Prospective Relations Between Life Stress, Emotional Clarity, and Suicidal Ideation in an Adolescent Clinical Sample

Abstract

Objective:

Although life stress has been linked to adolescent suicidal ideation, most past research has been cross-sectional, and potential processes characterizing this relation remain unclear. One possibility may be lack of emotional clarity. Informed by stress generation, the current study examined prospective relations between episodic life stress, lack of emotional clarity, and suicidal ideation in an adolescent clinical sample.

Methods:

The sample consisted of 180 youth (M_age = 14.89; SD = 1.35; 71.7% female; 78.9% White; 43.0% sexual minority) recruited from a psychiatric inpatient facility. Suicidal ideation severity was assessed at baseline and 18-month follow-up. Lack of emotional clarity and life stress were assessed at baseline, as well as 6-, and 12-month follow-ups. Two random-intercepts cross-lagged panel models were created to estimate within-person relations for variables of interest.

Results:

At the within-person level, lack of emotional clarity at baseline predicted greater 6-month impact of interpersonal dependent stressors (b = 0.29, p = .012, 95% CI [0.07, 0.52]), which subsequently predicted greater 12-month lack of emotional clarity (b = 0.41, p = .005, 95% CI [0.12, 0.70]). Next, 12-month lack of emotional clarity, but not interpersonal dependent stress, predicted greater 18-month suicidal ideation (b = 0.81, p = .006, 95% CI [0.23, 1.30]; R² = .24, p < .001). No significant relations were found for lack of emotional clarity and independent stress.

Conclusions:

Results support the stress generation hypothesis and suggest that future research should be conducted evaluating whether bolstering youth’s understanding of their emotional experiences may reduce subsequent suicidal ideation.

Introduction

Suicide is a leading cause of death among adolescents in the United States (CDC, 2020). Suicidal ideation (SI) is an important risk factor for suicidal behavior, and often emerges during adolescence (Voss et al., 2019). SI is also an important clinical outcome in its own right (Kleiman, 2020), and predictive of other negative outcomes, such as functional impairment and high healthcare utilization and cost (Babcock et al., 2021; Oppenheimer et al., 2022). Thus, uncovering mechanisms underlying the development of SI during adolescence may help inform targeted treatments and save young lives.

Life stress is associated with SI (Liu & Miller, 2014; Stewart et al., 2019). Nevertheless, what processes may underlie the life stress-SI relation have been understudied (Liu & Spirito, 2019). Specifically, a dearth of longitudinal studies in the current body of work precludes study of related processes, (e.g., lack of emotional clarity [EC]; Moriya & Takahashi, 2013; Liu & Miller, 2014). Furthermore, with cross-sectional analyses, it becomes impossible to disambiguate stress exposure (i.e., predictive effects of life stress on subsequent outcomes) and stress generation (i.e., the propensity for some individuals prospectively to experience greater rates of dependent life stressors, or more specifically, stressors that are at least partially influenced by their own behaviors ; Hammen, 1991; Hammen, 2020; Liu & Alloy, 2010; Liu et al., 2023; Rnic et al., 2023). To build upon existing work, the current study utilized a multi-wave design to test reciprocal relations between life stress and lack of EC, as well as association with subsequent SI across an 18-month period in an adolescent clinical sample.

Life Stress and Suicidal Ideation

Life stress has been conceptualized and assessed in various ways in relation to suicidal thoughts and behaviors since the field’s inception (for an overview, see Liu & Miller, 2014). Contextual threat approaches are considered the gold standard for assessing life stressors: this is an interview-based approach that prompts participants to provide a detailed narrative of the circumstances surrounding each life event, which is then rated by a panel of raters unaware of interviewee’s psychopathology and subjective response (Hammen, 2005; Monroe, 2008; Liu & Miller, 2014). This approach allows for accurate assessment of types of stressors in a fine-grained manner, such as dependent (i.e., events that are, in part, influenced by the individual’s behavior) versus independent (i.e., events that are not influenced by the individual’s behavior) stressors and their temporality (i.e., episodic versus chronic; Hammen, 1991; 2005). Finally, this approach also allows researchers to determine whether stressors are interpersonal (or not) in nature. This is important as interpersonal stressors specifically may increase suicide risk in this population (Buitron et al., 2016; Glenn et al., 2022; Stewart et al., 2019).

Life stressors are generally linked to SI among individuals across the lifespan, though results are mixed. In their review of 40 years of research, Liu and Miller (2014) found 27 studies demonstrating a positive relation between life stressors and SI. However, 9 studies demonstrated null results, particularly after inclusion of covariates in statistical models (Liu & Miller, 2014). Additionally, less than 5% of studies reviewed had used an interview-based assessment of life stressors, which is a significant limitation in part for reasons stated above (Liu & Miller, 2014; Liu & Spirito, 2019). More recently, in a community sample of youth and adults, Uliaszek et al. (2023) examined the potential, reciprocal relations between SI and episodic stress across a 1-year period. Specifically, they sought to test stress generation and exposure with three waves of data. In partial support of stress generation, 6-month SI predicted 12-month interpersonal stress. Baseline SI did not predict 6-month interpersonal stress. Results did not support stress exposure (i.e., stress predicting SI) at any timepoint.

