Dysfunctional Beliefs about Sleep in Suicidal Adolescents: Associations and Treatment Implications

Abstract

Cognitive Behavior Therapy for Insomnia has shown increasing promise in reducing suicidality in adults and may be effective in reducing suicidality in adolescents. Dysfunctional beliefs about sleep (DBs) are a primary target of such interventions. However, there are few studies of DBs in adolescents, let alone suicidal adolescents. Our study 1) assessed the sleep and suicide-related correlates of DBs, and 2) examined if change in DBs was associated with change in sleep, in a sample of acutely suicidal adolescents aged 12-18 years in an intensive cognitive behavior therapy informed treatment program that did not target sleep. Measures of DBs (using the 16-item Dysfunctional Beliefs About Sleep Scale’s total score and four subscale scores: Consequences, Helplessness, Expectations, and Medications), sleep disturbance, depressive symptoms, emotion regulation, and suicide ideation were administered at program admission and discharge. Total DBs were concurrently associated (all ps < .05) with sleep disturbance, depressive symptoms, emotion regulation, and suicide ideation, and reduction of DBs was associated with improvement in sleep over treatment. There was some variability in the association between types of DBAS and variables of interest with the Helplessness subscale showing the strongest and most consistent associations. Limitations include subjective measurements only, a lack of specific insomnia symptom measures, and a nondiverse sample. DBs show promise as a target in the management of suicidal adolescents. Our findings also suggest that the development of an adolescent-specific scale could enhance construct measurement validity.

Introduction

Suicide is the third leading cause of death among adolescents aged 15-19 years. In 2022, 13.6% of adolescents reported having made an attempt in the previous 12 months, which is an increase over 12% from 2015 (CDC, 2024). Research has shown that suicidal thoughts and behaviors (STBs) in adult and adolescent populations are associated with sleep impairment, particularly insomnia (Blake & Allen, 2020; Kearns et al., 2020). Interventions to improve sleep, such as Cognitive Behavioral Therapy for Insomnia (CBT-I), have shown promise in also reducing STBs in adults (Kalmbach et al., 2022; Hertenstein et al., 2022). In adolescents, these interventions have been effective in improving sleep, as well as anxiety and depression symptoms, but they have not been examined for efficacy in reducing STBs (Blake & Allen, 2020). One of the primary targets of these interventions is dysfunctional beliefs about sleep (DBs) – defined as unrealistic expectations that arouse negative affect which are proposed to underlie the relationship between insomnia and suicidal ideation (SI) (Parsons et al., 2021). However, little is known about DBs in adolescents, and to our knowledge, no studies have investigated DBs and their dynamics in adolescents experiencing STBs. This study examined the sleep and STB-related correlates of DBs in youth seeking treatment for recent suicide attempts and high levels of suicide ideation, and whether DBs changed in tandem with subjective sleep at the end of a program for STBs that did not target sleep.

Sleep disturbance is associated with SI in adolescents (Gazor et al., 2022; Thompson et al., 2020), and with suicide attempts up to six months after discharge from treatment (Lau et al., 2020). Though not all youth who experience SI make a suicide attempt, SI is a precursor and a risk factor for suicidal behaviors (Klonsky et al., 2021). Depression is also a well-established risk factor for suicide attempts and death by suicide (Jeong et al., 2020). Sleep disturbance is a symptom of depression but can also further exacerbate depressive disorders (Pandi-Perumal et al., 2020). Poor sleep is suggested to be indirectly related to affective disorders, such as depression, via emotional dysregulation (ED) (Vanek et al., 2020). Sleep disturbance is also strongly associated with ED in adolescents, with some noting a bidirectional relationship between the two: poor sleep causes ED and vice versa (Vanek et al., 2020). Furthermore, research has indicated that the relationship between sleep disturbance and STBs is due to ED (Perlis et al., 2016). Difficulties with ED are also independently associated with suicide attempts in adult and adolescent inpatients (Colmenero-Navarrete et al., 2022). Importantly, all these variables, including DBs, share an underlying feature: the cognitive process of distortions. Individuals with depressive disorders (Beck, 2008), ED (Deperrois & Combalbert, 2022) and SI (Naderi et al., 2023) have all been found to manifest specific cognitions that are dysfunctional and exacerbate their deficits. To our knowledge, the examination of DBs with subjective sleep and mood has not been reported in adolescents with STBs, leaving a gap in the literature.

