A Deeper Dive Into the Relation Between Psychotic-like Experiences and Suicidal Ideation and Behaviors in Children Across the United States

Abstract

Background and Hypothesis

Children who endorse psychotic-like experiences (PLEs) appear to be at a greater risk for suicidal ideation and behavior (SI/SB) compared to their peers who do not endorse PLEs. Despite evidence of differential relations among subtypes of PLEs and SI/SB, the research on which PLE subtypes produce the strongest associations remains mixed. Further, though there is evidence that general psychological distress may help explain the relation between PLEs and SI/SB, no research has investigated the role of distress specific to PLEs in this association.

Study Design

The present study sought to assess the associations among individual Prodromal Questionnaire-Brief Child Version (PQ-BC) items and SI/SB, as well as to explore the role of distress associated with PLEs as a mediator and/or moderator in a demographically diverse sample of children across the United States (N = 11 875).

Study Results

Results revealed that individual items of the PQ-BC may be differentially predictive of lifetime SI (ßs = 0.000–0.098) and SB (ßs=0.002–0.059), even when controlling for sociodemographic variables, internalizing symptoms, and traumatic experiences, with particularly strong associations observed among items indexing thought control, auditory hallucinations, suspiciousness, and nihilistic thinking/dissociative experiences. Item 13, nihilistic thinking/dissociative experiences, displayed the strongest effect sizes. Findings from moderation and mediation models provided evidence consistent with distress as both a partial mediator and moderator of the relation between total PLEs and individual PQ-BC items with SI and SB.

Conclusions

Distress specific to PLEs may be an important modifiable risk factor to target in suicide assessment, prevention, and intervention efforts.

Introduction

Suicide is the second leading cause of death for children in the United States, with 546 deaths among youth ages 8 to 14 in 2019.¹ Around 12% and 4% of early adolescents endorse suicidal ideation (SI) and suicidal behavior (SB), respectively.² Given the rising trend of suicide risk in youth,³ it is critical to understand which modifiable risk factors contribute to the increasing rate of SI and SB in this population. This study investigates associations between a well-documented risk factor, psychotic-like experiences (PLEs),^4–6 with SI and SB in middle childhood using Adolescent Brain Cognitive Development (ABCD) Study data.

PLEs and SI/SB

PLEs are subclinical psychosis symptoms that are generally less distressing and functionally impairing than threshold psychotic symptoms. PLEs are commonly reported in children, with about 10%–17% endorsing some form of PLEs.^7,8 A robust association between PLEs and SI/SB is now well established⁹ in child and adolescent samples,^10,11 including in the ABCD sample.^5,12,13 There is evidence that PLEs, above and beyond other forms of psychopathology, are associated with SI/SB in youth.^4,14–16

Subtypes of PLEs and SI/SB.

The relation between PLEs and SI/SB may differ based on the subtype of PLE endorsed. Perceptual abnormalities, including auditory hallucinatory experiences among community youth and visual distortions among patients at clinical high-risk for psychosis, may be related to SI/SB.^17–19 Other research has shown associations between persecutory ideation and delusional-like experiences with SI/SB over and above relevant demographic and clinical characteristics among community adolescents and adults, as well as hospitalized adolescents.^11,20,21 Although there is some evidence that certain subtypes of PLEs may be more strongly associated with SI/SB, the implicated PLE subtype varies across studies. Mixed findings may be attributable to factors such as sample characteristics (eg, clinical high-risk samples, adolescents in inpatient settings) or not controlling for other PLE subtypes. Additional research is needed to examine which PLEs are most strongly related to SI/SB to understand which symptoms to target in interventions.

The Role of Distress.

The increased risk of SI/SB experienced by individuals across the psychosis-spectrum may be due to distress resulting from novel symptoms.²² There is evidence that PLEs may only be associated with concurrent and future suicide attempts when they are accompanied by general psychological distress among community adolescents,²³ and that distress associated with PLEs may be predictive of a higher risk for SI among college students.²⁴ Psychological distress, however, may not fully account for this complex relation,²⁵ and there is even less clarity on whether distress is a mediator or a moderator of the relation between PLEs and SI/SB.