Interpersonal Life Stress, Lack of Emotional Clarity, and Suicidal Ideation

Considered a facet of emotion regulation (ER) difficulties, lack of EC is defined as challenges with understanding one’s own emotional experience (Gratz & Roemer, 2004). Importantly, preliminary theoretical and empirical work suggests that EC deficits influence the development of other facets of ER difficulties, such as lack of access to ER strategies and impulse-control difficulties (Vine & Aldao, 2014). Several studies have documented a positive relation between interpersonal life stress and lack of EC. For instance, in a sample of undergraduate students, Moriya & Takahashi (2013) found that interpersonal stress was contemporaneously related to lack of EC even after accounting for several facets of ER difficulties in the same model. Similarly, Freed et al. (2016) found a relation between baseline interpersonal stress, in the form of family dysfunction, and EC deficits one year later in an adolescent community sample. These findings parallel theoretical and neurobiological research positing that exposure to significant stress during adolescence, a developmentally sensitive period, may hinder the growth of ER capabilities, including EC (Haas et al., 2019; Lupien et al., 2009; McLaughlin et al., 2019). Thus, interpersonal life stress may hinder the ability for youth to understand their own emotions.

Although interpersonal life stress may increase lack of EC, it is equally possible that lack of EC may increase exposure to interpersonal stress. As noted previously, Hammen (1991; 2005) posits that individuals actively shape their social environment. Therefore, the occurrence and subsequent impact of some stressors, such as those that occur in the context of social relationships, will depend (at least in part) on that individual’s psychological characteristics (i.e., interpersonal dependent stressors). Existing studies suggest that lack of EC predicts future occurrence of interpersonal stressors. For instance, Hamilton et al. (2016) found that, among girls, lack of EC at baseline predicted risk of peer victimization, a form of interpersonal dependent stress, at 9-month follow-up in a community sample of youth. Thus, not only can interpersonal dependent stress negatively impact a youth’s ability to understand their own emotions, a lack of EC may also increase exposure to interpersonal dependent stressors. However, studies to date have not evaluated reciprocal relations between lack of EC and interpersonal dependent stressors within the same model.

Lack of EC and SI are also linked in the literature. A recent review of studies examining the relation between difficulties in ER and suicidal ideation and behavior across the lifespan found support for a link between EC deficits and SI (Colmenero-Navarrete et al., 2022). These results were consistent across community and clinical adolescent samples. Theoretical work also highlights the link between EC deficits and SI. Specifically, the integrated motivational-volitional model of suicidal behavior (O’Connor & Kirtley, 2018) posits that intrapersonal factors may increase the distress an individual experiences after a stressor. Specifically, if an individual has EC deficits, they may experience their emotions as overwhelming and/or unmanageable, potentially leading to thoughts of suicide as the only way to end their distress (Brausch et al., 2022; O’Connor & Kirtley, 2018). Of note, the integrated motivational-volitional model also highlights the role of interpersonal stressors as a precipitant of SI. As noted previously, interpersonal stressors may play a unique role in the onset and maintenance of SI, particularly among adolescents (Buitron et al., 2016; Glenn et al., 2022; Liu & Miller, 2014; Stewart et al., 2019). Taken together, existing theoretical and empirical work suggests that lack of EC may mediate the relation between interpersonal dependent stress and adolescent SI.

Statistical Considerations

A limitation of existing work is the reliance on statistical methods that confound between- and within-person processes. As noted by some researchers, stress generation, as most psychological theories, is a person-centered process (e.g., Curran & Bauer, 2011; Hammen, 2006). Recent work in methods research has highlighted the inability to distinguish between within- and between-person variance as a significant limitation of the predominant statistical model for testing bidirectional relations, the cross-lagged panel model (Hamaker et al., 2015). Instead, methodologists have recommended the use of random-intercepts cross-lagged panel models (RI-CLPM; Hamaker et al., 2015; Mulder & Hamaker, 2021). This approach allows for the disaggregation of between- and within-person variance, allowing direct tests of bidirectional, person-centered longitudinal processes (Hamaker et al., 2015; Mulder & Hamaker, 2021).

Current Study

The current study used a RI-CLPM framework to examine bidirectional relations between lack of EC and life stress as well as their subsequent impact on SI, with life stress assessed using gold standard contextual threat methods, in an inpatient adolescent sample, a group at high risk for suicide and life stress (Glenn et al., 2022). Specifically, lack of EC and interpersonal dependent life stress, but not independent stress, were hypothesized to exhibit bidirectional relations at the within-person level over a 1-year period (i.e., baseline, 6-, and 12-months). Lack of EC and interpersonal life stress were also hypothesized to predict SI severity 18 months post-baseline. Finally, the relation between baseline lack of EC and 18-month SI would be mediated by 6-month interpersonal dependent stress and 12-month lack of EC; the relation between interpersonal dependent stress and 18-month SI would similarly be mediated by 6-month lack of EC and 12-month interpersonal dependent stress (see Figures 1 and 2).

Figure 1.

Statistical diagram of within- and between-personal relations among interpersonal dependent stress, lack of emotional clarity, and suicidal ideation across 18-month follow-ups (i.e., Model 1).