In adults, the experimental investigation of sleep using CBT-I has resulted in a knowledge base about changes in DBs. There is reliable evidence that CBT-I reduces DBs (Thakral et al., 2020), and that CBT-I-induced change in insomnia is mediated by change in DBs (Parsons et al., 2021, Faaland et al., 2023). However, there is some inconsistent evidence regarding whether DBs decrease with insomnia improvement when they are not targeted directly. For example, some studies have shown that reduction of DBs is specific to a cognitive approach as opposed to behavioral (Lancee et al., 2022) or pharmacological (Morin et al., 2002) treatment. Contrastingly, in a trial comparing the effects of hypnotic medication versus placebo on SI (McCall et al., 2019), neither of which directly addressed DBs, both groups showed a significant improvement in insomnia and reduced DBs. Additionally, in Morin and colleagues’ (2002) study, regardless of whether DBs were targeted, an association was reported between change in subjective and objective measures of sleep and change in DBs.

In contrast to the body of studies in adults, much less is known about the correlates and dynamics of DBs in adolescents. A recent study (Robbins et al., 2022) elicited “myths” about sleep from adolescents and their parents, but the association of these beliefs with sleep quality was not demonstrated. A study of community adolescents suggested that DBs are associated with poor subjectively reported sleep and mood problems during some parts of the academic year but not during others (Bei et al., 2015). A longitudinal study of community adolescents showed that adolescents with higher DBs reported significantly lower positive (but not negative) affect when sleep duration declined (Chachos et al., 2021). Thus, there is a second gap in the literature: information about the association between changes in DBs and changes in sleep, mood, and suicide ideation following treatment of adolescents for suicide ideation and attempt.

The first aim of the study was to increase information about the presence of DBs at admission to treatment for suicidality, and their association with subjective sleep, mood, and emotion regulation. We hypothesized that DBs would be significantly and positively associated with sleep disturbance. We also expected that DBs would be associated with depression, ED, and SI, all of which are influenced by cognitive distortions. Further, we hypothesized that the different kinds of distortions, captured by the subscales of the instrument we used to assess DBs, would show the same pattern of correlation with related variables as have been reported in adult samples (Morin et al., 2007): The Consequences subscale, the Helplessness subscale, and the DBAS total would be more strongly associated with measures of sleep, mood, and emotional regulation, while the Expectations subscale and Medication subscale would be relatively weakly associated with these measures. We made the comparison to adult samples given the paucity of literature on DBs in adolescents.

Our second aim was to examine whether our study variables improved in response to the non-sleep-specific cognitive-behavioral treatment of suicidality, and whether change in DBs was associated with change in subjectively reported sleep improvement. Further, we assessed whether our findings were consistent with data from adult samples (Morin et al., 2007), such that change in DBs (total and subscales) and change in sleep improvement would be correlated, and that the effect sizes would be small but significant.

Method

Participants

This study used clinical data gathered from an Intensive Outpatient Program (IOP) for suicidal adolescents (aged 12 – 18). IOP admission required active SI and/or an attempt made within the past month. IOP treatment consisted of individual and family therapy, medication management, parent psychoeducation, and group therapy twice weekly for four to six weeks. Sleep is not routinely targeted in this program, and a database is maintained to evaluate program outcomes. This study used data collected at two time points: IOP admission and discharge.

All participants attended an intake session between September 25, 2020, and September 14, 2023. Of the 465 participants assessed for eligibility at IOP admission, 419 (90%) had a complete set of baseline data relevant to Aim 1 analyses. Of these 419 participants, 223 (53%) did not have discharge data for the following reasons: 50 were not admitted to the IOP, 47 did not complete IOP treatment, 37 were missing relevant discharge data, and 89 participated in a sleep intervention offered to a randomly selected subgroup of participants. The remaining 196 participants had complete baseline and exit data relevant to Aim 2 analyses. The individuals included in the Aim 2 sample (n = 196) were not significantly different from those excluded from the Aim 2 sample (n = 223) on variables of interest obtained at admission.

The use of clinical data for this study was approved by the Institutional Review Board of UT Southwestern. Our retrospective study of deidentified data was deemed exempt from consent and assent requirements.

Measures

All measures were administered at admission and discharge.