In a recent systematic review, Hielscher and colleagues²⁶ reported that psychological distress appears to partially account for the relation between PLEs and suicide attempts. These researchers found that psychological distress partially mediated the relation between auditory hallucinatory experiences and self-harm behavior, but fully accounted for the relation between both visual hallucinatory experiences and delusionary experiences with self-harm behavior among community youth.²⁷ Similar findings were reported in a recent longitudinal study with adolescents, where high psychological distress was found to mediate the relation between hallucinatory experiences and SB.²⁸

There is also evidence that distress may moderate the relation between PLEs and SI/SB. Among adolescents endorsing SI at baseline, the interaction between hallucinatory experiences and psychological distress was associated with a nine-fold increased risk of a suicide attempt one year later—though there was not strong evidence for distress mediating the relation between hallucinatory experiences and the transition from SI to SB.¹⁷ Importantly, to our knowledge, previous research has examined general psychological distress, and has not examined associations between distress inherent to PLEs with SI/SB among children.

Current Study

This present study used the demographically diverse ABCD Study sample and is the first to both examine associations between specific subtypes of PLEs and SI/SB, as well as to explore the role that distress specific to PLEs plays in these associations in middle childhood. First, as a hypothesis-driven aim, we examined the unique relations between each PLE item and lifetime SI and SB to determine which PLEs are most strongly related to SI and SB. We hypothesized that nonbizarre ideas, suspiciousness, and perceptual abnormalities would be most strongly related to SI and SB.^{11,17–20,27} Second, as an exploratory aim, we explored evidence for distress associated with PLEs explaining and/or strengthening the relation between PLEs and both SI and SB. Though exploratory, we anticipated that we would find evidence consistent with distress both mediating and moderating the relation between PLEs and both SI and SB, given that previous research examining this relation between general psychological distress has found evidence of both mediation and moderation.^{17,23,26–28}

Methods

Participants

Participants included 9–10-year-old children who participated in the baseline visit of the ABCD study (Data Release 3.0; Acknowledgements; Supplement for study-wide exclusion criteria) from 22 sites across the United States (N = 11 875). The current study examined available baseline data unless otherwise noted from participants with complete SI and SB data (n = 11 785; Supplemental Table 1 for sample sizes and descriptive statistics). The study was approved by a central Institutional Review Board at the University of California, San Diego. Parents and children provided written informed consent and assent, respectively.

Measures

PLEs.

The validated Prodromal Questionnaire-Brief Child Version (PQ-BC) was used to assess PLEs in children.²⁹ This self-report measure is a 21-item questionnaire assessing PLEs using a visual response Likert scale to measure distress. For this study, we examined individual PQ-BC items (0 = no; 1 = yes; table 1 for item text) and total scores, or the sum of items endorsed (Cronbach’s a = 0.865; Supplemental Table 2 for results examining PQ-BC distress scores [ie, total scores weighted by level of distress]). For mediation and moderation analyses, we examined the sum of distress associated with PQ-BC items.

SI/SB.

Lifetime SI and SB were assessed using the Kiddie-Structured Assessment for Affective Disorders and Schizophrenia (K-SADS) for DSM-5,³⁰ consistent with previous research.^5,31 Lifetime SI was calculated as the sum of ten dichotomous youth-rated variables including both past and present endorsement of passive SI (wishing you were dead), active, nonspecific SI (thoughts of killing oneself), active SI with a method, active SI with intent, and active SI with intent and a plan (α = .642). Lifetime SB was calculated as the sum of fifteen dichotomous youth-rated variables, including both past and present endorsement of preparatory behavior, aborted attempt, interrupted attempt, and actual attempts (α = .725).^5,30 See Supplemental Table 3 for individual items.

Demographic Variables.