Note. BL = baseline; 6M = 6-month follow-ups; 12M = 12-month follow-ups; 18M = 18-month follow-ups. SIQ-JR = Suicidal Ideation Questionnaire-JR; CDI = Children’s Depression Inventory-2; DERS Clarity = Difficulties in Emotion Regulation Scale Lack of Clarity subscale; LSI = UCLA Life Stress Interview. Paths that directly test hypotheses are in black. Significant paths are solid black while non-significant paths are dashed and black. See Table 2 for path estimates. Grey paths were estimated in the model, but values and their significance are not included here for ease of interpretation. b subscript denotes between-level component; w subscript denotes within-level component.

Figure 2.

Statistical diagram of within- and between-personal relations among independent stress, lack of emotional clarity, and suicidal ideation across 18-month follow-ups (i.e., Model 2).

Note. BL = baseline; 6M = 6-month follow-up; 12M = 12-month follow-up; 18M = 18-month follow-up. SIQ-JR = Suicidal Ideation Questionnaire-JR; CDI-2 = Children’s Depression Inventory-2; DERS Clarity = Difficulties in Emotion Regulation Scale Lack of Clarity subscale; LSI = UCLA Life Stress Interview. Paths that directly test hypotheses are in black. Significant paths are solid black while non-significant paths are dashed and black. See Table 3 for path estimates. Grey paths were estimated in the model, but values and their significance are not included here for ease of interpretation. b subscript denotes between-level component; w subscript denotes within-level component.

Method

Ethical Considerations

The Rhode Island Hospital institutional review board approved procedures noted below. Participants’ primary caregiver provided informed written consent, while written assent was obtained from youth.

Participants and Procedures

The current study consisted of a sample of 180 adolescents (M_age= 14.89; SD = 1.35; range: 13-17) recruited from a psychiatric inpatient unit located in the Northeastern U.S. Patients were eligible if they had been admitted for any reason. Exclusion criteria were: (1) non-fluency in English; (2) psychotic symptoms that precluded ability to provide informed assent or a diagnosis of pervasive developmental disorder; (3) IQ below 80 as assessed by the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999); (4) outside of the 13-17 age range; and (5) in custody of child protective services. Most of the sample identified as female (71.7%) and almost half identified as a sexual minority (43.0%). Regarding racial identities, 78.9% identified as White, 8.9% identified as Black, 8.9% identified as Multiracial, and 3.3% identified as Asian. Approximately 18% of the sample identified as Latino/a/x. Race and ethnicity data align with recent U.S. census data within the sample region which shows that 71.3% of the population identifies as White (non-Latino/a/x) and 18.7% identifies as Hispanic. Median family income fell within the $50,000-$74,000 range.

Participants completed various assessments at baseline during hospitalization and 6-, 12-, and 18-months post-discharge. Specifically, SI severity was assessed at baseline and 18-month. Life stress and EC deficits were assessed at baseline, 6-, and 12-months. Covariates (i.e., sex, sexual orientation, and depressive symptoms) were assessed at baseline (measure details are presented below). Baseline assessments were conducted at the hospital during the index admission. Six- and 12-month assessments were conducted in-person, whereas 18-month assessments were conducted remotely. Participants received monetary compensation for their time, and transportation costs were covered by the study as needed. Licensed clinicians with expertise in the assessment and treatment of suicidal youth were available at every assessment to conduct suicide risk assessments and ensure participant safety as needed. As adolescents’ recall of major life events begins to decline after about seven months (Monck & Dobbs, 1985), a 6-month interval between follow-ups was selected to maximize variability in the occurrence of life stressors while retaining youth’s ability to recall events reliably and with sufficient detail.

Measures

Suicidal Ideation Severity

The 15-item Suicidal Ideation Questionnaire-JR (SIQ-JR; Reynolds, 1987) was used to assess past month SI severity. The SIQ-JR is a self-report measure with excellent psychometric properties (Reynolds & Mazza, 1999). Items are rated on a 7-point scale from 0 (“I never had this thought”) to 6 (“Almost every day”). Ratings were summed to calculate a total score, with greater total scores indicating greater SI severity. This scale demonstrated strong internal consistency at baseline (ω = .95) and 18-months (ω = .94).

Lack of Emotional Clarity

The 5-item Lack of Emotional Clarity subscale of the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004) was used to assess youths’ understanding of their own emotional experiences. Only this subscale was utilized in the current study for three reasons. First, prior research suggests that lack of EC influence other facets of ER difficulties, (Vine & Aldao, 2014), highlighting the need to focus on EC deficits as a source of ER difficulties more broadly. Relatedly, the Lack of Emotional Clarity subscale most directly assesses a youth’s knowledge about what emotion(s) they experience at a given moment (Gratz & Roemer, 2004). Third, examining all 6 subscales and total score would have increased the risk of Type I error (Benjamini & Hochberg, 1995). Items for this subscale (e.g., “I have no idea how I am feeling”) are rated on a scale from 1 (“Almost never”) to 5 (“Almost always”), with some items reverse-scored; greater scores indicate greater lack of EC. The DERS has demonstrated good psychometric properties among adolescents (Neumann et al., 2010). The Lack of Emotional Clarity subscale demonstrated good internal consistency at baseline (ω = .81), 6- (ω = .85), and 12-months (ω = .83) in the present sample.

