Psychotic experiences, suicidality and non-suicidal self-injury in adolescents: Independent findings from two cohorts

A


Introduction
Psychotic experiences, including hallucinations and delusions, are common in children and adolescents from the general population (Kelleher et al., 2012;Maijer et al., 2018). Consistent with elevated suicide rates among adults with psychotic disorders (Laursen et al., 2019;Tanskanen et al., 2018), converging evidence suggests that psychotic experiences and suicidality often co-occur (Honings et al., 2016;Kelleher et al., 2013). A recent meta-analysis of prospective studies in adolescents and adults reported that psychotic experiences are associated with a two-fold increased odds of future suicidal ideation, a threefold increased odds of future suicide attempts, and a four-fold increased odds of suicide (Yates et al., 2019).
However, it is unclear whether the association between psychotic experiences and suicidality is causal or arises from shared risk factors (Hielscher et al., 2018), such as pre-existing mental health problems (Bolhuis et al., 2018;McGrath et al., 2016;Nock et al., 2013;Orri et al., 2020) or childhood trauma (Croft et al., 2019;Fisher et al., 2013;Gawęda et al., 2020;Geoffroy et al., 2016). Furthermore, given that prior studies have generally lacked baseline measures of suicidality, it is difficult to establish the temporal nature of the association. Therefore, it might also be plausible that reverse temporality exists, i.e., that suicidality is associated with the onset of later psychotic experiences (Murphy et al., 2018). Moreover, given that most previous studies involved adolescents or adults, the dynamic expression of psychotic experiences throughout early adolescence in the context of suicidality risk is unclear.
Additionally, little is known about the association between psychotic experiences and distinct expressions of self-harm, e.g., suicidal behavior versus non-suicidal self-injury (NSSI) (Hielscher et al., 2019). It is currently unclear whether psychotic experiences are differentially related to specific characteristics of NSSI, including frequency and severity. Manifestations of NSSI, such as cutting, severe scratching, and banging or hitting oneself, are important to assess given their varying degrees of tissue damage (Glenn and Klonsky, 2011). Such knowledge is crucial for a better understanding of the association between psychotic experiences and self-harm, and the consequent potential clinical implications.

Aims
We used data from two cohorts in the Netherlands with youth of comparable ages: a population-based cohort (Generation R Study) and a cohort of adolescents oversampled on their high risk of psychopathology (iBerry Study). The main research question was whether psychotic experiences were associated with suicidality above and beyond shared risk factors, such as sociodemographic characteristics, adverse life events, and emotional and behavioral problems. These potential confounders were harmonized across the two cohorts. While the overall cohort designs are different, they complement each other in meaningful ways: the Generation R Study has longitudinal data with the limitation of relatively brief questionnaires; the iBerry Study has a more detailed assessment of suicidality and psychotic experiences but the limitation of cross-sectional data. We investigated shared and study-specific research questions to provide a better characterization of the association between psychotic experiences and suicidality across general population and high-risk settings.
First, in the general population sample, we examined longitudinal associations between hallucinatory experiences and suicidality from childhood to adolescence. In addition, we investigated the persistence, remittance, and incidence of hallucinatory experiences in relation to suicidality. Second, in the risk-enriched sample, we examined the association between psychotic experiences and suicidality while adjusting for the same potential shared risk factors as in the general population sample. Furthermore, we examined associations between psychotic experiences and frequency and tissue damage of NSSI.

The Generation R Study
Generation R is a population-based prospective cohort from fetal life onwards (Kooijman et al., 2016). In the period 2002-2006, all pregnant women living in Rotterdam, the Netherlands, were eligible for inclusion, and approximately 61 % of the eligible population was included at baseline (N = 9778). Baseline data on hallucinatory experiences and suicidality were available for 4330 children aged 10 years, of whom 3435 (79.3 %) had follow-up data on these measures at age 14 years (see Supplementary Information for the attrition analysis). Delusional experiences were assessed in a sub-sample at age 14 years only (N = 2875). We conducted sensitivity analyses using mother-reported data on suicidality, which were available for 3275 children.

