Suicidal Ideation and Non-Suicidal Self-Injury Following Early Childhood Psychotic Experiences in Preadolescent Children at Familial High Risk of Schizophrenia or Bipolar Disorder—The Danish High Risk and Resilience Study, VIA 11

Abstract

Background and Hypothesis

Suicide is a leading cause of death in youth and is often preceded by suicidal ideation (SI) and non-suicidal self-injury (NSSI). Identifying early markers of risk for SI and NSSI could improve timely identification of at-risk individuals.

Study Design

Children (mean age 11.9, SD 0.2) at familial high risk of schizophrenia (N = 171), or bipolar disorder (N = 104), and controls (N = 174) were assessed for psychotic experiences (PE), SI, NSSI, and Axis I mental disorders in face-to-face interviews in early and middle childhood (age 7 and 11).

Study Results

Having 2 types of early childhood PE predicted middle childhood SI after accounting for previous SI, NSSI, and mental disorders (OR 2.8, 95% CI 1.1–6.9; P = .03). Two PE predicted NSSI (OR 3.0, 95% CI 1.2–7.7; P = .02) in excess of previous SI, NSSI, mental disorders, and familial risk. Persistent and incident PE predicted SI (OR 3.2, 95% CI, 1.1–8.8; P = .03; OR 3.8, 95% CI, 1.3–11.5; P = .02) in the fully adjusted model. Nineteen percent of children with persistent PE reported middle childhood SI vs 3.8% of those who never reported PE. In children with early childhood mental disorders, those who reported 2 PE had 4.4-fold increased odds of later SI (95% CI, 1.2–16.7; P = .03) after adjustments. PE were nondifferentially associated with outcomes across familial risk groups.

Conclusions

Early childhood PE index elevated risk for subsequent SI and NSSI beyond what can be attributed to presence of mental disorders. Mental health screenings and clinical assessments should include early childhood PE.

Introduction

Suicide is a major mental health concern worldwide and a leading cause of death in youth^1–3 making early prediction and prevention crucial. While accurate prediction of suicide is challenging,⁴ suicidal ideation (SI) and non-suicidal self-injury (NSSI) in children and adolescents are robustly associated with greater risk of concurrent and succeeding suicide attempts,^5–13 and these behaviors are viewed as a continuum.^14–16 Recent evidence shows that NSSI with onset before age 13 predicts greater severity of NSSI and higher likelihood of later suicide attempts compared with NSSI with later onset.¹⁷ Furthermore, SI in childhood is a marker of risk for mental disorders and poorer functioning later in life,^12,13 while youth who present to hospital with self-injury have excess risk of later being diagnosed with psychotic and bipolar disorders, with the highest risk for those presenting before age 18.¹⁸ Thus, identifying early markers of risk for these behaviors could enable targeted prevention with the aim of reducing risk of suicide and improving mental health outcomes.

SI and self-injury at different levels of severity are highly prevalent in individuals with psychotic disorders^19–24 and suicide is a common cause of premature death in this population.^25–28 Yet, evidence has accumulated that these behaviors are also widespread in individuals with psychotic experiences (PE), ie, hallucinations and delusions without a diagnosable psychotic disorder, which are particularly common in childhood and adolescence.²⁹ A number of studies have demonstrated that PE in preadolescence and onwards are associated with NSSI, SI, suicide attempts, and deaths.^30–41 In samples of adolescents with other psychopathology, those who also report PE more often have SI and suicide attempts than those without PE.^35,42,43

However, knowledge on early childhood markers of risk for suicidality is lacking, and although the association between PE and suicidality could be particularly strong in childhood,³⁹ to the best of our knowledge, no previous study has prospectively examined associations between early childhood PE and later SI and NSSI. Children and adolescents at familial high risk (FHR) of schizophrenia and bipolar disorder represent a high-risk group for psychotic and other mental disorders^44,45 which makes early identification of vulnerability within this population pertinent. Yet, while we have documented that PE are associated with mental disorders,^46,47 to the best of our knowledge, no study has examined PE in relation to SI and NSSI within this population.