Life Stress

Impact of episodic life stressors was assessed with the UCLA Life Stress Interview (LSI) adapted for adolescents (Hammen & Brennan, 2001). Administered as a semi-structured interview, the LSI applies a contextual threat method of life stress assessment, which highlights the context-dependent nature of life stressors across a range of domains (e.g., friendships, romantic relationships, physical health, etc.) in terms of their impact on mental health (Brown & Harris, 1978). Interviewers provided participants with structured initial probes and temporal anchors (e.g., holidays); a calendar was also provided to ensure that the events fell within the assessment period (i.e., 6-month period preceding baseline, 6-, and 12-month follow-ups). Narratives of stressors reported during the interview were presented to a rating team unaware of participants’ clinical history and subjective response to circumvent rating biases (Harkness & Monroe, 2016). The rating team then determined the impact score on a scale from 1 (“no significant threat or impact”) to 5 (“maximal negative impact or threat”), with half-point increments, for each event based on relevant contextual details (e.g., nature of the event, consequences, duration, etc.). Per the scoring protocol for the LSI, only events coded ≥1.5 were included in the present study. Furthermore, narratives were coded for independence versus dependence (1 = “entirely independent of the person” to 5 = “entirely dependent on the person”). Consistent with prior studies of life stress (Conway et al., 2012; Hammen et al., 2000), events with independence versus dependence ratings ≥ 3 were then categorized as dependent stressors. Last, events were categorized as primarily interpersonal (i.e., events that primarily involve other individual[s]) or non-interpersonal (i.e., events that primarily relate to other domains, such as physical health or academics).

Impact ratings of relevant events (i.e., interpersonal dependent versus independent stressors) were summed to create baseline, 6-, and 12-month scores for analysis. In the current study, reliability was determined by having a second team, unaware of ratings from the first team, re-rate approximately a third of participants’ narratives at baseline and 6-months, allowing for calculations of interclass correlations (ICC). Baseline and 6-month ratings for dependent versus independent stressors demonstrated excellent reliability (all ICC ≥ .97, p < .001). Baseline and 6-month impact ratings for both types of stressors also demonstrated excellent reliability (all ICC ≥ .97, p < .001).

Depressive Symptoms

The Children’s Depression Inventory-2 (CDI-2; Kovacs, 2011), a 28-item, self-report measure, was used to assess participants’ depressive symptom severity during the past two weeks (e.g., “Over the past 2 weeks, I am sad once in a while [0], many times [1], or all the time [2]”). To avoid overlap between this measure and the SIQ-JR, the sole item assessing SI was excluded from the total score in the present study (range: 0-54; greater scores indicate greater depressive symptom severity). The CDI-2 has demonstrated good psychometric properties with youth (Kovacs, 2011). At baseline, internal consistency (ω = .91) was high in the present sample.

Data Analysis Plan

First, descriptive statistics and bivariate correlations were conducted in SPSS 28. Next, missing data patterns were assessed using Little’s test, in which a non-significant chi-square value indicates that missingness does not differ from what would be expected if data were missing completely at random (MCAR; Little, 1988). Variance inflation factors (VIF) and tolerance values were then calculated to evaluate for high multicollinearity (i.e., VIF > 10; tolerance < 0.1; Cohen et al., 2003) by regressing 18-month SI severity on EC and life stress ratings, at baseline, 6-, and 12-months as well as baseline covariates (i.e., SI/depressive symptom severity, sex [male was coded “0” and female “1”], sexual orientation [heterosexual was coded as “0” and sexual minority as “1”]¹) for each model. These covariates were selected as prior research suggests that they are some of the strongest predictors of adolescent SI (Cha et al., 2018; López et al., 2023; Ribeiro et al., 2018). Power analyses were conducted to determine if the current sample size provided sufficient power to detect adequate model fit (see Supporting Information for more details).

To examine longitudinal relations among life stress, lack of EC, and SI, two separate RI-CLPM were created in Mplus version 8.10 (Muthén & Muthén, 1998–2017), with one model including interpersonal dependent life stress (i.e., Model 1; see Figure 1) and the other independent life stress (i.e., Model 2; see Figure 2)². As noted above, RI-CLPM can disaggregate longitudinal data into between- and within-person components using latent variables (Hamaker et al., 2015). Conceptually, the latent, between-person components represent stable differences between people across time while the latent, within-person components represent changes within an individual person over time (Hamaker et al., 2015). Time-invariant predictors (or covariates) and outcomes can be included in the RI-CLPM (Mulder & Hamaker, 2021). Specifically, baseline, 6-, and 12-month EC as well as life stress were used to generate between- and within-person latent variables. Consistent with their approach, factor loadings for latent components were fixed to 1. Between-person components and baseline within-person components were allowed to covary. At 6- and 12-month time points, residuals for within-person lack of EC and life stress were allowed to covary. Cross-lagged and autoregressive paths between within-person components were freely estimated as no specific hypotheses were posited regarding the temporal invariance of these paths (Mulder & Hamaker, 2021). Next, 18-month SI severity was regressed on 12-month lack of EC and life stress³ as well as baseline covariates. Covariances between all covariates were freely estimated within both RI-CLPMs as all covariates were significantly related at the bivariate level (see Table 1). Finally, mediation hypotheses were tested using bias-corrected bootstrapping with 10,000 bootstrap re-samples. Indirect path estimates are considered significant if the 95% confidence intervals (CI) do not contain zero (Hayes, 2017). Both models utilized full information maximum likelihood with a robust estimator (MLR; Muthén & Muthén, 1998–2017) to address missingness and potential non-normality. Model fit was assessed using several indices, including an adjusted χ² statistic (Satorra & Bentler, 2010), the comparative fit index (CFI), the Tucker-Lewis Index (TLI), and RMSEA (Hu & Bentler, 1999). A non-significant χ² test and CFI/TLI values ≥ .95 suggest good model fit (Hu & Bentler, 1999). RMSEA values between .05 and .02 suggest good/close model fit, while <.01 suggest great model fit (Hu & Bentler, 1999).