The iBerry Study
The iBerry Study is a population-based prospective cohort designed to investigate the transition from subclinical psychiatric symptoms into psychiatric disorders in the greater Rotterdam area, the Netherlands (Grootendorst- van Mil et al., 2021). In the period 2014-2016, 16,736 first-year secondary-school adolescents completed the Strengths and Difficulties Questionnaire-Youth (SDQ-Y) (van Widenfelt et al., 2003). The SDQ-Y includes subscales measuring emotional problems, conduct problems, hyperactivity/inattention, peer problems, and prosocial behavior. A high-risk group was formed from adolescents that scored among the highest 15 % SDQ-Y total scores, and a subsample was randomly selected from the 85 % lowest-scoring adolescents. The highrisk group was oversampled by a 2.5:1 ratio in the final cohort of 1022 adolescents. While the SDQ-Y does not include items on psychotic experiences, suicidality, and NSSI, the oversampling strategy likely resulted in a higher prevalence of these experiences because of their cooccurrence with emotional and behavioral problems. The present study used cross-sectional baseline data on psychotic experiences, suicidality, and NSSI available for N = 910 adolescents (89.0 %) aged 15 years (see Supplementary  At mean ages 10 and 14 years, hallucinatory experiences were assessed using two items from the Youth Self-Report (YSR): "I hear sounds or voices that other people think aren't there" and "I see things that other people think aren't there" (Achenbach and Rescorla, 2001). The items referred to the preceding six months and were rated on a three-point Likert scale: "not at all" (0), "a bit/sometimes" (1), and "clearly/often" (2). In line with our previous study (Steenkamp et al., 2021), the sum score was dichotomized into absence (score = 0) and presence (score > 0) of hallucinatory experiences due to high skewness. We also created four groups with different patterns of hallucinatory experiences over time: 1) absent (hallucinations at neither time point), 2) persistent (hallucinations at both time points), 3) remittent (hallucinations only at baseline), and 4) incident (hallucinations only at follow-up) (Steenkamp et al., 2021).

Delusional experiences.
At mean age 14 years, delusional experiences were assessed by six self-report items extracted from the Kiddie Schedule for Affective Disorders and Schizophrenia -Lifetime Version (K-SADS; Table S1) (Adriaanse et al., 2015;Kaufman et al., 1997). Three response options were available: "no" (0), "yes, probably" (1), and "yes, certainly" (2). Total scores were calculated by summing all items and dichotomized into absent (score < 3) and present (score ≥ 3) due to the skewed nature of the data. The scale had an internal consistency of α = 0.58.
2.2.1.3. Suicidality. At mean age 14 years, self-reported suicidality was assessed using two items from the YSR (Achenbach and Rescorla, 2001): "I think about killing myself" and "I deliberately try to hurt or kill myself." Similar to psychotic experiences, the sum score of the two items was dichotomized into absence (score = 0) and presence (score > 0) of suicidality. Given the young age of the participants during the age 10 assessment, self-reports were restricted to the item "I have thought about hurting myself," which was used as a proxy for suicidality. Selfharm ideation was recorded as absent (score of 0, "not at all") or present (score of 1, "a bit," or 2, "clearly"). Mother-reported suicidality was assessed using two items from the Child Behavior Checklist (CBCL/ 6-18) at mean ages 10 and 14 years (Achenbach and Rescorla, 2001), which were identical to the self-report items and handled in the same way in our analyses.

Psychotic experiences.
Hallucinatory and delusional experiences were assessed using two subscales of the Prodromal Questionnaire-16 (PQ-16), respectively 1) perceptual abnormalities/ hallucinations (nine items) and 2) unusual thought content, delusional ideas, and paranoia (five items) (Ising et al., 2012). All items were scored as "disagree" (0) or "agree" (1) and summed to a total subscale score. For descriptive purposes, we used a cut-off score of ≥5 on the 14item scale to identify adolescents as clinical high-risk for psychosis (adjusted from the original cut-off score of ≥6 on the 16-item scale) (Ising et al., 2012).

