model showed that higher illicit drug use at baseline predicted a slower reduction in cognitive distortions and suicidal ideation in the TAU group. These associations were not found in the CTSP + TAU group, suggesting an interruption of such risk from illicit drug use. Findings suggest that CTSP can reduce the risk of illicit drug use as a treatment barrier towards cognitive distortions and suicidal ideation among homeless youth, with implications to improve treatment efforts and to reduce premature mortality in a vulnerable population.

Homeless youth are a very vulnerable population who are at a significant risk for the onset of mental health concerns, including suicide (Kessler et al., 2007; Yoder et al., 2010). As the third leading cause of death among youth below the age of 24 years (Centers for Disease Control and Prevention, 2013), suicidal ideations and behaviors are even more common among homeless youth compared to their housed peers. Studies showed that over 30% of homeless youth reported a lifetime suicide attempt (Kidd, 2006; Yoder et al., 2010), in comparison to 7.8% among youth in U.S. (CDC, 2013). Among risk factors that predict and perpetuate youth suicidal ideation, the use of highly addictive and illegal substances is a significant factor that is associated with frequent suicidal ideation and completed suicidal attempts (Poorolajal et al., 2016; Swahn & Bossarte, 2007; Wu et al., 2004). More frequent drug use among homeless youth puts them at an elevated risk of suicide compared to their housed peers (Slesnick et al., 2015). The association between illicit drug use and suicidal ideation warrants special attention as it may influence treatment outcomes, and this association may be explained by change in maladaptive cognitions as cognitive therapy progresses. Given a dearth of suicide intervention research, the current study examined change in suicidal ideation associated with pretreatment illicit drug use in the context of cognitive therapy, among a sample of homeless youth.

Illicit Drug Use, Cognitive Distortions, and Suicidal Ideation

Illicit drug use confers significant risks towards the initiation and maintenance of psychiatric conditions such as mood disorders (Gold et al., 2018) and stress- and anxiety-related disorders (Najavits et al., 2020). Current views on the etiology for severe mental illness initiated by substance use disorders are usually rooted in the catecholamine hypothesis of schizophrenia or affective disorders (Mueser et al., 2007). These perspectives generally support a vulnerability preposition, such that substance use precipitates psychiatric disorders among vulnerable individuals through dysfunctions in brain reward circuits or psychosocial risk factors (Mueser et al., 2007). In contrast, limited research has examined the effect of illicit drug use in suicide research. This type of research is highly needed, especially in the context of suicide intervention, which may modify psychosocial risk factors that perpetuate suicidal thoughts and behaviors.

Illicit drug use may increase suicidal ideation through a variety of psychosocial factors. Illicit drug users may have increased psychological distress, which leads to suicidal thoughts and behaviors (Esposito-Smythers & Spirito, 2004). Illicit drug use may further constrain the thought patterns of youth experiencing suicidal ideation, enhancing distorted cognitions (Dalton, 2005; Wu et al., 2004). Defined as a distorted information process leading to maladaptive responses to stressful situations (Beck, 1976), cognitive distortions are an established risk factor for suicidal ideation and attempts (Fazakas-DeHoog et al., 2017; Jager-Hyman et al., 2014; Miller & Esposito-Smythers, 2013). Compared to non-suicidal individuals, suicidal individuals generally view their future, world, and oneself more negatively (Jager-Hyman et al., 2014; Miller & Esposito-Smythers, 2013), and tend to show higher levels of dysfunctional cognitive characteristics such as hopelessness (Brown et al., 2006), irrational beliefs (Ellis & Ellis, 2006), and deficits in cognitive problem-solving (Ghahramanlou-Holloway et al., 2012). As such, past research has revealed that a general cognitive distortion factor is related to suicidal ideation among adolescent and youth populations (Fazakas-DeHoog et al., 2017; Miller & Esposito-Smythers, 2013).

Research has recognized the positive association between illicit drug use and cognitive distortions among youth and adult samples (Dalton, 2005; Shoal & Giancola, 2001). Youth using illicit drugs can show a rigid cognitive pattern, generating polarized negative automatic thoughts (Esposito-Smythers & Spirito, 2004). Illicit drug users are also likely to selectively attend to negative events (Oliveira, 2007), memorizing adverse experiences (Pillersdorf, 2018), overgeneralizing personal memories (Gandolphe et al., 2013), and catastrophizing life events (Kneeland et al., 2019), which may reinforce a distorted negative view of the world and self. Additionally, homeless youth commonly experience high rates of trauma (Bender et al., 2012). A combination of illicit drug use and trauma experiences may contribute to an increased amount of cognitive distortions, possibly resulting from an individual’s need to create an alternative life narrative for self-healing (Najavits et al., 2004).