Table 1.

Means, standard deviations, and bivariate correlations for all study variables (N = 180)

Variables	M (SD)	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. 18M SIQ-JR	20.21 (14.69)	–
2. BL SIQ-JR	44.32 (24.29)	.37***	–
3. 12M DERS clarity	13.64 (4.52)	.25**	.08	–
4. 6M DERS clarity	15.00 (4.96)	.16	.16	.54***	–
5. BL DERS clarity	15.14 (4.52)	.15	.30***	.44***	.55***	–
6. 12M LSI dependent	5.92 (4.35)	.19*	.21*	−.10	−.04	−.13	–
7. 6M LSI dependent	6.37 (4.35)	.15	.15	.13	.13	−.02	.43***	–
8. BL LSI dependent	6.44 (4.91)	.11	.02	−.11	.02	−.04	.44***	.35***	–
9. 12M LSI independent	3.32 (3.51)	.01	.20*	−.05	−.05	.07	.23***	.19*	.07	–
10. 6M LSI independent	3.72 (3.27)	.18*	.22**	−.04	.01	.04	.28**	.16	.13	.29***	–
11. BL LSI independent	3.07 (3.66)	−.01	.07	−.08	−.05	.10	.27**	.15	.16*	.32***	.32***	–
12. BL CDI-2	21.48 (10.54)	.31***	.64***	.19*	.33***	.37***	.20*	.15	.05	.17	.24**	.16*	–
13. Sexual minority orientation	–	.25**	.24**	.04	.09	.02	−.06	.07	−.05	.03	.14	.01	.27***	–
14. Female sex	–	.26**	.27***	.17*	.22**	.21**	.22**	.20*	.17*	.12	.12	.12	.34***	.32***	–

Note. BL = baseline; 6M = 6-month follow-up; 12M = 12-month follow-up; 18M = 18-month. SIQ-JR = Suicidal Ideation Questionnaire-JR; DERS clarity = Difficulties in Emotion Regulation Scale Lack of Clarity subscale; LSI = UCLA Life Stress Interview; CDI-2 = Children’s Depression Inventory-2.

^*p < .05;

^**p < .01;

^***p < .001.

Results

Descriptive Statistics, Bivariate Correlations, Retention Rates, and Power Analyses

Table 1 contains descriptive statistics for all study variables and bivariate relations among them. The retention rate at each timepoint were: 88.4% at 6-month, 88.3% at 12-month, and 87.2% at 18-month follow-up. Clinically, 58.3% (n = 105) of participants endorsed having made at least one lifetime suicide attempt at baseline. Participants also presented with a variety of psychiatric disorders and concerns at baseline, although almost two-thirds (n = 118) met criteria for major depressive disorder (see Table S1 for more details). On average, youth reported experiencing approximately three interpersonal dependent stressors at baseline, 6-, and 12-months. Relatedly, youth experienced approximately two independent stressors during the same period. Little’s test was non-significant for both models (${\chi }_{Model1}^2\left(138\right)=128.79,p=.701;{\chi }_{Model2}^2\left(138\right)=135.77,p=.538$), suggesting data were MCAR. VIF (range = 1.19-2.32) and tolerance values (range = .43-.85) for both models did not indicate high multicollinearity. Power was above 0.80 ($1-\mathrm{\beta }=0.86$) to detect acceptable model fit (i.e., RMSEA = 0.08; MacCallum et al., 1996; 2010).

Relations Among Lack of Emotional Clarity, Interpersonal Dependent Stress, & Suicidal Ideation

Model 1 demonstrated good fit to the data ${(\chi }^2\left(21\right)=17.73,p=.666$; CFI/TLI = 1.00; RMSEA = .00, 90% CI [.00, .05]) and accounted for 24% of the variance in 18-month SI (R² = .24, p < .001), a medium effect size (f² = .32; Cohen et al., 2003). Table 2 contains results for path estimates. Briefly, at the between-person level, baseline depressive symptoms were positively associated with lack of EC (unstandardized b = 0.13, p < .001, 95% CI [0.06, 0.20]). Furthermore, females, relative to males, also reported greater lack of EC (b = 1.34, p = .042, 95% CI [0.05, 2.65]) and experienced greater interpersonal dependent stress (b = 1.83, p = .004, 95% CI [0.60, 3.09]). No other covariates were significantly related to lack of EC or interpersonal dependent stress at the between-person level. At the within-person level, baseline lack of EC predicted greater 6-month interpersonal dependent stress (b = 0.29, p = .012, 95% CI [0.07, 0.52]), which then predicted greater 12-month lack of EC (b = 0.41, p = .005, 95% CI [0.12, 0.70]). Finally, 12-month lack of EC predicted greater 18-month SI (b = 0.81, p = .006, 95% CI [0.23, 1.39]), and the indirect path from baseline lack of EC to 18-month SI through 6-month interpersonal dependent stress and 12-month lack of EC was significant ($\mathrm{\beta }=.02$; b = 0.10, 95% CI [0.01, 0.34])⁴. Conversely, baseline dependent stress did not predict 6-month lack of EC (b = 0.20, p = .145, 95% CI [−0.07, 0.46]), which did not subsequently predict 12-month interpersonal dependent stress (b = 0.18, p = .278 [−0.15, 0.51]). Twelve-month interpersonal dependent stress did not predict 18-month SI (b = 0.42, p = .220, 95% CI [−0.23, 1.25]), and the indirect path from baseline interpersonal dependent stress to 18-month SI through 6-month lack of EC and 12-month interpersonal dependent stress was not significant ($\mathrm{\beta }=.00$; b = 0.02, 95% CI [−0.01, 0.14]). Of the covariates simultaneously included in Model 1, only baseline SI severity significantly predicted 18-month SI severity (b = 0.16, p = .004, 95% CI [0.05, 0.27]).