PE are usually perceived as markers of risk for SI and NSSI but could also have a causal role.³³ The aims of the current study were descriptive, ie, to describe associations between PE and SI and NSSI rather than to explain the etiology of these behaviors.

The aims of the current study were (1) to examine whether early childhood PE predict middle childhood SI and NSSI in children at FHR of schizophrenia or bipolar disorder and population-based controls, (2) to examine associations between pathways of PE from early to middle childhood and SI and NSSI in middle childhood within this population, (3) in exploratory analyses, to examine whether PE in children with early childhood mental disorders are associated with subsequent SI and NSSI, and (4) to examine whether PE are differentially associated with SI and NSSI across children at FHR of schizophrenia, bipolar disorder, and population-based controls.

Methods

Study Design

The VIA 11 Study is the first follow-up of The Danish High Risk and Resilience Study, a Danish nationwide, prospective, longitudinal cohort study. The original cohort comprised 522 7-year-old children with at least one biological parent with a schizophrenia spectrum disorder, (FHR-SZ) (N = 202, ICD-10 codes: F20, F22, F25, or ICD-8 codes: 295, 297, 298.29, 298.39, 298.89, and 298.99), or bipolar disorder, (FHR-BP) (N = 120, ICD-10 codes: F30, F31, or ICD-8 codes: 296.19, 269.39), and population-based controls (hereafter controls) where neither parent was diagnosed with these disorders (N = 200). Controls were matched to children at FHR-SZ on age, sex, and municipality. Children at FHR-BP were an unmatched sample comparable to the other groups on age at inclusion and sex. The cohort was retrieved from Danish national registers.^48,49 Only children born in Denmark were considered eligible for inclusion. No other exclusion criteria were employed. The study was approved by the Danish Data Protection Agency. The Danish Committee on Health Research Ethics deemed ethical approval unnecessary due to the observational nature of the study. Written informed consent from the parent or guardian and assent from the children were obtained upon explanation of the procedures. Data collection took place from 2013 to 2016 and 2017 to 2020 in Copenhagen and Aarhus, Denmark. Child assessors were blinded to parental mental illness. The cohort and study design are described in detail elsewhere.^50,51

Measures

Psychotic Experiences

PE were assessed with the psychosis supplement from The Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL).⁵² Nine types of hallucinations and 13 types of delusions during early (until age 7) and middle (age 7–11) childhood were assessed through separate interviews with the child and the primary caregiver at the first assessment (age 7) and at follow-up (age 11). When a symptom was endorsed and given a K-SADS-PL score of subthreshold (ie, 2, suspected or likely) or threshold (ie, 3, definite) on either the child’s or caregiver’s response, additional questions were asked (frequency, duration, degree of conviction, distress, and impact on functioning). The symptom was then given a final score on a 7-point scale ranging from 0 = absent to 6 = severe and psychotic. PE scores on the 7-point scale were then recoded into 0–1 = “absent or possible PE” and 2–6 = “definite PE.”

For the analyses, only definite PE were considered. Having “any PE” was defined as any type of hallucination or delusion.

Five types of auditory hallucinations were collapsed into one, resulting in 5 types of hallucinations: Auditory, visual, tactile, olfactory, and loud thoughts. Delusions comprised 13 types: Grandiose, guilt/sin, control, somatic, nihilism, thought broadcasting, thought insertion, thought withdrawal, messages from tv/radio, persecution, mind reading, reference, and other delusions. A categorical variable with number of different types of PE (ie, number of hallucinations and/or delusions) was constructed as a proxy of severity, in keeping with previous studies^30,47: No PE, 1 type of PE, 2 types of PE, or 3 or more types of PE. If a child had visual hallucinations only, it was considered 1 type of PE, whereas if a child had both visual and auditory hallucinations or visual hallucinations and grandiose delusions, it was considered 2 types, etc.