As such, illicit drug use, cognitive distortions, and suicidal ideation together likely contribute to a sequential pathway for risk transmission, among homeless youth. However, current evidence on the associations among illicit drug use, cognitive distortions, and suicidal ideation is mostly cross-sectional (e.g., Fazakas-DeHoog et al., 2017; Jager-Hyman et al., 2014; Miller & Esposito-Smythers, 2013). Longitudinal evidence on these factors remains scarce despite its significance, especially in the context of intervention research. Investigating factors associated with the process of change during intervention can provide clinically important information for therapists to understand efficacious elements in treatment, as well as to practice evidence-based interventions.

Intervention Research on Suicidal Ideation among Homeless Youth

Despite the high risk for suicide among homeless youth, intervention efforts for this population remain under-developed. Additionally, there is a severe shortage in intervention research for suicide in youth populations that is rigorously evaluated by experimental research designs, such as randomized controlled trials (Calear et al., 2016; Linehan, 2008). A few recent studies have shown support for the Cognitive Therapy for Suicide Prevention (CTSP; Wenzel et al., 2009). CTSP specifically focuses on reducing suicide behaviors and has shown efficacy among youth and young adults samples (Brown et al., 2005; Stanley et al., 2015). CTSP is based upon the theoretical assumption that people’s interpretation of life events determines their emotional and behavioral responses, and thus, it focuses on restructuring maladaptive cognitions associated with suicidal ideation (Wenzel et al., 2009). CTSP also involves techniques of cognitive therapy including practicing coping methods, gaining social support, as well as identifying reasons to live. As youth restructure their distorted cognitions and modify their maladaptive coping behaviors, their suicidal ideation is expected to decrease. As such, CTSP has been found to reduce homeless youth’s suicidal ideation (Slesnick et al., 2020).

In intervention research, it is important to understand the process of change. That is, whether the treatment can reduce the likelihood of risk transmission by interrupting longitudinal pathways from risk factors to treatment outcomes. Whereas initial evidence has been present for CTSP in reducing suicidal ideation through reshaping maladaptive cognitions, questions persist concerning the process of change for CTSP. As a major goal of CTSP is to intervene in distorted cognitive processing, it is unclear whether CTSP would reduce the risk of illicit drug use on suicidal ideation directly (i.e., through other elements in CTSP such as coping or obtaining social support) or through processes such as reducing cognitive distortions. Additionally, it is unclear that CTSP would intervene similarly on the risk transmission pathways preceding cognitive distortions (e.g., reducing the effect of illicit drug use on cognitive distortions by interrupting the process where illicit drug users develop constrained cognitive flexibility and form a rigid, polarized cognitive pattern) and proceeding cognitive distortions (e.g., lowering how already-formed cognitive distortions increase future suicidal ideation by disrupting the reasoning process that the life is not worth living due to youth’s negative, static world view and lowered self-image). Gaining an understanding of these questions may help to further reveal the effectiveness of specific elements in CTSP. As such, in the current study, we examined the effectiveness of CTSP by investigating whether specific longitudinal risk-transmission pathways apply similarly in both treatment groups engaged in the randomized controlled trial. These pathways included the preceding pathway from initial illicit drug use to change in cognitive distortions over time, the proceeding pathway from initial cognitive distortions to change in suicidal ideation, as well as the direct effect from initial illicit drug use to change in suicidal ideation (Figure 1a).

Figure 1. The hypothesized model of the current study.

(a) The conceptual model. H1, H2, and H3 refer to the study Hypotheses 1, 2, and 3.

(b) The statistical model.

Interventionists remain keenly interested in identifying factors that accelerate/hinder not only the mean level of a treatment outcome, but also the change, or improvement in treatment outcome over time. Studying changes in intervention response can provide robust information on longitudinal treatment effectiveness, while revealing important factors that facilitate or impede the improvement in treatment. In the current study, we modeled not only the change but also the parallel growth (or co-development over time) among illicit drug use, cognitive distortion, and suicidal ideation. This approach extends the current literature on suicidal ideation and increases the current understanding of the factors facilitating/hindering suicide intervention outcomes. Information gained from modeling the parallel growth between risk factors and treatment outcomes can help discern whether the changes in the treatment target (e.g., suicidal ideation) are associated with the pretreatment risk factors (e.g., illicit drug use), in addition to the changes in such risk factors throughout treatment.

The Current Study

The current study is a secondary analysis of a randomized controlled trial of CTSP among a sample of homeless youth experiencing suicidal ideation. In the major outcome analysis, youth showed a faster reduction in suicidal ideation in the group receiving CTSP added on to treatment as usual (TAU), compared to youth in the TAU group (Slesnick et al., 2020). The current study extended previous findings by investigating the longitudinal associations among illicit drug use, cognitive distortions, and suicidal ideation over 9 months, and tested whether these associations were different between the CTSP+TAU and the TAU groups.