Table 2.

Parameter estimates for paths in model including interpersonal dependent stress, lack of emotional clarity, suicidal ideation, and covariates.

Cross-lagged paths among within-person components	$\mathit{\beta }$	b (se)	p	95% CI
BL Emotional clarity → 6M Interpersonal dependent stress	.30	0.29 (0.12)	.012	.05, .56
6M Interpersonal dependent stress → 12M Emotional clarity	.35	0.41 (0.15)	.005	.10, .60
BL Interpersonal dependent stress → 6M Emotional clarity	.20	0.20 (0.14)	.145	−.07, .47
6M Emotional clarity → 12M Interpersonal dependent stress	.23	0.18 (0.17)	.278	−.17, .64
Lagged paths predicting outcome (i.e., 18M suicidal ideation)
12M Emotional clarity (within-person component)	.21	0.81 (0.30)	.006	.08, .35
12M Interpersonal dependent stress (within-person component)	.09	0.42 (0.34)	.220	−.05, .23
BL Suicidal ideation	.26	0.16 (0.05)	.004	.10, .42
BL Depressive symptoms	.08	0.12 (0.14)	.414	−.11, .28
Female sex	.11	3.64 (1.35)	.179	−.06, .28
BL Sexual minority orientation	.13	3.72 (2.45)	.129	−.04, .29
Autoregressive paths among within-person components
BL Emotional clarity → 6M Emotional clarity	.24	0.28 (0.17)	.109	−.06, .53
6M Emotional clarity → 12M Emotional clarity	.17	0.16 (0.16)	.320	−.16, .50
BL Interpersonal dependent stress → 6M Interpersonal dependent stress	−.07	−0.05 (0.12)	.660	−.36, .23
6M Interpersonal dependent stress → 12M Interpersonal dependent stress	−.15	−0.14 (−0.68)	.495	−.57, .27
Between-person components regressed on covariates
BL Suicidal ideation → Emotional clarity	.02	0.01 (0.02)	.880	−.27, .31
BL Depressive symptoms → Emotional clarity	.50	0.13 (0.04)	<.001	.20, .80
Female sex → Emotional clarity	.22	1.34 (0.66)	.043	.01, .44
BL Sexual minority orientation → Emotional clarity	.17	−0.94 (0.58)	.104	.37, .03
BL Suicidal ideation → Interpersonal dependent stress	.13	0.02 (0.01)	.255	−.08, .33
BL Depressive symptoms → Interpersonal dependent stress	.05	0.01 (0.03)	.636	−.16, .27
Female sex → Interpersonal dependent stress	.29	1.83 (0.64)	.004	.09, .49
BL Sexual minority orientation → Interpersonal dependent stress	−.19	−1.09 (0.58)	.061	−.37, −.01
Contemporaneous paths among within-person components	r		p	95% CI
BL Emotional clarity ↔ BL Interpersonal dependent stress	.20		.102	−.04, .43
6M Emotional clarity ↔ 6M Interpersonal dependent stress	.45		.003	.25, .76
12M Emotional clarity ↔ 12M Interpersonal dependent stress	.08		.511	−.16, .32
Contemporaneous paths among covariates
BL Suicidal ideation ↔ BL Depressive symptoms	.64		<.001	.55, .73
BL Suicidal ideation ↔ Female sex	.27		<.001	.14, .41
BL Suicidal ideation ↔ BL Sexual minority orientation	.24		.001	.10, .37
BL Depressive symptom severity ↔ Female sex	.34		<.001	.21, .47
BL Depressive symptom severity ↔ BL Sexual minority orientation	.27		<.001	.13, .40
Female sex ↔ BL Sexual minority orientation	.32		<.001	.20, .45

Note. BL = baseline; 6M = 6-month follow-up; 12M = 12-month follow-up; 18M = 18-month follow-up. se = robust standard error. CI = confidence interval for standardized parameter estimate.