A variable with pathways of PE containing 4 groups was created based on presence or absence of PE during early and middle childhood: (1) Children who never reported PE, (2) Children with PE during early childhood but not during middle childhood (remittent PE), (3) Children with PE during middle childhood but not during early childhood (incident PE), and (4) Children with PE during both early and middle childhood (persistent PE).

Methods and results regarding PE are described in detail elsewhere.^46,47

Suicidal Ideation and Non-Suicidal Self-Injury

SI and NSSI during early (until age 7) and middle (age 7–11) childhood were assessed through interviews with the screening section from the K-SADS-PL conducted with the child and the primary caregiver separately. SI was considered present when thoughts of completing suicide (item 4b) were endorsed (ie, given a summary score of subthreshold, infrequent thoughts, or threshold, recurrent thoughts). Very few children planned or carried out suicidal acts in middle childhood (N < 10, items 4c and 4d) and even fewer had intent to die (N < 5, as indicated by a summary score above threshold on item 4c), and this could not be examined as independent outcomes. Therefore, these behaviors were included as SI.

NSSI was considered present when intentional, self-inflicted damage to the body without intent of killing oneself, eg, cutting or burning oneself (item 4e), was endorsed (subthreshold, rarely or no serious injury, or threshold, often or significant injury).

In cases where symptoms warranting further attention, such as worrying SI, were reported during the interview, the child and caregivers were informed of relevant services and if necessary, referrals were made. If the information came from the child only, assent was obtained from the child before informing the caregivers.

Mental Disorders

DSM-IV and DSM-5 Axis I mental disorders during early (until age 7) and middle (age 7–11) childhood were assessed through interviews with the screening section and relevant supplements from the K-SADS-PL conducted with the child and the primary caregiver separately.

In the current cohort, diagnosed mental disorders included: Affective disorders, psychotic disorders, anxiety disorders, disruptive behavior disorders, attention deficit hyperactivity disorder (ADHD), autism spectrum disorders, posttraumatic stress disorder, stress and adjustment disorders, and tic disorders. Elimination disorders, transient and unspecified tics, and specific phobias were excluded.

For the analyses “any Axis I disorder” included any diagnosed mental disorder. Methods and results regarding mental disorders are described in detail elsewhere.^53,54

Intelligence Quotient

Intelligence Quotient (IQ) was measured with Reynolds Intellectual Screening Test.⁵⁵

Statistical Analyses

Differences in demographic and clinical background characteristics between children at FHR-SZ, FHR-BP, and controls were analyzed with chi-square tests and one-way analyses of variance. Frequencies and percentages were calculated using crosstabs. Dropout analyses were performed with chi-square.

In the main analyses, logistic regression was used to examine the relationship between early childhood PE and middle childhood SI and NSSI. For each of the 2 outcomes we first used the binary variable of any early childhood PE, then number of types of PE as predictor. Children with no early childhood PE were used as reference. We then used pathways of PE as predictor with children who never reported PE as reference. Analyses were successively adjusted for early childhood SI and NSSI, any Axis I mental disorder during the child’s life, and FHR-group.

In exploratory analyses, to examine associations between Axis I mental disorders with and without PE in early childhood and SI and NSSI in middle childhood, we used logistic regressions restricted to children with any Axis I mental disorder in early childhood. We used any PE and number of PE as predictors. Adjustments were carried out similarly to the analyses for the entire cohort, with the exception that, since only children with early childhood mental disorders were included, these analyses were adjusted for presence of any Axis I disorder during middle childhood.

To check for interaction effects between FHR-group and PE we added an interaction term to unadjusted models. This was subsequently removed in case of nonsignificant interaction.

Due to the risk of over-adjusting, we did not adjust for socioeconomic status or IQ as these factors are intrinsically associated with familial risk.^56,57

Alpha was set at <.05. Data were analyzed using SPSS version 25.