Given the general expectation that CTSP would interrupt the risk transmission pathways from illicit drug use to suicide ideation, we hypothesized that the following longitudinal pathways would appear more salient (i.e., bigger in size) in the TAU group compared to the CTSP+TAU group:

1) Higher pretreatment levels of illicit drug use would slow down the decrease in cognitive distortions (the preceding pathway to cognitive distortions);

2) Higher initial cognitive distortions would also slow down the decline in suicidal ideation (the pathway proceeding cognitive distortions);

And 3) higher initial levels of illicit drug use would impede the reduction in suicidal ideation (the direct effect from illicit drug use to suicidal ideation). We illustrated the conceptualized model in Figure 1a, and the hypothesized model in Figure 1b.

Method

Participants

Homeless youth (N = 150) were recruited from a drop-in center in a large Midwestern city in the U.S. In order to be eligible for the randomized controlled trial, youth had to be between the ages of 18 and 24 years be able to provide informed consent, and score above 16 on the Scale for Suicide Ideation – Worst Point (SSI-W; Beck et al., 1999). Youth were considered ineligible if they screened positive for a psychotic spectrum disorder as determined by the Structured Clinical Interview for DSM-5 disorders psychotic screening (SCID; First et al., 2015), or required hospitalization due to suicidal ideation. We included demographic information on the current sample in Table 1.

Table 1.

Demographic Characteristics of the Current Sample

Variables	n (%)	Mean (S.D.)
Age		20.99 (1.96)
Sex
Female	61 (40.7%)
Male	89 (59.3%)
Race/Ethnicity
American Indian or Alaskan Native	1 (0.7%)
Asian, Asian-American, or Pacific Islander	1 (0.7%)
Black or African American	57 (38.0%)
Hispanic, Other Latin American	2 (1.3%)
White, not of Hispanic origin	59 (39.3%)
Other	30 (20.0%)
Highest degree received
Vocational	4 (2.7%)
High School Diploma	79 (52.7%)
GED	13 (8.7%)
Associate’s Degree	0 (0%)
Bachelor’s Degree	1 (0.7%)
Other	6 (4.0%)
None	47 (31.3%)
Current marital status
Single, never married	142 (94.7%)
Legally married	4 (2.7%)
Divorced	4 (2.7%)
Number of children
0	106 (70.7%)
1	26 (17.3%)
2 and more	18 (12.0%)
Received substance abuse treatment
Inpatient	23 (15.3%)
Outpatient	22 (14.7%)
Received treatment for emotional difficulties
Inpatient	85 (56.7%)
Outpatient	86 (57.3%)
Number of lifetime suicide attempts		6.11 (9.69)
Length of homelessness (days)		126.74 (198.82)
CTSP group: number of total sessions		5.01 (6.08)
CTSP sessions		2.36 (2.57)
TAU sessions		3.56 (4.84)
TAU group: number of TAU sessions		3.32 (4.63)

Procedures

Homeless youth were recruited and screened at the drop-in center by a trained research assistant. Interested youth were administered the SCID section on psychosis and the SSI-W to determine formal eligibility. Those meeting the inclusion criteria for the study consented and continued with the assessment battery. Youth determined at imminent risk for suicide were taken to a local hospital for psychiatric evaluation and/or crisis assessment. Youth who were ineligible were given a care package and continued receiving services at the drop-in center.

Eligible youth completed a baseline assessment with a research assistant and then were randomly assigned to receive treatment as usual (TAU) through the drop-in center (n = 75) or CTSP+TAU (n = 75). Youth did not differ in their demographic characteristics or baseline scores between groups. Licensed therapists at the drop-in center provided all interventions, and tracked the number of sessions that each participant completed. During the first 6 months post-baseline, youth in the CTSP+TAU group were able to receive up to 10 CTSP sessions and 9 CTSP booster sessions upon request, in addition to TAU services. After 6 months, all youth were able to continue to receive TAU services through the drop-in center. In addition to treatment sessions, we conducted follow-up assessments to evaluate the treatment process and outcomes for all participants, at 3-months (retention rate 89.4%), 6-months (86.6%), and 9-months post-baseline (85.9%), regardless of participation in treatment (i.e., an intent-to-treat design). Youth received a $5 gift card for each attended therapy session, and a $40 gift card for completing each baseline or follow-up assessment to compensate for their time of participation. All study procedures were approved by the University’s Institutional Review Board.

Intervention

Treatment as Usual (TAU).