Dysfunctional Beliefs and Attitudes About Sleep Scale - 16-item (DBAS)

To examine self-reported dysfunctional beliefs about sleep, we selected the DBAS-16 scale (Morin et al., 2007) which is reported to be the most widely used measure of dysfunctional beliefs about sleep (Thakral et al., 2020). The scale has been widely used with adults and has also been used in studies of DBs with adolescents (Chachos et al., 2023). The DBAS is a 16-item self-report measure designed to examine sleep-related cognitions, including causes and appraisals of the consequences of insomnia, and issues of worry and helplessness about insomnia. Higher scores reflect a greater presence of DBs. We used the total score (average of all items) and the four subscale scores (average of items on respective subscales) for analyses: “Perceived Consequences of Insomnia” (Consequences), “Worry/Helplessness about Insomnia” (Helplessness), “Sleep Expectations” (Expectations), and “Medications” (Medications). In a sample of community adolescents, the average DBAS score was 3.91 (SD = 1.48) (Chachos et al., 2023). The DBAS total score had strong internal consistency in our sample (16 items; α = .88). For the DBAS subscales, internal consistency was also strong for Consequences (5 items; α = .82) and Helplessness (6 items; α = .81). However, Expectations (2 items; α = .60) and Medications (3 items; α = .67) subscales had lower internal consistency.

Pittsburgh Sleep Quality Index (PSQI)

To assess for self-reported sleep disturbance, we selected the PSQI (Buysse et al., 1989), which is a subjective measure of sleep quality. For our study, we conceptualized poor sleep quality as indicative of sleep disturbance. This scale uses 19 items to make up seven components that are summed to provide a total score. The internal consistency was low for the original total PSQI score (7 components; α = .46).

The PSQI has been normed with community adolescents (Raniti et al., 2018), and suicidal adolescents may show different response patterns. To extract a scale with acceptable reliability, we conducted a factor analysis using Principal Component Analysis with an oblimin rotation. The analysis yielded three factors of eigenvalues greater than 1.0. The majority of items fell on a single factor with a loading of .40 or greater. The remaining two items, habitual sleep efficiency and use of sleep medication, fell on two separate factors, and were poorly correlated with the first factor and with each other. We excluded these items, resulting in a modified PSQI five-item scale that showed much better reliability (α = .65).

Quick Inventory of Depressive Symptomatology - Adolescent Version (QIDS - A)

To evaluate depressive symptoms, we employed the validated self-report version of the Quick Inventory of Depressive Symptomatology, Adolescent Version (QIDS-A₁₇-SR) total score (Haley et al., 2023). This self-report measure gauges depressive symptoms in accordance with DSM-IV criteria, which also maps onto the current DSM-V. Higher scores indicate more severe depressive symptoms. This scale uses 17 items, 12 of which are combined into four intermediate scores via the maximum of their constituent items that are then summed with the remaining five items. For the purposes of our analyses, the four items that address sleep (items 1-4) were excluded from the total score to avoid overlap with the DBAS on sleep-related items. The resulting modified QIDS score had strong internal consistency in our sample (8 items; α =.82).

Difficulties in Emotion Regulation Scale – Short Form (DERS)

The DERS-SF is a well-validated and widely used 18-item self-report measure of emotional dysregulation among adolescents and retains the strong psychometric properties of the original 36-item measure (Kaufman et al., 2016). Higher scores reflect greater difficulty with ED. The DERS-SF total score, obtained by summing all items, was used for this study and had strong internal consistency in our sample (18 items; α =.88).

Concise Health Risk Tracking Self-Report Scale (CHRT-SR)

Suicide ideation was assessed by the three items that sum to comprise the Suicidal Thoughts subscale of the CHRT Self Report (Nandy et al., 2023). The three items, rated on a five-point scale, address thoughts, methods, and plans to kill oneself. Overall, the CHRT Suicidal Thoughts subscale had strong internal consistency in our sample (3 items; α =.87).

Analysis

Aim 1:

Pearson correlations were conducted to examine the relationships between the DBAS total, DBAS subscales, and outcome measures at admission (Table 2). To examine the hypothesized difference in the strength of the correlations between the Expectations subscale and outcomes and the other DBAS subscales and outcomes, we used seven two-tailed z-tests adjusted for multiple testing using the Bonferroni correction.