Results

Sample Characteristics

A total of 449 (47.0% girls) 11-year-old (mean 11.9, SD 0.2, range 10.9–12.7) children participated. Only children with data on both early childhood PE and middle childhood SI and NSSI were included in the study. The 3 groups of children did not differ regarding sex or age at inclusion. Children at FHR-SZ and FHR-BP had a significantly higher prevalence of early and middle childhood Axis I mental disorders, early childhood NSSI, and middle childhood SI than controls. Children at FHR-SZ had a higher prevalence of early childhood PE than controls and of middle childhood PE than children at FHR-BP and controls. Children at FHR-SZ had lower IQ than controls in middle childhood. Children at FHR-SZ had lower socioeconomic status than children at FHR-BP and controls (table 1). There were no sex differences in the examined outcomes (SI, χ²(1) = 0.228, P = .63; NSSI, χ²(1) = 0.674, P = .41).

Table 1.

Demographic and Clinical Background Characteristics of Study Population

	Familial Risk Group			P Value	P Value for Pairwise Comparisons
	FHR-SZ (N = 171)	FHR-BP (N = 104)	Controls (N = 174)		FHR-SZ vs Controls	FHR-BP vs Controls	FHR-SZ vs FHR-BP
Female, n (%)	83 (48.5%)	46 (44.2%)	82 (47.1%)	.79a	NA	NA	NA
Early childhood (until age 7)
Age at inclusion, years, mean (SD)	7.8 (0.2)	7.9 (0.2)	7.8 (0.2)	.06b	NA	NA	NA
Any psychotic experiences, n (%)	83 (48.5%)	46 (44.2%)	61 (35.1%)	.04a,*	.01*	.13	.49
Suicidal ideationd, n (%)	23 (13.5%)	8 (7.7%)	14 (8.1%)	.16a	NA	NA	NA
Non-suicidal self-injurye, n (%)	31 (18.6%)	18 (17.5%)	14 (8.1%)	.01a,*	.004*	.02*	.82
Any Axis I disorder (K-SADS-PL), n (%)	62 (36.3%)	36 (34.6%)	28 (16.1%)	<.001a,*	<.001*	<.001*	.78
Main disorder categories
Affective disorders, n (%)	3 (1.8%)	4 (3.8%)	2 (1.1%)	.29a	NA	NA	NA
Anxiety disorders, n (%)	20 (11.7%)	12 (11.5%)	9 (5.2%)	.07a	NA	NA	NA
ADHD, n (%)	31 (18.1%)	10 (9.6%)	13 (7.5%)	.007a,*	.003*	.53	.06
Disruptive behavior disorders, n (%)	9 (5.3%)	3 (2.9%)	2 (1.1%)	.09a	NA	NA	NA
Stress and adjustment disorders, n (%)	9 (5.3%)	10 (9.6%)	3 (1.7%)	.01a,*	.07	.003*	.17
IQf,g, mean (SD)	102.9 (11.3)	105.2 (8.8)	105.2 (9.8)	.07b	NA	NA	NA
Middle childhood (age 7–11)
Age at inclusion, years, mean (SD)	12.0 (0.3)	11.9 (0.2)	11.9 (0.2)	.55b	NA	NA	NA
Any psychotic experiencesh, n (%)	54 (31.8%)	21 (20.4%)	32 (18.4%)	.009a,*	.004*	.68	.04*
Suicidal ideation, n (%)	19 (11.1%)	12 (11.5%)	6 (3.4%)	.01a,*	.006*	.008*	.91
Non-suicidal self-injury, n (%)	20 (11.7%)	9 (8.7%)	8 (4.6%)	.06a	NA	NA	NA
Any Axis I disorder (K-SADS-PL), n (%)	83 (48.5%)	41 (39.4%)	39 (22.4%)	<.001a,*	<.001*	.002*	.14
Main disorder categories
Affective disorders, n (%)	14 (8.2%)	10 (9.6%)	2 (1.1%)	.003a,*	.002*	.001*	.68
Anxiety disorders, n (%)	17 (9.9%)	15 (14.4%)	11 (6.3%)	.08a	NA	NA	NA
ADHD, n (%)	35 (20.5%)	11 (10.6%)	10 (5.7%)	<.001a,*	<.001*	.14	.03*
Disruptive behavior disorders, n (%)	7 (4.1%)	3 (2.9%)	5 (2.9%)	.79a	NA	NA	NA
Stress and adjustment disorders, n (%)	16 (9.4%)	11 (10.6%)	6 (3.4%)	.04a,*	.03*	.02*	.74
IQf,i, mean (SD)	95.1 (10.5)	97.2 (9.6)	98.2 (10.4)	.02b,*	.006*	.45	.10
Educational level, primary caregiver j , k
Primary/lower secondary, n (%)	58 (33.9%)	19 (18.3%)	36 (20.8%)	.006c,*	.006*	.51	.003*
Upper secondary, vocational, short-cycle tertiary, n (%)	39 (22.8%)	22 (21.2%)	39 (22.5%)
Bachelor degree, equivalent, or higher, n (%)	74 (43.3%)	63 (60.6%)	98 (56.6%)