The drop-in center provides services to meet youth’s basic needs, including food, laundry, toiletries, and shower facilities, social interactions and recreational activities such as television, books, games, and art, and linkage to community resources as needed. Additionally, the drop-in center employs two on-site therapists that provide non-directive client-centered therapy, their standard practice for suicide prevention.

Cognitive Therapy for Suicide Prevention (CTSP) + TAU.

CTSP is a manualized intervention (Wenzel et al., 2009) that was added on to TAU to determine the intervention effects above and beyond TAU, and focuses solely on suicide prevention. CTSP includes constructing and continually developing a crisis plan, assessing suicidal thoughts and behaviors, and engaging in cognitive restructuring and behavioral change to decrease risk factors linked to suicide. The treatment process is structured into three phases: 1) education about the cognitive model for suicide and case conceptualization (sessions 1-3); 2) restructuring cognitions and changing behaviors to address suicide-specific risk factors (sessions 4-7); and 3) practicing the newly acquired skills through guided imagery to prevent treatment relapses (sessions 8-10).

Ongoing Supervision, Fidelity, and Session Attendance.

Two independently licensed therapists attended a three-day onsite training of the CTSP intervention by Dr. Wenzel and read the CTSP manual (Wenzel et al., 2009). Dr. Wenzel provided ongoing weekly virtual supervision during the intervention. All therapy sessions were digitally recorded, and therapist competence was evaluated using the Cognitive Therapy Rating Scale (Young & Beck, 1980) by Dr. Wenzel, using a 7-point scale (0 = poor, 6 = excellent). A sum score of the 11 items was generated (range 0-66). Dr. Wenzel rated 53 sessions. The mean score for therapist “a” was 31.17 (SD = 5.54, range 20-41), and for therapist “b” was 32.26 (SD = 6.59, range 13-42). At 6-months, the average number of sessions was 5.01 (SD = 6.08) among participants in the CTSP+TAU condition and 3.32 (SD = 4.65) among participants in the TAU condition. Youth in the CTSP+TAU condition attended more meetings with their therapist in the first 3 months, t(148) = 2.81, p < 0.01. Additional information about the treatment, supervision, and session attendance can be found in Slesnick et al. (2020).

Measures

Suicidal ideation

was measured via the Scale for Suicidal Ideation – Worst Point (SSI-W; Beck et al., 1997) at all four time points. The SSI-W is a 19-item rating scale administered by the research assistant to measure the intensity of one’s attitudes, behaviors, and plans to complete suicide in the previous 90 days. Sample items include “I wish to die” and “I desire to make an active suicide attempt,” rated on a 3-point Likert-type scale (e.g., 0 = “none”, 2 = “moderate to strong”). A total score was generated by summing participants’ responses on all items, with a high score indicating elevated suicidal ideation. The SSI-W has shown good internal consistency (Cronbach alpha = .88) and established validity (Beck et al., 1997). In the current study, alpha levels for the SSI-W at baseline, 3-, 6-, and 9-months ranged from .69 to .89.

Cognitive distortions

were assessed at each time point using the Inventory of Cognitive Distortions (ICD; Yurica, 2002). The ICD is a 69-item self-report questionnaire with 11 subscales, each designed to assess a distinct cognitive distortion in clinical populations. Items are rated on a 5-level Likert scale (1 = “never”, 5 = “always”). Sample items include “To feel good, I need others to recognize me,” “I call myself negative names,” and “My negative predictions usually come true.” Previous investigations have supported the measure of cognitive distortions using the ICD as a unitary construct (e.g., Jager-Hyman et al., 2014; Strohmeier et al., 2016). Thus, a final score was calculated by summing all items, and higher scores indicated more severe levels of cognitive distortions. The ICD has demonstrated great internal consistency reliability (α = .96; Jager-Hyman et al., 2014) as well as strong concurrent validity (Yurica, 2002). Alpha levels for the ICD in the current study ranged from .96 to .97.

Illicit drug use

was measured at all time points using the Addiction Severity Index (ASI; McLellan et al., 1992). The ASI is a structured clinical interview administered by a research assistant that assesses the frequency, type, and amount of substance use in the past 30 days. Illicit drug use was assessed by drug class (13 items), including the use of heroin, methadone, other opiates, barbiturates, other sedatives, cocaine, amphetamines, cannabis, hallucinogens, and poly drugs. Sample items include, “How many days in the past 30 have you experienced problems with drug use?” A composite score was created following the procedures by McGahan et al. (1986) accounting for both the frequency and the severity of drug use, with higher scores indicating more problematic use of illicit drugs. The ASI has shown good internal reliability (α ranging from .70 to .87), moderate concurrent validity (r = .31 to .46), and moderate to high test retest reliability with homeless samples (Zanis et al., 1994).

Covariates

were assessed via a demographic questionnaire administered at baseline. In the current study, covariates included youth’s sex (1 = male, 2 = female), race (1 = White, 0 = non-White), and education in years.