Suicidality.
We used four items from the Questionnaire on Suicide and Self-harmshort version (VOZZ-10): "I have thought about ending my life", "I have attempted suicide", "In the past seven days, I thought about ending my life", and "In the past seven days, I thought that suicide would be a solution for my problems" (Kerkhof, 2016). Items were scored on a five-point Likert scale ranging from "never/no days" (1) to "often/every day" (5) and summed to a total score.

Non-suicidal self-injury (NSSI).
The self-report Inventory of Statements about Self-Injury (ISAS) assesses self-injurious behaviors without lethal intent (Klonsky and Glenn, 2009). Adolescents reported the lifetime frequency of ten listed behaviors, which were summed across methods into a total score. Furthermore, data on methods were classified into three groups based on their potential for causing tissue damage, with scores of "1" (least severe: superficial tissue damage, e.g. biting, pinching), "2" (moderate: bruising or light tissue damage, e.g., banging or hitting self) or "3" (most severe: severe tissue damage, e.g. cutting, burning) (Whitlock et al., 2008). A total tissue damage score (range 1-19) was calculated by summing scores for all endorsed behaviors.

Covariates
We included the following covariates in our analyses: sex, age, national origin (Dutch/Non-Western/Other-Western), maternal educational level (low/middle/high), maternal self-reported psychopathology, child non-verbal IQ, mother-reported adverse life events of the child, and self-reported internalizing and externalizing problems. Most covariates were similar in the two samples. A more detailed description of the covariates and their collection methods can be found in the Supplementary Information.

Statistical analyses
In the general population sample (Generation R Study), we first conducted logistic regression to examine the association between hallucinatory experiences at age 10 years and suicidality at age 14 years, while adjusting for self-harm ideation at baseline (age 10 years). Second, we conducted logistic regression to investigate potential reverse temporality. More specifically, we examined the association between selfharm ideation at age 10 years and hallucinatory and delusional experiences at age 14 years, while adjusting for hallucinatory experiences at age 10 years. Delusional experiences were assessed only at age 14 years, and could therefore be studied only as an outcome. Third, we examined associations between longitudinal patterns of hallucinatory experiences (i.e., absent, persistent, remittent, and incident) and suicidality at age 14 years. Fourth, in a sensitivity step, we repeated all analyses using mother-reported suicidality.
In the risk-enriched sample (iBerry Study), we examined crosssectional associations of hallucinatory and delusional experiences with 1) suicidality, 2) NSSI frequency, and 3) NSSI tissue damage. We used multivariate linear regression because these were continuous outcome variables (rather than dichotomous outcome variables as in the general population sample). All continuous outcomes were square root transformed and subsequently converted to z-scores.
In both samples, analyses were adjusted for covariates in a step-wise manner. First, we included the covariates sex, age, national origin, maternal education level, non-verbal IQ, and maternal psychopathology (Model 1). Subsequently, in three separate steps, we additionally adjusted for 1) adverse life events, 2) externalizing problems, and 3) internalizing problems. In a final step, we adjusted for all covariates simultaneously (Model 2). Missing data on covariates (not exposures) were handled using multiple imputation by chained equations (MICE), and estimates were pooled using Rubin's rules (Rubin, 2004;Van Buuren and Groothuis-Oudshoorn, 2011). Statistical analyses were performed in R version 4.1.0 (Generation R Study) and SPSS version 25 (iBerry Study).  Table 1 summarizes the study sample characteristics. In the general population sample (Generation R Study), the prevalence of hallucinatory experiences decreased from 31.9 % to 11.8 % between the ages of 10 and 14 years. Delusional experiences were reported by 11.3 % of adolescents aged 14 years (no data available at age 10 years). Self-harm ideation was reported by 21.2 % of children aged 10 years. Suicidality was reported by 3.9 % of adolescents aged 14 years, and more often by girls than boys (4.7 % vs. 3.1 %).