Race/ethnicity was calculated as a five-level factor variable: White, Black, Hispanic/Latinx, Asian/Asian American, Multiracial/Multiethnic. Financial adversity was calculated by summing seven items that indexed financial hardship.³² We include age, sex, race/ethnicity, and financial adversity as covariates in analyses, given research implicating these variables in previous studies examining SI and SB.^{29,32,33,34–39}

Statistical Analysis

Analyses used hierarchical linear models (HLMs) conducted in R lme4 package,⁴⁰ with family unit and research site modeled as random intercepts, and age, sex, race/ethnicity, and financial adversity included as covariates. Due to significant skew and zero inflation of PLE, SI, and SB scores, we also examined log-transformed scores with consistent results. First, we examined the unique associations with each of the 21 PQ-BC item endorsements in the model as predictors with covariates, with models separately examining SI and SB as the outcome (Variance Inflation Factor variable values ranged from 1.01 to 1.41; Supplemental Table 2 for associations with PQ-BC distress scores; Supplemental Table 4 for each of the PQ-BC items examined in separate models). Multiple comparisons were False Discovery Rate corrected (FDR-corrected; Benjamini and Hochberg false discovery rate of 0.05). Given previous associations between SI/SB, PLEs, and psychopathology,^6,41 we also examined whether unique associations with individual PQ-BC items and both SI and SB remained after accounting for internalizing symptoms and adverse childhood experiences (Supplement). Second, we examined which PQ-BC items showed stronger unique associations with SI and SB compared to each of the other PQ-BC items using Meng’s z-tests.⁴²

Third, we performed mediation analyses using the lavaan package in R⁴³ to explore evidence that the sum of distress associated with PLEs mediates associations between total PLE endorsement and SI/SB. For follow-up mediation analyses, only individual items that were significantly uniquely associated with SI/SB (table 1) were included in models examining individual items as predictors. Fourth, we examined whether there was evidence that distress moderated associations between PLEs and SI or SB by running a model including an interaction between total PLEs and the sum of distress associated with PLEs. For these analyses, we evaluated PQ-BC items with three levels of distress: low distress (ie, no distress associated with PLEs), average distress (ie, at least one PLE rated in 1–2 range of distress), and high distress (ie, at least one PLE rated in 3–5 range of distress). We were not able to examine the interaction between individual PQ-BC items and distress because of the structure of the PQ-BC, as participants who did not endorse a PLE could not report that it was distressing. Instead, we examined whether there was evidence that the associations with SI and SB were stronger for those who report distress versus those who do not among children who endorsed each PQ-BC item (Supplemental Table 5).

Results

PQ-BC Items Associations With SI/SB

Findings revealed that several PQ-BC items were uniquely associated with both SI and SB, with all items and covariates together accounting for 8% of the variance in SI and 3% of the variance in SB (table 1; when modeled individually, all PQ-BC items showed significant positive associations with both SI and SB, Supplemental Table 4). Importantly, significant relations generally remained when using PQ-BC item distress scores (Supplemental Table 2) and when accounting for comorbid internalizing symptoms and ACEs, except that the association between item 10 (losing concentration due to distant sounds not normally aware of) and SI was no longer significant (Supplemental Table 6). Results generally remained consistent when examining estimates from models including all PQ-BC items predicting current SI/SB (Supplemental Table 7).

In general, results from Meng’s Z analyses indicated that the relation between items 13 (feeling like you do not exist or that you are dead) and 18 (feeling mistrustful or suspicious of others) and SI, and item 13 and SB, appear to be significantly stronger than the relation between most other items and SI and SB (figures 1a and 1b). See Supplemental Table 8 for means, standard deviations, and effect sizes for associations between PQ-BC item endorsement with SI and SB.

Distress as a Mediator/Moderator of PLEs and SI/SB

Results portrayed evidence consistent with distress inherent to PLEs indirectly linking relations between total PQ-BC scores and SI and SB (figures 2a and 2b), as well as most individual PQ-BC items that had a significant, unique association (tables 2 and 3 for indirect effects of items; table 1 for item description). There was not strong evidence consistent with distress indirectly linking the relation between item 5 (feeling like you are not in control of your thoughts) and SI nor the relation between items 3, 18, or 20 (things looking differently than normal, feeling mistrustful or suspicious, or seeing things that others do not see) and SB.