Analytic Plan

Analyses were conducted using Mplus 8.1 (Muthén & Muthén, 2017). We used full information likelihood (FIML) estimation (Enders & Bandalos, 2001) with robust standard errors (MLR) to handle missingness, as Little’s MCAR test (Little, 1988) revealed that data were Missing Completely at Random, χ²(105) = 114.49, p = .25.

We used Multilevel Structural Equation Modeling (MSEM) to test our hypotheses, because the data were hierarchically structured, with repeated measures (i.e., level 1) nested within individuals (i.e., level 2). The MSEM was set up in a stepwise way. In the first step, the unconditional (i.e., no predictor) growth model was estimated to determine the degree of variance at the within-person and the between-person levels. The time points were coded as 1, 2, 3, and 4 for baseline, 3-, 6-, and 9-month assessments to estimate a linear growth curve model on the within-person level. Next, we added between-person predictors, with the initial levels of illicit drug use and cognitive distortions predicting the change in suicidal ideation, and the initial levels of illicit drug use predicting the change in cognitive distortions. Youth’s sex, race, and education were included as covariates at the between-person level.

We illustrated the MSEM for suicidal ideation (SI) as below. At the within-person level, the equation was:

$${\text{SI}}_{it}={\beta }_{0i}+{\beta }_{1i}({\text{TimePoint}}_{it})+{\gamma }_{it}$$

Where SI_it refers to suicidal ideation reported by participant i at time point t, β_0i refers to the intercept of suicidal ideation for participant i, and β_1i refers to the linear slope of suicidal ideation for participant i.

At the between-person level, predictors were added to the intercepts and linear slopes of suicidal ideation, and the models for participant i were:

$${\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}({\text{Sex}}_i)+{\gamma }_{02}({\text{Race}}_i)+{\gamma }_{03}({\text{Education}}_i)+u_{0i}$$

$${\beta }_{1i}={\gamma }_{10}+{\gamma }_{11}({\text{CDIntercept}}_i)+{\gamma }_{12}({\text{IDUIntercept}}_i)+{\gamma }_{13}({\text{Sex}}_i)+{\gamma }_{14}({\text{Race}}_i)+{\gamma }_{15}({\text{Education}}_i)+u_{1i}$$

Where CD refers to cognitive distortions, and IDU refers to illicit drug use. The model for cognitive distortions was set up similarly, with the intercept of illicit drug use and covariates as between-individual predictors. The model for illicit drug use only had demographic covariates as between-individual predictors. Note the three models were estimated simultaneously, with covariances among variables estimated both on within- and between-person levels, to account for the concurrent associations within each time point and parallel growth across time points.

Further, multiple-group models were tested within the MSEM to investigate the differences between the two treatment groups on interested associations. These associations included hypothesized paths from the initial levels of illicit drug use to the change in cognitive distortions (Hypothesis 1), from the initial level of cognitive distortions to the change in suicidal ideation (Hypothesis 2), and from the initial level of illicit drug use to the change in suicidal ideation (Hypothesis 3). To fully understand the multifaceted associations among the three variables across time, we also tested group differences on concurrent associations among illicit drug use, cognitive distortions, and suicidal ideation in the within-person level, and the parallel growth (or covariances among the slopes) of illicit drug use, cognitive distortions, and suicidal ideation at the between-person level (Figure 1b). A multiple-group approach is commonly used to compare differences in treatment effects in randomized controlled trials, with explanatory models (i.e., paths set to be equal among groups) nested within the full model (i.e., all paths set to be the different across groups; Emsley et al., 2010). Chi-square tests were used to compare the model fits between two nested models, one with the target path free and the one with the path constrained to be equal between groups. The path would be kept free if the model fit of the constrained model was significantly worse than the unconstrained model. The final models included freely estimated paths which were statistically different between groups, with other paths set to be equal between groups. We included the Mplus code for our model in the Supplemental Material to increase clarity.

Results

Descriptive statistics of study variables are presented in Table 2. Unconditional growth models were first estimated. Over time, youth’s illicit drug use (B = −0.01, SE = 0.003, t = −3.16, p = .002), cognitive distortions (B = −5.67, SE = 1.35, t = −4.22, p < .001), and suicidal ideation (B = −5.51, SE = 0.25, t = −21.96, p < .001) showed significant reductions in the entire sample. Illicit drug use and cognitive distortions showed significant variance on the between-person level (illicit drug use intercept variance = 0.01, SE = 0.002, t = 5.94, p < .001, slope variance = 0.001, SE = 0.000, t = 4.56, p < .001; cognitive distortions intercept variance = 1548.63, SE = 309.35, t = 5.01, p < .001, slope variance = 113.27, SE = 32.05, t = 3.53, p < .001), whereas the variances for the intercept (variance = 2.39, SE = 23.18, t = 0.10, p = .92) and the slope (variance = 0.49, SE = 3.10, t = 0.16, p = .88) of suicidal ideation were not significant in the entire sample.