^*Indicates significance at alpha level <.05.

ADHD, attention deficit hyperactivity disorder; FHR-BP, children at familial high risk of bipolar disorder; FHR-SZ, children at familial high risk of schizophrenia spectrum disorders; IQ, intelligence quotient; K-SADS-PL, Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version.

NA, Not Applicable.

^aChi square test.

^bOne-way ANOVA test with post hoc Least Significant Difference.

^cLinear-by-Linear Association p value is reported when an ordinal variable has more than 2 categories.

^dIncludes 170 children at FHR-SZ, 104 children at FHR-BP, and 173 controls.

^eIncludes 167 children at FHR-SZ, 103 children at FHR-BP, and 173 controls.

^fMeasured with Reynolds Intellectual Screening Test (RIST).

^gIncludes 171 children at FHR-SZ, 104 children at FHR-BP, and 172 controls.

^hIncludes 170 children at FHR-SZ, 103 children at FHR-BP, and 174 controls.

ⁱIncludes 168 children at FHR-SZ, 103 children at FHR-BP, and 173 controls.

^jThe primary caregiver is defined as the caregiver currently spending the most time with the child.

^kIncludes 171 primary caregivers of children at FHR-SZ, 104 caregivers of children at FHR-BP, and 173 caregivers of controls.

There were no significant differences between children included in the current study and those not included regarding prevalence of any PE (χ²(1) = 1.457, P = .23), any Axis I disorder (χ²(1) = 1.479, P = .22), or SI (χ²(1) = 0.363, P = .55) during early childhood, sex (χ²(1) = 0.518, P = .47), or FHR-group (χ²(2) = 0.515, P = .77). Children lost to follow-up more often engaged in early childhood NSSI (23.8%) than those who remained (14.2%, χ²(1) = 3.889, P = .05). Retention was 86%.

Early Childhood PE and Middle Childhood SI and NSSI

Having 2 types of early childhood PE was associated with increased odds of middle childhood SI. Associations were attenuated, yet remained significant when adjusting for mental disorders, but were no longer statistically significant when adjusting for familial risk. Of children with 2 PE, 20.8% reported middle childhood SI. Of children with 3 or more PE, 16.2% reported SI, which was not statistically significantly higher than in those without PE (6.2%) after adjustments (table 2).

Table 2.