Results from a series of moderation models indicated that there was a stronger effect of distress on both SI and SB as total PLE endorsement increased (figures 3a and 3b). Exploratory analyses revealed that generally, relations between PQ-BC items and SI and SB were numerically stronger for children who endorsed distress compared to children who did not (Supplemental Table 5). Our findings provide evidence consistent with distress inherent to PLEs both mediating and moderating relations between total PLEs and SI/SB.

Discussion

This study was the first to examine relations between specific subtypes of PLEs and SI/SB using individual PQ-BC items, as well as the first to explore the role that distress specific to PLEs plays in these associations. Consistent with hypotheses, several individual PQ-BC items were differentially predictive of lifetime SI and SB—in particular, items indexing thought control, auditory hallucinations, and nihilistic thinking/dissociative experiences. Importantly, these results remained when controlling for sociodemographic variables, internalizing symptoms, and ACEs. In our exploratory models, we found evidence consistent with distress inherent to PLEs as both a mediator and moderator of the relation between total PLEs and SI and SB, as well as evidence that distress inherent to certain items were significant mediators. These results provide a greater understanding of the relation between subtypes of PLEs, distress, and SI/SB in children, and support for assessing PLEs when determining suicide risk.

Unique PQ-BC Item Associations With SI/SB

In models examining all PQ-BC items, eleven out of 21 items were uniquely associated with SI and three out of 21 items were uniquely associated with SB. Several items significantly predicted SI but not SB, including items indexing components of visual distortions, delusional ideas, and suspiciousness. In contrast, items measuring thought control, auditory hallucinations, and nihilistic/dissociative experiences were significant predictors of both SI and SB, consistent with some previous research.^11,44,45 Overall, results suggest there may be specificity to the associations between PLEs with SI and SB.

Strength of the Association of PQ-BC Items With SI/SB

Several findings emerged when comparing the strength of associations between each item with SI or SB. First, an item assessing suspiciousness generally showed stronger associations with SI compared to other items. An item indexing unusual thought content or magical thinking also showed stronger associations with SI compared to other items. These findings are consistent with research reporting that unusual thought content and delusional-like experiences are related to SI among older teenagers and young adults in the general population,²¹ although these findings do not necessarily generalize to clinical high-risk samples.^19,46 An item indexing distracting new sounds and an item indexing sensing something is nearby both showed stronger associations with SB compared to other PQ-BC items, consistent with our hypothesis that hallucinatory experiences may display strong relations with SI/SB, and with previous findings including with samples of hospitalized adolescents and youth from the general population.¹⁹ Thus, items indexing suspiciousness, unusual thought content, and hallucinatory experiences were more strongly associated with SI/SB compared to other items, with an item assessing nihilistic ideas or dissociative experiences showing the strongest associations with both SI and SB. These findings suggest that these particular forms of PLEs may be especially relevant when assessing suicide risk among youth in the general population.

Strength of the Association of Item 13 With SI/SB

Item 13, assessing nihilistic ideas or dissociative experiences, stood out as a relevant PLE for both SI and SB, consistent with literature examining this association among adolescents who were referred for mental health treatment.⁴⁵ This association has also been explored in philosophical and sociological writings.^47–49 There are several possibilities for why this item was particularly strongly associated with SI and SB. First, there may be a clinically meaningful link between nihilistic thinking and/or unusual thought content and SI and SB. The Structured Interview for Psychosis-Risk Syndromes⁵⁰ contains a similarly worded item within the unusual thought content section,⁵¹ indicating our results are potentially consistent with previous evidence for an association between unusual thought content and SI/SB among community youth and adults²¹ (although this has not been demonstrated in clinical high-risk samples^19,46). Although speculative, there may be something more severe or distressing about nihilistic ideas that may contribute to the strength of this relation, consistent with our finding evidence that distress inherent to nihilistic ideas may mediate relations between nihilistic ideas and both SI and SB.