Table 2.

Correlations Between DBAS Scales and Relevant Measures at Program Admission (n = 419)

Variables	DBAS a
	Total	Consequences	Helplessness	Expectations	Medications
PSQI b	.42 ***	.28***	.53 ***	−.01	.32 ***
QIDS c	.41 ***	.28***	.45 ***	.10	.32 ***
DERS d	.45 ***	.36 ***	.45 ***	.17***	.30 ***
CHRT-SR e	.12*	.01	.16***	.01	.17***

Note.

^*p < .05,

^**p < .01,

^***p < .001.

Medium effects are bolded. r = .10 – .29 indicates a small effect, r =.30 – .49 indicates a medium effect, and r = .50 – .99 indicates a large effect.

The self-report versions of the below scales were used.

^aDysfunctional Beliefs and Attitudes about Sleep, total and subscale scores.

^bPittsburgh Sleep Quality Index, 5-component score.

^cQuick Inventory of Depressive Symptomatology – Adolescent Version, total score without sleep items.

^dDifficulties in Emotions Regulation Scale – Short Form, total score.

^eConcise Health Risk Tracking Scale, Suicidal Thoughts subscale score.

Aim 2:

We first examined whether all measures changed over treatment. A one-way repeated-measures multivariate ANOVA (MANOVA) was conducted to examine change in outcome measures across treatment. This was followed by a two-way repeated ANOVA on the DBAS subscale scores across time points. Mauchly’s test indicated that the assumption of sphericity had been violated for the interaction effect and main effect of the subscales. Therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity. Lastly, to examine the relationship between subjective sleep quality and DBs over treatment, five bivariate Pearson correlations were then run to examine relationships between the change in DBAS total/subscale scores and the change in PSQI total score. Change scores were calculated as the difference between scores at discharge and scores at admission. Resulting p values were adjusted for multiple testing using the Bonferroni correction.

Results

A demographic breakdown of the Aim 1 sample (n = 419) is shown in Table 1. The score averages for each scale are also shown. The average DBAS score was 0.24 standard deviations above the community sample mean.

Table 1.

Characteristics of the Entire Sample at Program Admission (n = 419)

Variables	M	SD	n	%
Age (12 – 18)	14.71	1.54
Sex at Birth (F : M)			330 : 89	78% : 21%
Race/Ethnicity
White non-Hispanic			209	50%
White Hispanic			108	26%
Black non-Hispanic			46	11%
Asian non-Hispanic			14	3%
Other †			42	10%
Lifetime Attempt (Yes : No)			262 : 157	63% : 37%
Number of Program Sessions Completed	8.89	4.31
DBAS a	4.26	1.92
Consequences	4.41	2.38
Helplessness	3.86	2.27
Expectations	5.39	2.79
Medications	4.06	2.64
PSQI b	6.83	2.64
QIDS c	13.65	4.94
DERS d	57.26	13.38
CHRT-SR e	5.64	3.32

Note.

^†The “Other” category is comprised of races/ethnicities representing < 2% of the Aim 1 sample (< 9 individuals per each of the 12 combined races/ethnicities). Comparisons between inclusions and exclusions from the Aim 2 sample used the full racial/ethnic breakdown.

The self-report versions of the below scales were used.

^aDysfunctional Beliefs and Attitudes about Sleep, total score (first row) and subscale scores.

^bPittsburgh Sleep Quality Index, 5-component score.

^cQuick Inventory of Depressive Symptomatology – Adolescent Version, total score without sleep items.

^dDifficulties in Emotions Regulation Scale – Short Form, total score.

^eConcise Health Risk Tracking Scale, Suicidal Thoughts subscale score.

Aim 1

Significant correlations were found between the DBAS total scale and all outcome measures (PSQI, QIDS, DERS, and CHRT-Suicidal Thoughts) with small to medium effect sizes (Table 2). Significant correlations were also found between the Helplessness and Medications subscales and outcome measures, with small to medium effect sizes. The correlations of the Consequences subscale showed a similar pattern of significance with all outcomes except CHRT-Suicidal Thoughts. Lastly, the Expectations subscale was shown to be only significantly correlated with the DERS and with a small effect size.