Associations between Early Childhood Psychotic Experiences and Middle Childhood Suicidal Ideation and Non-Suicidal Self-Injury

	Early Childhood Psychotic Experiences
	Any Psychotic Experiences (N = 190)d			One type of Psychotic Experiences (N = 105)e			Two types of Psychotic Experiences (N = 48)			Three or more types of Psychotic Experiences (N = 37)
	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value
Suicidal ideation, middle childhood
Model 1a	1.6	0.8–3.2	.18	0.7	0.2–1.9	.48	3.3	1.4–8.1	.008*	2.5	0.9–7.1	.08
Model 2a,b	1.3	0.6–2.6	.49	0.6	0.2–1.6	.30	2.8	1.1–6.9	.03*	1.7	0.6–4.8	.34
Model 3a,b,c	1.2	0.6–2.5	.61	0.6	0.2–1.7	.30	2.5	1.0–6.2	.06	1.4	0.5–4.3	.50
Non-suicidal self-injury, middle childhood
Model 1a	1.3	0.6–2.7	.48	0.2	0.1–1.1	.06	3.6	1.5–8.7	.004*	1.9	0.7–5.7	.23
Model 2a,b	1.0	0.5–2.2	.90	0.2	0.0–0.9	.04*	3.0	1.2–7.5	.02*	1.4	0.5–4.1	.58
Model 3a,b,c	1.0	0.5–2.2	.96	0.2	0.0–0.9	.04*	3.0	1.2–7.7	.02*	1.3	0.4–4.0	.63

Children with no psychotic experiences in early childhood (N = 259) were used as reference.

^*Indicates significance at alpha level <.05.

^aAdjusted for suicidal ideation and non-suicidal self-injury in early childhood (age 7).

^bAdjusted for any Axis I disorder during the child’s life at age 11.

^cAdjusted for familial risk.

^dSignificant Covariates with any Psychotic Experiences:

Suicidal Ideation:

In models 2 and 3, a significant association between Axis I disorders during the child’s life and suicidal ideation in middle childhood was found. Having an Axis I disorder was associated with higher odds of suicidal ideation in middle childhood.

Non-Suicidal Self-Injury:

In model 1, a significant association between non-suicidal self-injury (NSSI) in early childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood was associated with higher odds of NSSI in middle childhood.

In models 2 and 3, a significant associations between NSSI in early childhood and Axis I disorders during the child’s life and NSSI in middle childhood was found. Reporting NSSI in early childhood and having an Axis I disorder was associated with higher odds of NSSI in middle childhood.

^eSignificant Covariates with Number of Types of Psychotic Experiences:

Suicidal Ideation:

In models 2 and 3, a significant association between Axis I disorders during the child’s life and suicidal ideation in middle childhood was found. Having an Axis I disorder was associated with higher odds of suicidal ideation in middle childhood.

Non-Suicidal Self-Injury:

In model 1, a significant associations between NSSI in early childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood was associated with higher odds of NSSI in middle childhood.

In models 2 and 3, a significant association between NSSI in early childhood and Axis I disorders during the child’s life and NSSI in middle childhood was found. Reporting NSSI in early childhood and having an Axis I disorder was associated with higher odds of NSSI in middle childhood.

Children with 2 types of early childhood PE had increased odds of middle childhood NSSI compared with children with no PE after adjusting for previous SI and NSSI, mental disorders, and familial risk. Of children with 2 PE, 25% reported NSSI compared with 6.6% of children with no early childhood PE. Children with one type of PE had decreased risk of NSSI (table 2). Although a higher percentage of children with 3 or more early childhood PE (16.2%) reported middle childhood NSSI compared with those without PE, this was not statistically significant after adjustments (table 2).

Pathways of PE and Middle Childhood SI and NSSI

Persistent and incident PE were associated with a significantly higher risk of middle childhood SI than never having PE. Associations were attenuated, but remained significant when adjusting for mental disorders and familial risk (table 3). Nineteen and 16.3% of children with persistent and incident PE, respectively, reported middle childhood SI, compared with 3.8% of children who never reported PE.

Table 3.