Sample characteristics
Using identical questionnaires, in the risk-enriched sample (iBerry Study), hallucinatory experiences were reported by 18.4 % and suicidality was reported by 10.5 % of adolescents aged 15 years. When using more detailed instruments, 22.4 % of adolescents were identified as clinical high risk for psychosis and 32.6 % reported NSSI (girls 39.2 %; boys 25.7 %).

The co-occurrence between psychotic experiences and suicidality
At age 10 years, the prevalence of self-harm ideation was nearly three times higher in children with hallucinatory experiences compared to children without hallucinatory experiences (38.1 % vs. 13.2 %; Table S2). At age 14 years, the prevalence of suicidality was nearly five times higher in adolescents with hallucinatory or delusional experiences compared to adolescents without hallucinatory or delusional experiences (13.0 % vs. 2.7 % and 12.6 % vs. 2.7 %, respectively; Table S2).

The longitudinal association between hallucinatory experiences and suicidality
Children who reported hallucinatory experiences at age 10 years had a nearly two-fold increased odds of suicidality at age 14 years, even after adjusting for self-harm ideation at baseline, sociodemographic characteristics, non-verbal IQ, and maternal psychopathology (Table 2). This association attenuated, but remained, after further adjustment for adverse life events and internalizing and externalizing problems at baseline (OR = 1.62, 95 % CI 1.11,2.38; Table 2). Consistent with selfreported suicidality, hallucinatory experiences were prospectively associated with mother-reported suicidality (OR = 2.34, 95 % CI 1.44,3.81; Table S3). This association attenuated in the fully adjusted model (OR = 1.65, 95 % CI 0.98,2.79; Table S3).

Patterns of hallucinatory experiences in relation to suicidality
Adolescents with persistent or incident, but not remittent, hallucinatory experiences exhibited a higher suicidality risk at age 14 years compared to adolescents without hallucinatory experiences (Table 2). In the fully adjusted model, adolescents with persistent hallucinatory experiences had a greater burden of suicidality than adolescents with remittent hallucinatory experiences (OR = 5.08, 95 % CI 3.00,8.58). These findings were consistent in sensitivity analyses using motherreported suicidality (Table S3).

The longitudinal association between self-harm ideation and psychotic experiences
After adjustment for covariates and hallucinatory experiences at baseline, children with self-harm ideation at age 10 years reported more hallucinatory and delusional experiences at age 14 years (Table 3). These associations attenuated by further adjusting for adverse life events, internalizing problems, and externalizing problems at baseline (OR hallucinations = 1.29, 95 % CI 0.99,1.67; OR delusions = 1.37, 95 % CI 1.03,1.82; Table 3). We found no evidence for an association between mother-reported suicidality and subsequent hallucinatory or delusional experiences (Table S4). Results from the Generation R sample (N = 3435). Model 1 is adjusted for age, sex, national origin, maternal education, non-verbal IQ, maternal psychopathology, and self-harm ideation (as proxy for suicidality) at age 10 years. Model 2 is additionally adjusted for adverse life events and baseline internalizing and externalizing problems (mutually adjusted). a Suicidality is reported by n = 134 (3.9 %). b Hallucinatory experiences are reported by n = 1096 (31.9 %), "no hallucinatory experiences" is the reference category.

Table 3
Longitudinal associations of self-harm ideation with hallucinatory experiences and delusional experiences in the general population sample. Results from the Generation R sample (N = 3435 for hallucination sample and N = 2875 for delusion sample). Model 1 is adjusted for age, sex, national origin, maternal education, non-verbal IQ, maternal psychopathology, and hallucinatory experiences at age 10 years. Model 2 is additionally adjusted for adverse life events and baseline internalizing and externalizing problems (mutually adjusted). a Hallucinatory experiences are reported by n = 407 (11.8 %). b Delusional experiences are reported by n = 326 (11.3 %). c Self-harm ideation is reported by n = 728 (21.2 %; hallucination sample) and n = 612 (21.3 %; delusion sample), "no self-harm ideation" is the reference category.