Relations Among Lack of Emotional Clarity, Independent Stress, & Suicidal Ideation

Model 2 also demonstrated good fit ${(\chi }^2\left(21\right)=18.27,p=.631$; CFI/TLI = 1.00; RMSEA = .00, 90% CI [.00, .05]) and accounted for 24% of the variance in 18-month SI (R² = .24, p < .001), a medium effect size ($f$² = .32; Cohen et al., 2003). Table 3 contains results for path estimates. At the between-person level, depressive symptoms were positively associated with lack of EC (b = 0.13, p = .002, 95% CI [0.05, 0.20]) and independent stress (b = 0.05, p = .042, 95% CI [0.001, 0.10]). No other covariates were significantly related to lack of EC or interpersonal dependent stress at the between-person level. Furthermore, at the within-person level, all cross-lagged paths between lack of EC and independent stress were non-significant. Twelve-month lack of EC, but not independent stress, predicted 18-month SI (b = 0.87, p = .013, 95% CI [0.19, 1.55]). The indirect path from baseline lack of EC to 18-month SI through 6-month independent stress and 12-month lack of EC was non-significant ($\mathrm{\beta }=.00$; b = 0.00, 95% CI [−0.06, 0.04]). The indirect path from baseline independent stress to 18-month SI through 6-month lack of EC and 12-month independent stress was also non-significant ($\mathrm{\beta }=.00$; b = −0.01, 95% CI [−0.08, 0.01]). Of the covariates simultaneously included in Model 2, only baseline SI severity significantly predicted 18-month SI severity (b = 0.17, p = .002, 95% CI [0.06, 0.28]).

Table 3.

Parameter estimates for paths in model including independent stress, lack of emotional clarity, suicidal ideation, and covariates.

Cross-lagged paths among within-person components	$\mathit{\beta }$	b (se)	p	95% CI
BL Emotional clarity → 6M Independent stress	−.01	−0.01 (0.13)	.983	−.34, .33
6M Independent stress → 12M Emotional clarity	−.06	−0.09 (0.17)	.620	−.31, .18
BL Independent stress → 6M Emotional clarity	−.11	−0.14 (0.13)	.269	−.31, .09
6M Emotional clarity → 12M Independent stress	−.11	−0.08 (0.09)	.368	−.33, .12
Lagged paths predicting outcome (i.e., 18M suicidal ideation)
12M Emotional clarity (within-person component)	.21	0.86 (0.35)	.013	.06, .36
12M Independent stress (within-person component)	−.07	−0.36 (0.42)	.401	−.22, .09
BL Suicidal ideation	.27	0.17 (0.05)	.002	.11, .44
BL Depressive symptoms	.08	0.11 (0.14)	.425	−.11, .27
Female sex	.11	3.73 (2.71)	.168	−.06, .28
BL Sexual minority orientation	.12	3.44 (2.43)	.157	−.05, .28
Autoregressive paths among within-person components
BL Emotional clarity → 6M Emotional clarity	.21	0.24 (0.19)	.212	−.12, .53
6M Emotional clarity → 12M Emotional clarity	.26	0.24 (0.14)	.093	−.04, .55
BL Independent stress → 6M Independent stress	−.01	−0.01 (0.12)	.931	−.30, .28
6M Independent stress → 12M Independent stress	.10	−0.11 (0.20)	.581	−.44, .24
Between-person components regressed on covariates
BL Suicidal ideation → Emotional clarity	.03	0.01 (0.02)	.824	−.24, .30
BL Depressive symptoms → Emotional clarity	.44	0.13 (0.04)	.001	.16, .73
Female sex → Emotional clarity	.19	1.25 (0.66)	.061	−.01, .39
BL Sexual minority orientation → Emotional clarity	−.11	−0.68 (0.58)	.242	−.31, .08
BL Suicidal ideation → Independent stress	.07	0.01 (0.01)	.508	−.13, .26
BL Depressive symptoms → Independent stress	.27	0.05 (0.03)	.042	.01, .53
Female sex → Independent stress	.11	0.49 (0.39)	.210	−.06, .28
BL Sexual minority orientation → Independent stress	−.02	−0.08 (0.44)	.862	−.23, .20
Contemporaneous paths among within-person components	r		p	95% CI
BL Emotional clarity ↔ BL Independent stress	.11		.368	−.12, .33
6M Emotional clarity ↔ 6M Independent stress	−.06		.653	−.32, .20
12M Emotional clarity ↔ 12M Independent stress	−.02		.816	−.22, .17
Contemporaneous paths among covariates
BL Suicidal ideation ↔ BL Depressive symptoms	.64		<.001	.55, .73
BL Suicidal ideation ↔ Female Sex	.27		<.001	.14, .41
BL Suicidal ideation ↔ BL Sexual minority orientation	.24		.001	.10, .37
BL Depressive symptom severity ↔ Female sex	.34		<.001	.21, .47
BL Depressive symptom severity ↔ BL Sexual minority orientation	.27		<.001	.13, .40
Female sex ↔ BL Sexual minority orientation	.32		<.001	.19, .45

Note. BL = baseline; 6M = 6-month follow-up; 12M = 12-month follow-up; 18M = 18-month follow-up. se = robust standard error. CI = confidence interval for standardized parameter estimate.

Discussion

Interpersonal dependent stress and adolescent SI have been linked, but research is lacking that employs gold standard methods assessing life stress within longitudinal designs that permit temporal assessments of life stress in relation to SI. Furthermore, existing research has often relied on statistical methods that do not disaggregate within- and between-person variance. To extend extant literature, this study used a multi-wave design and structural equation modeling to test longitudinal, within-person relations among lack of EC, life stress, and SI in high-risk youth.