Through seven two-tailed z-tests, the Expectations subscale was found to have the weakest correlation among the DBAS subscales with the QIDS (z = −2.71, p < .05) and PSQI (z = −4.31, p < .001) (Table 2). When examining DBAS subscale correlations with the DERS, the Expectations subscale was significantly weaker than the Consequences (z = −2.83, p = .03) and Helplessness subscales, but was not significantly weaker than the Medications subscale. The Expectations subscale correlation with the CHRT-Suicidal Thoughts subscale was not significantly different from any other DBAS subscale correlation with the CHRT-Suicidal Thoughts subscale.

Aim 2

In the one-way repeated-measures MANOVA examining DBAS and the outcome measures between treatment time points, the multivariate main effect of time point was significant with a large effect size, indicating a multivariate decrease in scores between time points, F(5, 191) = 47.71, p < .001, η_p² = .56. Post-hoc pairwise comparisons (Table 3, row 1 and rows 6 – 9) indicated that all outcome measures significantly decreased between time points at p < .001. The PSQI and DBAS total scales decreased with medium effect sizes, while the QIDS, DERS, and CHRT-Suicidal Thoughts scales decreased with large effect sizes.

Table 3.

Changes in DBAS Scales and Outcome Measures Over the Course of Treatment (n = 196)

	Admission		Discharge		Change (ηp2)
Variables	M	SD	M	SD
DBAS a	4.36	1.78	3.93	2.01	.06 ***
Consequences	4.61	2.26	4.09	2.36	.05**
Helplessness	3.87	2.16	3.27	2.27	.08 ***
Expectations	5.52	2.70	5.40	2.64	.00
Medications	4.10	2.57	3.91	2.71	.01
PSQI b	6.72	2.58	5.74	2.57	.16 ***
QIDS c	13.66	4.85	9.00	4.87	.52 ***
DERS d	56.64	12.19	50.89	13.44	.18 ***
CHRT-SR e	5.79	3.24	2.96	2.90	.42 ***

Note.

^**p < .01,

^***p < .001.

Medium to large effects are bolded. η_p² = .01 indicates a small effect, η_p² = .06 indicates a medium effect, and η_p² = .14 indicates a large effect.

The self-report versions of the below scales were used.

^aDysfunctional Beliefs and Attitudes about Sleep, total (first row) and subscale scores.

^bPittsburgh Sleep Quality Index, 5-component score.

^cQuick Inventory of Depressive Symptomatology – Adolescent Version, total score without sleep items.

^dDifficulties in Emotions Regulation Scale – Short Form, total score.

^eConcise Health Risk Tracking Scale, Suicidal Thoughts subscale score.

In the two-way repeated-measures ANOVA examining the DBAS subscales between treatment time points, the interaction effect between time point and subscale was significant with a small effect size, F(2.53, 493.09) = 3.06, p = .04, η_p² = .015. The main effects of time point (F(1, 195) = 8.53, p = .004, η_p² = .042) and subscale (F(2.20, 429.47) = 50.01, p < .001, η_p² = .20) were also significant, with small and large effect sizes, respectively.

After adjustment for multiple testing using the Bonferroni correction, post-hoc pairwise comparisons (Table 3, rows 2 – 5) indicated that the Consequences and Helplessness subscales decreased between time points with small and medium effect sizes, respectively. There were no significant changes in the Expectations and Medication scales between time points.

In the correlations between DBAS total/subscale and PSQI total change scores (Table 4), the DBAS total change score was significantly correlated with the PSQI total change score. Subscale analyses indicated that the Helplessness subscale change score was the only DBAS subscale change score significantly correlated with the PSQI total change score.

Table 4.

Bivariate Pearson Correlations of DBAS Total, DBAS Subscales, and PSQI Total Change Score

ΔPSQI Total	ΔDBAS Consequences	ΔDBAS Helplessness	ΔDBAS Expectations	ΔDBAS Medication	ΔDBAS Total
r	.18*	.37***	−.04	.11	.26***

^*p < .05,

^**p < .01,

^***p < .001

Discussion

In the present study, we explored the relationship between DBs and outcome measures for adolescents in an IOP for suicide prevention and treatment. Our findings revealed significant associations between DBs and self-reported sleep disturbance, depressive symptoms, ED, and SI. We also observed changes in DBs over the course of treatment, and a small but significant association between reduction of DBs and reduction of sleep disturbance.