Discussion
We found convincing evidence for associations of psychotic experiences with suicidality and NSSI in young adolescents. Several key findings emerged. First, hallucinatory experiences in childhood were prospectively associated with increased risk for suicidality in adolescence, even when adjusted for self-harm ideation at baseline and various covariates, including childhood adversity and pre-existing mental health problems. Second, adolescents with persistent and incident, but not remittent, hallucinatory experiences exhibited a higher risk for suicidality than adolescents without hallucinatory experiences. Third, childhood self-harm ideation was prospectively associated with a higher burden of psychotic experiences, albeit only for self-report. Fourth, in the risk-enriched sample, psychotic experiences were associated with suicidality as well as with a higher frequency and more extensive tissue damage of NSSI.
Our longitudinal findings extend prior research on the association between psychotic experiences and suicidality (Hielscher et al., 2018;Kelleher et al., 2013;Martin et al., 2015;O'Hare et al., 2021;Yates et al., 2019) in a sample of (pre-)adolescents. The studied age range is a particularly critical period because suicidality and NSSI typically emerge during the early teenage years (Jacobson and Gould, 2007;Voss et al., 2019). Consistent with findings from the population-based sample, hallucinatory experiences were cross-sectionally associated with suicidality in the risk-enriched sample of adolescents aged 15 years. Furthermore, in line with prior research (Hielscher et al., 2021a;Martin et al., 2015;Rimvall et al., 2020), persistentbut not remittenthallucinatory experiences were associated with suicidality. This suggests that transient childhood hallucinations may not be a substantial risk factor for future suicidality, which is supported by increasing evidence of a greater severity and clinical relevance of persistent psychotic experiences (Karcher et al., 2022a;Linscott and van Os, 2013;Steenkamp et al., 2021). Like persistent psychotic experiences, incident psychotic experiences that emerged during adolescence were associated with a substantially increased risk of suicidality, indicating that their contemporaneous presence may be particularly important in clinical settings. Nevertheless, persistent childhood-onset psychotic experiences may reflect different psychopathological and developmental vulnerabilities than adolescent-onset psychotic experiences (Steenkamp et al., 2021). Finally, the associations of hallucinatory and delusional experiences with a more frequent occurrence of NSSI and a greater degree of tissue damage inflicted by NSSI suggest that psychotic experiences may be clinical predictors of NSSI severity.
Additionally, we found evidence in support of reverse temporality as self-harm ideation was prospectively associated with a higher risk of psychotic experiences. This finding is consistent with a six-year followup study that observed bidirectional associations between self-injurious behaviors and psychotic experiences from early to late adolescence (Murphy et al., 2021), as well as a recent study that reported such bidirectional associations between the ages of 14 and 16 years (Stanyon et al., 2022). In contrast, another study found that suicide attempts and NSSI at age 14 years were not associated with incident psychotic experiences two years later (Hielscher et al., 2021a). In the present study, the association with psychotic experiences was observed for self-reported self-harm ideation, but not for mother-reported suicidality. Consistent with prior observations that parents tend to underreport their children's suicidality (Jones et al., 2019), only few mothers (1.6 %) reported suicidality of their children, which may have resulted in insufficient statistical power to detect an association. Furthermore, the association with Table 4 Cross-sectional associations of hallucinatory and delusional experiences with suicidality and non-suicidal self-injury (NSSI) in the risk-enriched sample. Results from the iBerry Study sample. Model 1 is adjusted for age, sex, national origin, maternal education, non-verbal IQ, and maternal psychopathology. Model 2 is additionally adjusted for adverse life events and co-occurring internalizing and externalizing problems (mutually adjusted).
self-reported self-harm ideation may have suffered from shared-method variance, because psychotic experiences were also assessed through selfreport. Finally, the discrepancy may be explained by the different exposure measures, i.e., maternal assessment of suicidal ideation and self-harm compared to self-report assessment of ideation about selfharm. Youth who engage in self-harm may differ from those who think about self-harm (O'Connor et al., 2012), and thoughts of harming oneself can involve various cognitive, coping, and emotional regulatory functions (Laye-Gindhu and Schonert-Reichl, 2005), which might partially overlap but are not equivalent to those of suicidality (Butler and Malone, 2013). Taken together, more well-powered longitudinal studies are needed to draw conclusive results about the bidirectional nature of the association between psychotic experiences and suicidality and self-harm. A causal relationship between psychotic experiences and suicidality and NSSI could exist if suicidality or NSSI emerges or worsens due to command hallucinations (Hielscher et al., 2018;Kelleher et al., 2013). This is supported by a recent study exploring the content of psychotic experiences in adolescents with co-occurring suicidality, in which over one-third of psychotic experiences were characterized by suicidal/selfinjurious or death-related themes and content (Murphy et al., 2021). Alternatively, psychotic experiences may lead to suicidality and NSSI via indirect pathways, such as through distress, anxiety, or depression (Hielscher et al., 2021b;Jang et al., 2014), or through social isolation or stigma (Lehmann et al., 2020). A bidirectional relationship may exist if negative consequences of engaging in self-harm, such as feelings of guilt, shame, and low self-esteem (Bachtelle and Pepper, 2015) exacerbate psychiatric problems and lead to the development of psychotic experiences. In addition, self-harm could be considered a stressful or traumatic event (Stanyon et al., 2022), which, in turn, could increase the risk of psychotic experiences (Trotta et al., 2015). Furthermore, the severe distress and internal threat caused by self-harm may trigger psychotic experiences (Murphy et al., 2018). According to this 'suicidal drive hypothesis,' psychotic experiences may emerge as an externalizing, adaptive strategy to regulate the internal threat of self-harm and protect oneself.