Table 2.

Descriptive Statistics of the Study Variables.

Variable	CTSP+TAU Mean (SD)	TAU Mean (SD)	Total sample Mean (SD)	Skewness Total sample	Kurtosis Total sample
Suicidal ideation
Baseline	23.15 (5.03)	22.68 (4.66)	22.85 (4.90)	0.37	−0.11
3 months	6.58 (5.14)	7.74 (7.80)	7.17 (6.63)	1.01	0.90
6 months	5.25 (5.27)	6.81 (6.33)	6.02 (5.85)	0.91	0.10
9 months	4.91 (5.60)	6.00 (6.16)	5.45 (5.89)	0.97	0.05
Illicit drug use
Baseline	0.46(0.10)	0.49(0.11)	0.49(0.10)	1.43	1.89
3 months	0.43(0.07)	0.38(0.08)	0.43(0.08)	1.76	3.21
6 months	0.43(0.07)	0.38(0.07)	0.43(0.07)	1.96	3.83
9 months	0.44(0.07)	0.38(0.07)	0.44(0.07)	2.04	4.82
Cognitive distortions
Baseline	212.21 (47.07)	200.84 (43.46)	206.53 (45.51)	0.04	0.27
3 months	195.88 (47.98)	193.06 (44.37)	194.44 (46.02)	−0.02	0.31
6 months	198.75 (51.12)	183.36 (38.83)	191.12 (45.92)	0.65	1.49
9 months	191.54 (51.59)	186.31 (44.05)	188.95 (47.88)	0.27	0.24

Next, between-person predictors were added to the conditional growth model. Multiple-group tests indicated several significant between-group differences (Table 3). First, consistent with Hypothesis 1, the path from the intercept of illicit drug use to the slope of cognitive distortions differed significantly between two treatment groups, Δχ²(1) = 4.11, p = .04. Greater initial levels of illicit drug use were associated with a slower reduction in cognitive distortions only in the TAU group (B = 49.38, SE = 18.73, t = 2.64, p = .01), not in the CTSP+TAU group (B = 8.21, SE = 16.39, t = 0.50, p = .62). Inconsistent with Hypothesis 2, the initial levels of cognitive distortions were not associated with the reduction of suicidal ideation (B = 0.00, SE = 0.01, t = −0.72, p = .47), in both groups. Additionally, consistent with Hypothesis 3, the path from the intercept of illicit drug use to the slope of suicidal ideation differed significantly between two treatment groups, Δχ²(1) = 6.92, p = .008. Higher initial levels of illicit drug use significantly predicted a slower decline of suicidal ideation only in the TAU group (B = 8.31, SE = 3.39, t = 2.45, p = .01), not in the CTSP+TAU group (B = −0.77, SE = 1.79, t = −0.43, p = .67).

Table 3.

Multilevel Model Results.

	CTSP+TAU			TAU
	B	SE	t	B	SE	t	Δχ2(1)
Within-person level
Concurrent association between SI and CD	29.30	10.49	2.79**	29.30	10.49	2.79**	ns
Concurrent association between SI and IDU	0.07	0.02	3.33***	0.07	0.02	3.33***	ns
Concurrent association between CD and IDU	0.20	0.12	1.74	0.20	0.12	1.74	ns
Between-person level
SI intercept
Youth sex	0.56	1.56	0.36	1.80	1.44	1.25	--
Youth race	−0.11	1.72	−0.06	0.89	1.44	0.62	--
Youth education	0.06	0.43	0.15	0.33	0.30	1.13
SI slope
CD intercept	0.00	0.01	−0.72	0.00	0.01	−0.72	ns
IDU intercept	−0.77	1.79	−0.43	8.31	3.39	2.45 *	6.92**
Youth sex	−0.39	0.66	−0.58	−0.38	0.58	−0.66	--
Youth race	0.58	0.77	0.75	0.27	0.63	0.43	--
Youth education	0.14	0.16	0.88	0.16	0.15	1.04
CD intercept
Youth sex	22.25	11.56	1.93	10.15	10.86	0.94	--
Youth race	−2.05	12.91	−0.16	−9.31	11.38	−0.82	--
Youth education	8.27	3.96	2.09*	−2.76	2.87	−0.96
CD slope
IDU intercept	8.21	16.39	0.50	49.38	18.73	2.64 *	4.11*
Youth sex	−10.12	4.00	−2.53*	−2.82	3.66	−0.77	--
Youth race	5.49	4.63	1.19	−0.36	3.89	−0.09	--
Youth education	0.04	1.48	0.03	1.22	1.10	1.10
IDU intercept
Youth sex	−0.01	0.03	−0.45	−0.03	0.03	−1.27	--
Youth race	−s0.05	0.03	−1.60	0.02	0.03	0.73	--
Youth education	−0.01	0.01	−0.97	−0.01	0.01	−1.28
IDU slope
Youth sex	0.00	0.01	−0.16	0.00	0.01	0.10	--
Youth race	0.02	0.01	2.29*	0.00	0.01	0.27	--
Youth education	0.01	0.00	1.55	0.00	0.00	1.37
Mean initial levels of SI	23.22	5.99	3.88***	17.43	3.99	4.37***	--
Mean initial levels of CD	89.07	50.36	1.77	225.34	35.29	6.39***	--
Mean initial levels of IDU	0.28	0.19	1.52	0.30	0.11	2.61**	--
Mean changes in SI	−6.18	2.41	−2.57**	−6.64	2.56	−2.60**	--
Mean changes in CD	5.73	19.96	0.29	−20.74	14.21	−1.46	--
Mean changes in IDU	−0.09	0.06	−1.51	−0.04	0.03	−1.71	--
Parallel growth between SI and CD	4.88	1.83	2.67**	4.88	1.83	2.67**	ns
Parallel growth between of SI and IDU	0.00	0.00	0.88	0.00	0.00	0.88	ns
Parallel growth between of CD and IDU	0.00	0.03	0.13	0.00	0.03	0.13	ns