Second, though item 13 is intended to measure nihilist thinking and/or unusual thought content, this item likely taps into other psychopathological domains, including dissociation, consistent with literature linking dissociative experiences and suicide risk.^45,52,53 Dissociation may attenuate fear and pain, which can contribute to risk for SI and SB due to increasing the capability for suicide.^52,54 It is also possible that dissociative experiences increase fear and anxiety, which in turn could increase suicide risk. Associations between dissociation and SI/SB may also be partially attributable to other psychopathology variables (eg, posttraumatic stress symptoms).⁵³ For instance, dissociation can be a quintessential symptom reported by those who have experienced ACEs or trauma,^55,56 which is common among children with PLEs.^57,58 Indeed, trauma may be a shared risk factor of both PLEs and SI/SB among college students.^24,59

Third, item 13 may be endorsed by individuals experiencing passive SI. As this question asks participants whether they feel as if they are “dead”, similarity in constructs and language may contribute to the observed association.⁶⁰ Two evidence-based theories, the Three-Step Theory and the Interpersonal Theory of Suicide posit that deaths by suicide may result in part from acquired habituation to death or pain.^61,62 Thus, recurrent thoughts of death or familiarity with death may increase risk for SI and SB. Overall, future research should clarify the nature of associations between this item and SI/SB to help apply findings to assessment and intervention efforts.

Distress as a Mediator/Moderator of PLEs and SI/SB

Findings from our exploratory analyses provide evidence that distress inherent to PLEs plays a role in relations between PLEs and SI/SB, supporting previous findings for general psychological distress as both a mediator and moderator of relations between PLEs and SI/SB among a number of community samples.^{26,28,44,63,64} Specifically, we found evidence that distress related to PLEs mediated the relation between all items analyzed and SI—except a thought control item—and all items analyzed with SB—except several items indexing certain types of visual distortions, suspiciousness, and disorganized communication. In particular, distress inherent to PLEs was found to indirectly link items indexing auditory hallucinations, sensing that a person or force is around, and dissociative experiences to both SI and SB, consistent with work findings hallucinatory experiences and SI and SB are mediated by general psychological distress.^26–28 To our knowledge, however, this is the first study to find that distress inherent to PLEs accounts for a portion of associations between other subtypes of PLEs and SI/SB among children. This is potentially critical for assessment and treatment, as this could imply that mitigating distress may be an actionable target for reducing SI and SB among individuals endorsing these PLE subtypes.

Exploratory analyses also found evidence of moderation: there was a steeper slope for the association between PLEs with both SI and SB for those endorsing PLEs with higher distress in contrast to those endorsing PLEs with average and lower distress. Though preliminary, our findings suggest that associations between PLEs and SI/SB are stronger in the context of high distress than in the context of average or low distress. For those with lower distress, however, there are still associations between PLEs and SI/SB, albeit weaker. This was also evident in sensitivity analyses predicting SI among individuals who endorsed specific PQ-BC items (Supplemental Table 5). This indicates that although distress inherent to PLEs may be a modifier of the strength of the relation between PLEs and SI/SB, this relation (albeit attenuated) may exist outside the context of distress. Previous research has found that subtypes of psychosis-spectrum symptoms (eg, hallucinatory experiences, hallucinations, delusions) may be moderated by distress in both community and psychosis samples,^17,63,64 though no studies have examined subtypes at this granular level. One explanation for our findings is that suicide risk may be particularly heightened for the people endorsing distress with certain PLE subtypes, including nihilistic ideas and hallucinatory experiences. This is consistent with research with clinical-high risk samples showing that SI/SB is linked to psychosis-risk symptoms (ie, visual distortions,¹⁹ suspiciousness)^46,65 and suicide risk factors that are nonspecific to psychosis (ie, stigma, social isolation).⁶⁶

Our findings from the mediation and moderation models provide evidence that distress inherent to PLEs may partially explain associations between PLEs and SI/SB, but also that in the context of high distress the association between PLEs and SI/SB is strengthened. Though it is clear that distress inherent to PLEs is relevant, additional longitudinal research is needed to parse these pathways. It is possible that distress is uniquely influential during middle childhood, perhaps in part due to maturational changes (eg, neural, hormonal) influencing these associations. Future research should longitudinally examine the role that age, puberty, and the development of psychotic experiences (eg, clinical high-risk) play in relations between PLEs and SI/SB.