Rather than explained by a causal framework, evidence for a bidirectional association could indicate that psychotic experiences and suicidality share common underlying factors, including psychological distress (Martin et al., 2015), genetic or neurodevelopmental vulnerabilities (Hielscher et al., 2018), borderline personality traits (D'Agostino et al., 2019), impulsivity or emotional dysregulation (Grattan et al., 2021), social dysfunction (Chau et al., 2019), poor coping skills (Lin et al., 2011), or low self-esteem (DeVylder et al., 2015Yates et al., 2019), at least some of which may not have been fully captured by our set of covariates. For example, the recently posed 'neural efficiency threshold model' of psychotic experiences suggests that impaired efficiency of cooperation between brain hubs, such as those involved in problem-solving skills, may contribute to the emergence of both psychotic experiences and suicidal behavior (Kelleher and Cannon, 2021). Although we adjusted for non-verbal IQ in our analyses, neural efficiency might be better assessed by measures of executive functioning (e. g., processing speed), cerebral white matter integrity, or functional brain connectivity.
Future studies are needed to investigate whether the prevention or reduction of psychotic experiences would in turn lead to a reduction in suicidality as suggested by recent observational studies (Yates et al., 2019). However, even if psychotic experiences are not causally related to suicidality, they can still be important clinical predictors of suicide risk. Given that psychotic experiences are likely an indicator of the severity of underlying psychopathology and distress, the co-occurrence with suicidality may indicate a greater experienced burden of illness, warranting clinical attention. We observed that adolescents with psychotic experiences had a nearly five times higher prevalence of cooccurring suicidality than adolescents without psychotic experiences, highlighting the importance of assessing these experiences in suicide risk assessments in clinical practice. This is supported by the metaanalysis by Yates et al. (2019) who reported that the populationattributable fraction of psychotic experiences for suicide attempts and suicide is ~25 %. In fact, preliminary evidence suggests that psychotic experiences, particularly in the context of distress, may contribute to the transition from suicidal thoughts to suicidal behavior (Hielscher et al., 2020;Karcher et al., 2022b). Furthermore, among subtypes of psychotic experiences, several studies found that auditory hallucinations were most strongly associated with suicidality (Hielscher et al., 2021a;Núñez et al., 2018), emphasizing that these experiences may be particularly useful to assess in clinical practice. Conversely, expressions of self-harm may be an indicator of future risk for severe psychopathology including psychotic and bipolar disorders . The above underlines that clinicians should be aware that psychopathology in adolescence is often multifaceted and requires personalized treatment.
An important strength of our study is the integrated use of a longitudinal population-based cohort and a risk-enriched cross-sectional study of similarly-aged adolescents. Furthermore, both samples included the same comprehensive range of potential confounders to carefully assess their role in the association between psychotic experiences and suicidality/NSSI. However, several limitations should be discussed. First, the assessment of suicidality and psychotic experiences in the population-based sample was restricted to brief questionnaires. The delusion scale had a low internal consistency, although Cronbach's alpha tends to underestimate this if scales have fewer items. In addition, our self-report measure of self-harm ideation at age 10 years may not have been an optimal proxy for baseline suicidality, which is supported by the high prevalence of self-harm ideation observed (21.2 %). Furthermore, due to the phrasing of one question of the age 14 suicidality assessment in the Generation R Study, some children may have been included when they endorsed (thoughts of) NSSI without suicidal ideation. However, in the risk-enriched sample, more comprehensive assessments of suicidality, NSSI, and psychotic experiences were available, allowing for an extension of the observations found in the prospective sample. Second, the analyses of the risk-enriched sample relied on cross-sectional data, precluding a direct comparison/replication of the findings from the Generation R Study. Future waves of the iBerry Study will permit the assessment of longitudinal associations of psychotic experiences with suicidality and NSSI across middle and late adolescence. Third, the adjustment for emotional and behavioral problems can be regarded as overcorrection in the cross-sectional sample as the temporal sequence of symptoms is difficult to establish, and therefore, these problems can be considered a mediator rather than a confounder in the association between psychotic experiences and suicidality/NSSI.