Consistent with hypotheses, 6-month interpersonal dependent stress and 12-month lack of EC serially mediated the relation between baseline lack of EC and 18-month SI severity at the within-person level. Specifically, the more a youth experienced EC deficits at baseline, the greater interpersonal dependent stress they experienced 6-months later. In turn, greater interpersonal dependent stress at 6-months predicted greater 12-month EC difficulties, which ultimately predicted SI severity 18-months post-baseline. This study provides the most comprehensive examination to date of this question within the same model, and its findings are consistent with prior empirical research that has examined portions of this model. Specifically, existing work has demonstrated that EC deficits can predict interpersonal dependent stress (e.g., Hamilton et al., 2016) and vice versa (e.g., Freed et al., 2016). They also parallel existing work on the direct relation between EC deficits and SI (Colmenero-Navarrete et al., 2022). Importantly, results are consistent with the stress generation hypothesis, which posits that individuals with pre-existing vulnerabilities (in this case, lack of EC) may interact with their environments in ways that inadvertently increase the rates of prospective stress they experience, triggering a positive feedback loop (Hammen 1991; 2005). Indeed, several recent meta-analyses found support for dependent stress as a partial mechanism underlying baseline, and subsequent, levels of psychopathology symptoms, broadly defined (Liu et al., 2023; Rnic et al., 2023). Also supportive of the stress generation hypothesis, no significant relations were found between EC deficits and independent stress and the (relatively) weak to null relations between independent stress and psychopathology (Liu et al., 2023; Rnic et al, 2023).

Clinical Implications

Results from the current study can be used to inform suicide interventions for high-risk adolescents. First, assessment of EC deficits at the beginning of, and throughout, care may help inform treatment planning. Indeed, periodic assessment throughout the course of treatment is an important component of evidence-based care (Lewis et al., 2019). Furthermore, existing research has demonstrated that improvements in ER, including EC, are an important mechanism underlying changes in risk for self-injurious thoughts and behaviors among adolescents (Asarnow et al., 2021). Second, present findings may be used to provide psychoeducation to clients about potential consequences of EC deficits. Specifically, clinicians may highlight to youth how difficulties understanding one’s own emotional experience may increase their prospective risk for experiencing interpersonal dependent stress. Experiencing this stress, may, in turn, increase EC deficits, and ultimately increase risk for greater SI severity. Third, suicidal youth with EC deficits may benefit from treatments that incorporate mindfulness-based components, which have been shown to bolster EC among adolescents (Cooper et al., 2018). Of note, dialectical behavior therapy for adolescents (DBT-A), an effective treatment for self-injurious thoughts and behavior among youth, incorporates mindfulness-based strategies and interpersonal skills (Kothgassner et al., 2021; Miller et al., 2006). From a DBT-A perspective, ER difficulties, including lack of EC, are central treatment targets (Asarnow et al., 2021; Miller et al., 2006). To determine how ER difficulties may lead to interpersonal stress and thus increase suicide risk, a clinician may work with the client to complete a chain analysis (i.e., a step-by-step, factual description of contextual details that prompted a “problem” behavior; Miller et al., 2006). Missing links analysis, aimed at identifying effective behaviors that were not used during the situation, is a complementary tool (Miller et al., 2006). These tools may bolster a client’s awareness of self-perpetuating cycles of ineffective behavior and help a client generate a list of additional skills (e.g., “What” and “How” skills of mindfulness; see Miller et al., 2006) needed to extricate themselves from such patterns in the future.

Strengths, Limitations, and Future Directions

Although this study has significant strengths, including a multi-wave, multi-method design and rigorous analytic strategy in a high-risk sample, findings are to be interpreted in the context of some limitations. First, while the use of an adolescent inpatient sample is an important strength, it is uncertain whether results may generalize to less severe samples. Relatedly, while almost half of youth in the sample identified as a sexual minority, the majority identified as White, non-Latino/a/x, and female. Thus, future research including less severe and more diverse samples would permit evaluations of generalizability of the present findings. Second, the current study focused solely on SI as an outcome. The low number of youths who reported an SA between 12- and 18-months (n = 11) hindered use of SA as an outcome. Third, while the 6-month lags between follow-ups was selected to maximize variability in exposure to life stressors and minimize challenges recalling major life events among participants (Monck & Dobbs, 1985), the design precluded study of whether relations between EC deficits and life stress lead to proximal suicide risk (Nock et al., 2019). Use of intensive longitudinal designs in future research (Rabasco & Sheehan, 2022) may help clarify whether this mechanism impacts proximal risk. Finally, although the use of RI-CLPM is a strength, results from such models cannot be used to make definitive conclusions about causality (Orth et al., 2021).

Conclusion

This study provides the most comprehensive test to date on longitudinal relations among EC deficits, life stress, and SI using a gold standard assessment of life stress and a statistical method that allowed for a direct test of the hypothesized within-person process based on stress generation theory. Results suggest that the relation between lack of EC and SI severity is accounted for, in sequential fashion, by greater interpersonal dependent stress and EC deficits. Therefore, interventions intended to bolster EC among adolescents may reduce interpersonal dependent stress, subsequent EC deficits, and ultimately, SI severity.