Several findings were notable in our exploration of the associations between DBs and the other constructs. We compared some of our unexpected findings to those with adults to determine if they were developmentally specific. First, the Expectation subscale demonstrated a consistently weak relationship with the other variables in this study. This scale consists of two items addressing beliefs about the need 1) for eight hours of sleep, and 2) to “catch up” with daytime naps when the threshold for adequate sleep is not achieved. This finding aligns with previous research in adults (Morin et al., 2007). In the paper validating the short form of the scale, Morin and colleagues (2007) found weak discrimination offered by these items between those who report good and poor sleep quality but decided to keep these items because they might “represent important treatment target given their potential role in turning a situational sleep problem into chronic insomnia” (p. 1552). The second notable finding was that the association between DBs and SI was weaker than anticipated – barely reaching a small effect size for the overall scale. Given that there are many potential mediators between DBs at one end of the chain acting via insomnia, to suicidal acts on the other end leading to STBs, the weak association is possibly due to the distal nature of SI as a variable. Furthermore, whereas DBs would be expected to be relatively stable without intervention, SI is a state variable that fluctuates over time. However, in adult psychiatric patients, the association between DBs and SI as measured by the Beck Scale for Suicide Ideation (r = .49) was much stronger than in our study, reaching a medium to large effect size (McCall et al., 2013). Whether this discrepancy is due to a developmental difference in the pathway between insomnia symptoms and SI, or because of different methodology to assess SI, requires further study.

Our finding that DBs decreased over treatment suggests that non-sleep-specific cognitive-behavioral interventions can improve sleep-related beliefs, a finding which is consistent with some (e.g., McCall et al., 2019) but not all (e.g., Morin et al., 2002) literature in adult populations. In our study, change in the DBAS total after treatment was largely driven by the Consequences and Helplessness subscales. It is notable that these two subscales incorporate cognitions that most clearly relate to anxiety and negative mood arousal (e.g., “I worry that I might lose control over my ability to sleep”; “I have little ability to manage the negative consequences of disturbed sleep”). In contrast, Expectations and Medications subscales, which did not exhibit significant changes over time, are rooted in more factual misinformation (e.g., “I believe insomnia is the result of a chemical imbalance”; “I believe I need 8 hours of sleep a night”). The treatment these youth received in the IOP was largely informed by CBT, and thus, there may have been a secondary effect more broadly on the management and improvement of negative mood-provoking cognitions. Change in knowledge might require more specific modification, as might occur in sleep-specific cognitive interventions like CBT-I.

In a related finding, when we examined whether changes in DBs occurred in tandem with changes in subjective sleep, we found a significant correlation between the DBAS and PSQI total change score. Interestingly, in our study, the correlation between change in DBs and change in sleep disturbance was entirely driven by the DBAS Helplessness subscale. It is possible that these subscale beliefs are most amenable to the broad effects of treatment in our adolescent IOP program, where the focus of treatment centers around teaching adolescents to manage their negative affect in order to improve their mood and emotion regulation skills.

It is important to note that although the DBAS is widely used and validated in adults, and has been used in multiple studies with adolescents, it has not been validated in adolescents. Many of our findings parallel the adult literature, although in many cases the associations were weaker in our sample. We note that the scale was not developed in the context of adolescence; given that the items of the scale were proposed based on expert opinion and experience in adults, these items may not overlap with domains that tap into sleep disturbance in adolescents. In considering the influences of negative affect and rumination in this age group, differences in findings may have less to do with developmental differences and more to do with measurement differences. In addition to age-specific manifestations of DBs, changes in relevant contemporary challenges since the original scale was created may result in an underspecified model for measuring DBs in adolescents. For example, the influences of social media and device use on emotional function (Keles et al., 2020) and sleep (Alonzo et al., 2021) in adolescents have been of interest and concern in the literature. Future directions might include the development of an adolescent-specific scale based on cognitions elicited from youth to optimize measurement of DBs.