Associations Between Pathways of Psychotic Experiences and Middle Childhood Suicidal Ideation and Non-Suicidal Self-Injury

	Pathways of Psychotic Experiences
	Remittent Psychotic Experiences (N = 131)d			Incident Psyhotic Experiences (N = 49)			Persistent Psychotic Experiences (N = 58)
	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value
Suicidal ideation, middle childhood
Model 1a	1.8	0.7–4.8	.22	4.9	1.7–14.0	.003*	4.9	1.8–13.1	.002*
Model 2a,b	1.5	0.5–3.9	.46	3.7	1.3–11.1	.02*	3.4	1.2–9.2	.02*
Model 3a,b,c	1.4	0.5–3.7	.54	3.8	1.3–11.5	.02*	3.2	1.1–8.8	.03*
Non-suicidal self-injury, middle childhood
Model 1a	0.6	0.2–1.7	.34	0.9	0.2–3.5	.87	3.0	1.2–7.2	.02*
Model 2a,b	0.5	0.2–1.3	.15	0.6	0.1–2.4	.48	2.0	0.8–5.0	.14
Model 3a,b,c	0.5	0.2–1.3	.14	0.6	0.1–2.4	.45	1.9	0.8–4.8	.18

Two children who did not provide data on psychotic experiences in middle childhood (FHR-SZ = 1, FHR-BP = 1) were not included in these analyses, therefore N = 447.

Children who never reported psychotic experiences (N = 209) were used as reference.

^*Indicates significance at alpha level <.05.

FHR-SZ, Children at familial high risk of schizophrenia spectrum disorders; FHR-BP, Children at familial high risk of bipolar disorder.

^aAdjusted for suicidal ideation and non-suicidal self-injury in early childhood (age 7).

^bAdjusted for any Axis I disorder during the child’s life at age 11.

^cAdjusted for familial risk.

^dSignificant Covariates with Pathways of Psychotic Experiences:

Suicidal Ideation:

In models 2 and 3, a significant association between Axis I disorders during the child’s life and suicidal ideation in middle childhood was found. Having an Axis I disorder was associated with higher odds of suicidal ideation in middle childhood.

Non-Suicidal Self-Injury:

In model 1, a significant association between non-suicidal self-injury (NSSI) in early childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood was associated with higher odds of NSSI in middle childhood.

In models 2 and 3, a significant association between NSSI in early childhood and Axis I disorders during the child’s life and NSSI in middle childhood was found. Reporting NSSI in early childhood and having an Axis I disorder was associated with higher odds of NSSI in middle childhood.

Persistent PE were associated with higher odds of middle childhood NSSI when adjusting for previous SI and NSSI (table 3). Of children with persistent PE, 22.4% reported middle childhood NSSI vs 6.2% of children who never reported PE. However, these associations were rendered nonsignificant when additionally adjusting for mental disorders and familial risk.

Middle Childhood SI and NSSI in Children with Early Childhood Axis I Disorders with and without PE

In exploratory analyses, children with early childhood mental disorders who also reported 2 types of PE had significantly increased risk of middle childhood SI compared with those without PE. Forty percent of children with 2 PE reported middle childhood SI vs 10% of children without PE. This association was attenuated but remained significant when adjusting for middle childhood mental disorders and familial risk (table 4). Of children with 3 or more types of PE, 26.3% reported SI. This was not significantly higher than in those without PE.

Table 4.

Associations Between Psychotic Experiences and Suicidal Ideation and Non-Suicidal Self-Injury in Children with Early Childhood Mental Disorders