Note. Number of parameters = 83. Loglikelihood = −3868.02. SI= suicidal ideation; IDU = illicit drug use; CD = cognitive distortions. Race: White =1, other races = 0. Sex: male = 1, female = 2. Bold font indicates significant between-group differences.

^*p < .05

^**p < .01

^***p < .001 (two tailed).

Other associations among youth’s illicit drug use, cognitive distortions, and suicidal ideation did not yield significant between-group differences, and were thus constrained as equal between groups (Table 3). At the within-person level, positive concurrent associations were found among illicit drug use, cognitive distortions, and suicidal ideation, in both groups. As for covariates, females showed a faster reduction in cognitive distortions compared to males in the CTSP+TAU group (B = −10.12, SE = 4.00, t = −2.53, p = .01). White youth had a slower reduction in illicit drug use than non-White youth in the CTSP+TAU group (B = 0.02, SE = 0.01, t = 2.29, p = .02).

Discussion

The current study investigated the effects of CTSP on modifying the longitudinal pathways from illicit drug use and cognitive distortions to suicidal ideation in a randomized controlled trial, among a sample of homeless youth experiencing suicidal ideation. We found evidence suggesting positive treatment effects of CTSP in interrupting risk-transmission pathways. Findings of the current study have implications to improve conceptualization of suicide as well as intervention efforts for this high-risk, vulnerable population, as discussed below.

The current study provided evidence that CTSP interrupted the risk transmission pathway preceding cognitive distortions (i.e., reduced effects of illicit drug use on cognitive distortions in the CTSP+TAU group). Supporting our Hypothesis 1, we found that the initial level of illicit drug use predicted a slower reduction in cognitive distortions. Significant between-group differences were found in this association, which was only present in the TAU group but nonexistent in the CTSP+TAU group. As CTSP requires youth to reflect on and challenge their faulty cognitions, it may reduce the attention bias towards negative events, the polarized thought processes, as well as overgeneralization and catastrophizing in interpreting life events that are commonly seen in illicit drug users (Oliveira, 2007; Pillersdorf, 2018). By gathering evidence on a positive self-image, youth attending CTSP may be able to reconstruct their core beliefs about self-worth and reduce self-debasing feelings, thus alleviating distorted cognitions. Our propositions are supported by empirical evidence in the current study, whereas more process-focused research on how CTSP potentially reshapes youth’s cognitions is needed in future investigations.

Inconsistent with Hypothesis 2, we did not find that CTSP interrupted the risk transmission pathway proceeding cognitive distortions (i.e., there was no difference in the association between initial level of cognitive distortions and the change in suicidal ideation between two treatment groups). Rather, these two processes were tightly related concurrently and declined in a parallel manner in both treatment groups. Beyond current evidence that CTSP reduces cognitive distortions and suicidal ideation, no known research has tested whether CTSP can intervene in the association between these two factors. Our finding has two possible explanations. First, CTSP may not be able to interrupt the link between cognitive distortions and suicidal ideation, of which the robustness has been well recognized (e.g., Brown et al., 2006; Jager-Hyman et al., 2014; Miller & Esposito-Smythers, 2013). Rather, CTSP may be effective in reducing the formation of cognitive distortions (i.e., the preceding pathway), through which it lessens suicidal ideation. Second, this finding may be specific to our sample. As homeless youth tend to experience high levels of financial and interpersonal stress, mental health concerns, and traumatic events (Wu et al., 2020), these risks may have reinforced their negative views of the world and self, which appear to be inseparable from their suicidal thoughts. Both propositions warrant research attention in future studies.