Conclusion
In conclusion, we found that children with hallucinatory experiences are more likely to express suicidality during adolescencethis association was independent of suicidality at baseline and a range of shared risk factors. We also found tentative support for reverse temporality, suggesting that vulnerabilities shared between psychotic symptomatology and suicidality could explain their bidirectional association. Our findings also suggest that psychotic experiences are associated with specific NSSI attributes, including frequency and tissue damage, which could be of relevance to clinicians. Findings were consistent across crosssectional, clinical high-risk and longitudinal, general population samples, further underscoring the generalizable importance of these findings. Overall, our findings highlight the need to routinely assess psychotic experiences in youth clinical practice as a risk marker for NSSI and suicidality.

Funding
This work was supported by the Erasmus MC University Medical L.R. Steenkamp et al. Center (Mrace 2016 107569) to L. Steenkamp, H. Tiemeier, S. Kushner, M. Hillegers, L. Blanken, and K Research and Development (ZonMw). The iBerry Study is funded by the Erasmus MC University Medical Center and the following institutes of mental health care (GGz): Parnassia Psychiatric Institute Antes, GGz Breburg, GGz Delfland, GGz Westelijk Noord-Brabant and Yulius. All funding organizations participate in the Epidemiological and Social Psychiatric Research Institute (ESPRi), a consortium of academic and non-academic research groups. The study was also financially supported by a grant from the Janivo Foundation. The funding sources had no involvement in the analysis or interpretation of data, the writing of the report, and the decision to submit the article for publication.

Declaration of competing interest
The authors have no relevant financial or non-financial interests to disclose.