Limitations

In terms of limitations, first, although it is still informative to conduct mediation analyses on cross-sectional data,⁶⁷ causal inferences are limited. Second, distress inherent to PLEs was recent/current, while SI and SB were assessed as lifetime. This study utilized lifetime SI and SB because focusing on recent SI and SB would have limited analyses due to the lower frequency of items endorsed. Nonetheless, we examined these results in the context of recent SI/SB and found similar results (Supplemental Table 7). Third, we decided to include covariates based on their relevance to analyses, though we were unable to fully address the importance of some of these variables due to the scope of this study. In particular, race/ethnicity may play a role in the association between PLEs and SI/SB due to the documented effects of systematic racism on health disparities in the United States. See Supplemental Table 9 for analyses exploring race/ethnicity groups as moderators of associations between total PLEs with SI and SB (note, though it is possible that race/ethnicity may play a role in the relation between PLEs and SI/SB, in this sample, race did not moderate the relation between total PLEs and SI/SB).⁶⁸ Additional research is needed to examine the relation between subtypes of PLEs using longitudinal and qualitative data to further understand the nuances of these complex relations, and to examine how these relations evolve with age, including the degree to which associations change with worsening of psychotic experiences and associated distress (eg, clinical high-risk). These limitations notwithstanding, this study utilized a large, diverse sample to examine the unique associations between individual PLEs and SI/SB using a validated measure of PLEs.

Clinical Implications and Future Directions

Findings provide evidence for the assessment of PLEs when determining suicide risk in children and early adolescents,^4,14,69 and when identifying which symptoms to target in prevention, assessment, and treatment of suicide risk in this population. In particular, nihilistic thinking and/or dissociative experiences may be an important, modifiable risk factor to explore in future research by utilizing a comprehensive measure of dissociative experiences and thoroughly examining associations between PLEs, ACES, and SI/SB. Though more research is necessary to determine the best clinical approach, clinicians could note the distress level related to PLEs when assessing and treating suicide risk in children with PLEs. Clinicians could also be mindful that certain types of PLEs may be more strongly related to SI/SB than others (eg, nihilistic ideas/dissociative experiences, unusual thought content, and auditory hallucinations).

Conclusion

To our knowledge, this is the first study to examine the relation between individual PQ-BC items and SI/SB in children, as well as to clarify the role of distress inherent to PLEs. Findings suggest that individual PQ-BC items may be differentially predictive of lifetime SI and SB, even when controlling for sociodemographic variables, internalizing symptoms, and traumatic experiences. Though findings are preliminary, this study helps identify which subtypes of PLEs may be clinically meaningful, like unusual thought content, auditory hallucinations, suspiciousness, and nihilistic thinking/dissociative experiences. Distress inherent to PLEs appears to play a role, as results were consistent with distress as both a partial mediator and moderator of the relation between PLEs with SI and SB. Distress specific to PLEs may be a particularly relevant modifiable factor to target in suicide assessment, prevention, and intervention efforts.

Funding

This work was supported by the National Institute of Mental Health (K23MH121792-01). Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development (ABCD) Study (https://abcdstudy.org), held in the NIMH Data Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 children aged 9-10 and follow them over 10 years into early adulthood. The ABCD Study is supported by the National Institutes of Health and additional federal partners under award numbers U01DA041022, U01DA041028, U01DA041048, U01DA041089, U01DA041106, U01DA041117, U01DA041120, U01DA041134, U01DA041148, U01DA041156, U01DA041174, U24DA041123, U24DA041147, U01DA041093, and U01DA041025. A full list of supporters is available at https://abcdstudy.org/federal-partners.html. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/wp-content/uploads/2019/04/Consortium_Members.pdf. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time. The ABCD data used in this report came from DOI 10.15154/1519007.

Acknowledgments

The Authors have declared that there are no conflicts of interest in relation to the subject of this study. S.Y.J., J.S., and N.R.K. developed the study concept. N.R.K. analyzed the data and created the figures. S.Y.J. and N.R.K. drafted the manuscript. N.R.K. supervised the writing of the manuscript. J.S., R.G., K.O., M.K., and J.D. provided critical feedback. All authors approved the final version of the manuscript for submission.

References