Last, consistent with Hypothesis 3, we found a direct association between the pretreatment level of illicit drug use and a slower decrease in suicidal ideation over 9 months. This association showed significant between-group difference, in that it was only salient in the TAU group but not in the CTSP+TAU group. It seems that beyond interrupting the link between illicit drug use and distorted cognitions, CTSP provides additional benefits. Possibly, in addition to cognitive restructure, changes on the behavioral level such as gaining adaptive coping methods and forming a supportive social network can also shield youth from the negative influences of illicit drugs. This finding has further clinical implications. Considering that homeless youth tend to experience dual risks associated with both illicit drug use and suicide (Yoder et al., 2010; Slesnick et al., 2015), an integrated intervention approach is recommended not only catering to basic needs (e.g., housing, food, clothing, money, transportation, education, identification, legal aid, medical care and job training) but also including treatments towards both substance use and suicide (Slesnick et al., 2009). As the strong association between these two significant risk factors can be interrupted by CTSP, interventionists may consider using CTSP in conjunction to other addiction treatment approaches to reduce both illicit drug use behaviors and suicidal ideation.

Strengths, Limitations, and Clinical Implications

This is the first known study showing the potential of CTSP in producing positive change in suicidal ideation through reducing the negative effects of illicit drug use. Strengths of this study include 1) utilizing a randomized controlled trial design; 2) testing a manualized, empirically supported intervention; 3) long-term follow-ups for treatment effects with 4 assessment points over 9 months; and 4) engagement and treatment of a high-risk, underserved, and very vulnerable sample. The current study revealed that CTSP reduced risk-transmission pathways from pretreatment illicit drug use to suicidal ideation, shedding light upon the effective elements and the scope of application of CTSP.

Several limitations should be considered when interpreting the findings. First, this study included a convenience sample of homeless youth accessing a drop-in center, who might have greater motivation and skills to obtain resources and improve their living condition, compared to other street youth. As such, these findings may not generalize to those who do not use drop-in centers. Second, this study utilized a self-report measure for cognitive distortions, due to the amount of information to be collected in a short timeframe. We also used an interviewer-administrated measure for suicidal ideation, which was necessary as homeless youth may be reluctant in disclosing their suicidal thoughts and it is difficult to gather such information from their family and friends (Fulginiti et al., 2020). However, youth’s responses to the questionnaires/interviews can be influenced by their social desirability. Additionally, in the current study we did not test a mediation hypothesis (e.g., whether illicit drug use predicted changes in cognitive distortion then changes in suicidal ideation), as the parallel growth process was not equivalent to the preceding or proceeding pathways of interest, and did not afford testing for a mediation hypothesis due to a lack of temporal sequence. Testing such a hypothesis is warranted by further research for more robust study conclusions. Meanwhile, youth in the CTSP group may have attended only a portion of CTSP sessions (with an average number of fewer than three sessions), and thus it is possible that cognitive restructuring was not completed in CTSP sessions for some youth. Another limitation is the relatively lower ratings of therapist competence in some sessions, limiting the interpretation of the findings. Finally, illicit drug use usually brings about multifaceted changes on neurobiological, cognitive, affective, and behavioral levels, which may interact in complex pathways to impact youth’s suicidal thoughts and attempts. Future research needs to consider these pathways and to investigate the mechanisms underlying how treatment may ultimately reduce suicide.

The current study adds to the evidence base supporting the efficacy of CTSP with homeless youth, highlighting the importance of considering their illicit drug use behaviors when working with this population at high risk of suicide (Yoder et al., 2010). This study provides clinically significant information that can be used towards treatment planning, risk assessment, and process monitoring by therapists. This study also suggests the potential for incorporating CTSP with techniques from substance-use interventions such as Motivational Interviewing (Baker et al., 2002) and Ecologically-Based Family Therapy (Slesnick & Prestopnik, 2005), aiming to reduce their illicit drug use behaviors and improve mental health. Finally, this study supports the use of CTSP in reducing suicidal ideation in homeless youth who did not seek treatment, which may ultimately reduce the public health burden due to suicide among a very vulnerable population.

Highlights.

• Cognitive therapy attenuates the link from illicit drug use to suicidal ideation.

• Cognitive therapy mitigates the link from illicit drug use to cognitive distortions.

• Cognitive distortions covary with suicidal ideation among homeless youth.