Whether direct effects to target DBs result in changes in sleep and STBs has yet to be demonstrated. Nevertheless, our findings have some implications for clinicians managing suicidal youth. Although sleep impairments are well recognized in these youth, awareness that they appear to be associated with DBs opens up some possibilities for eliciting and addressing these distortions both in sleep-specific treatment, and for clinicians who may not be experts in CBT-I but use cognitive-behavioral strategies in treatment. Suicidal adolescents who report having a lack of control over sleep may benefit from efforts that reduce helplessness about sleep by eliciting and addressing dysfunctional beliefs, and using psychoeducation strategies to correct them. The intertwined changes we document in beliefs and sleep, and the concurrent associations among DBs and mood, and Suicidal Thoughts encourage clinicians to consider CBT-I for the adolescents they manage who have sleep impairments. Clinicians using the Dysfunctional Beliefs about Sleep (DBAS) scale should be cautious with the Expectations subscale, which has shown poor reliability in both adults and adolescents. However, the Helplessness subscale has shown strong reliability, and we found that adolescents with initial Helplessness subscale elevations demonstrated significant improvements in both the associated beliefs and in sleep over the course of CBT-informed treatment which did not target sleep.

Limitations and Future Directions

Although this study revealed interesting relationships between DBs and various health measures within a large patient sample, limitations must be acknowledged. All measures were self-report, resulting in the biases that result from single method variance. Second, our sample was comprised of mostly female patients (78%) and largely White (about 50%). Third, findings from our sample all come from IOP for adolescents at high risk for suicide, and thus, may not be generalizable to other types of clinical populations. Although this sample is typical for adolescents in treatment for depression and suicidality, future research could benefit from using more diverse populations, along with objective sleep measures. We used a modified scale to assess Sleep Efficiency, and the reliability was not optimal. The modification is consistent with the literature that has noted that the items we removed from the PSQI function poorly in adolescent samples. Research suggests that the Sleep Medication component may not measure sleep quality well in adolescents (Passos et al., 2017). Research also suggests that participants may be prone to making “data errors” (such as confusing AM with PM on the 12-hour clock), which can contribute to erroneous calculation of Sleep Efficiency (Shaffer et al., 2023). Fourth, we did not include a specific measure of insomnia, instead assessing sleep quality more broadly. Finally, we examined all constructs in a design that was not optimal to assess directional or causal questions which could increase our understanding of the potential underlying mechanisms among these constructs.

To our knowledge, the variables that we have reported on in our study have not been tested in a single longitudinal model, though the interrelationships among some of them have been tested in cross-sectional or partial models. One (of possible) logical and sequential pathways is that 1) Depression and 1) DBs impair 2) Sleep Quality which increases 3) Emotional Dysregulation which mediates the effects of poor sleep quality on 4) Suicide ideation. Multiwave, longitudinal data would offer an opportunity to test the robustness of such models, and determine direct, indirect, and more complex relationships among the variables. These findings could then inform future treatment studies. For example, if the association between Depression at the first wave (T1) and Sleep Quality at T3 were largely mediated by DBs at T2, CBT-I might be tested as a first line of treatment to improve sleep outcomes in patients regardless of the severity of their depression. On the other hand, concurrent Depression might act as a moderator of the relationship between DBs at T1 and Sleep Quality at the next wave. DBs influence on Sleep Quality might be stronger when depressive symptoms are present at mild or lower levels and weak at moderate or high levels of depression. In this case, it may be better to aggressively treat moderate and severe depression with a broader range of tools, addressing the sleep impairment that remains only after depression decreases.

Conclusions

Overall, our study showcased the varied nature of DBs, especially in its associations with sleep correlates. This study offers preliminary information specific to this population for whom CBT-I is being increasingly implemented and suggests that targeting DBs may be promising in management of suicidal adolescents. This claim aligns with suggestions from prior research and highlights the potential of DBs in the treatment of patients with sleep disturbance (Morin et al., 2007). Given the variation in the sensitivity of the subscales to associations with related constructs, it may be helpful to develop an adolescent-specific scale that could enhance construct measurement validity. Such research would strengthen studies that seek to determine whether targeting dysfunctional sleep beliefs offers hope for improved outcomes for suicidal adolescents.

Highlights.

• Few studies examine dysfunctional beliefs about sleep (DBs) in suicidal adolescents

• DBs were studied in adolescents enrolled in a suicide prevention program

• DBs were linked to sleep, depression, suicidality, and poor emotion regulation

• Reducing DBs was associated with improved sleep in adolescents

• Addressing DBs may help manage suicidality in adolescents

DATA AVAILABILITY:

The participants of this study did not give written consent for their data to be shared publicly, so due to the sensitive nature of the research supporting data is not available.