	Early Childhood Psychotic Experiences
	Any Psychotic Experiences (N = 66)d			One type of Psychotic Experiences (N = 27)e			Two types of Psychotic Experiences (N = 20)			Three or more types of Psychotic Experiences (N = 19)
	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value	Odds ratio	95% CI	P Value
Suicidal ideation, middle childhood
Model 1a	2.5	0.9–7.1	.08	0.7	0.1–3.7	.67	5.8	1.7–20.3	.006*	3.0	0.8–11.5	.11
Model 2a,b	2.3	0.8–6.6	.13	0.7	0.1–3.7	.63	4.6	1.3–16.8	.02*	2.6	0.6–10.7	.18
Model 3a,b,c	2.1	0.7–6.2	.17	0.7	0.1–3.8	.65	4.4	1.2–16.7	.03*	2.5	0.6–10.3	.22
Non-suicidal self-injury, middle childhood
Model 1a	1.8	0.7–4.8	.22	0.3	0.0–2.2	.22	3.8	1.1–12.5	.03*	2.9	0.8–10.2	.10
Model 2a,b	1.6	0.6–4.5	.33	0.3	0.0–2.2	.21	2.9	0.8–10.2	.09	2.7	0.7–9.8	.14
Model 3a,b,c	1.6	0.6–4.6	.34	0.2	0.0–2.1	.19	3.4	0.9–12.6	.06	3.2	0.8–12.7	.11

Only children with mental disorders in early childhood (N = 126) were included in these analyses.

Children with no psychotic experiences in early childhood (N = 60) were used as reference.

^*Indicates significance at alpha level <.05.

^aAdjusted for suicidal ideation and non-suicidal self-injury in early childhood (age 7).

^bAdjusted for any Axis I disorder during middle childhood (age 7–11).

^cAdjusted for familial risk.

^dSignificant Covariates with any Psychotic Experiences:

Suicidal Ideation:

In models 2 and 3, associations between Axis I disorders during middle childhood and suicidal ideation could not be estimated.

Non-Suicidal Self-Injury:

In model 1, a significant association between non-suicidal self-injury (NSSI) in early childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood was associated with higher odds of NSSI in middle childhood.

In models 2 and 3, a significant association between NSSI in early childhood and Axis I disorders during middle childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood and having an Axis I disorder was associated with higher odds of NSSI in middle childhood.

^eSignificant Covariates with Number of Types of Psychotic Experiences:

Suicidal Ideation:

In models 2 and 3, associations between Axis I disorders during middle childhood and suicidal ideation could not be estimated.

Non-Suicidal Self-Injury:

In model 1, a significant association between NSSI in early childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood was associated with higher odds of NSSI in middle childhood.

In models 2 and 3, a significant association between NSSI in early childhood and Axis I disorders during middle childhood and NSSI in middle childhood was found. Reporting NSSI in early childhood and having an Axis I disorder was associated with higher odds of NSSI in middle childhood.

Having 2 types of PE was associated with increased odds of middle childhood NSSI when adjusting for previous SI and NSSI. This, however, was no longer statistically significant when adjusting for middle childhood mental disorders and familial risk (table 4). Of children with 2 PE, 40.0% reported NSSI vs 13.3% of children with no PE. Although not significantly higher than in those without PE, 31.6% of children with 3 or more PE reported NSSI.

Interaction Between PE and Familial Risk

There was no interaction between PE and FHR-group, ie, early childhood PE were nondifferentially associated with outcomes across FHR-groups (P-values for test of interaction >.15) as were pathways of PE (P-values >.51). No interaction was found for children with early childhood mental disorders either (P-values >.25).

Discussion

In this prospective longitudinal cohort study, we found that children who reported 2 types of early childhood PE had 2.8-fold increased risk of middle childhood SI after accounting for mental disorders and previous SI and NSSI. This association was no longer statistically significant when adjusting for familial risk. Having 2 types of early childhood PE conferred 3-fold increased risk of subsequent NSSI compared with having no PE. Risk was increased in excess of that explained by previous SI and NSSI, mental disorders, and familial risk. In the fully adjusted model, children with persistent and incident PE had more than 3-fold higher risk of middle childhood SI compared with children who never reported PE. In children with early childhood mental disorders, those who reported 2 types of PE had more than 4-fold increased risk of subsequent SI compared with children without PE, while increase in risk of NSSI was not statistically significant, after accounting for previous SI and NSSI, middle childhood mental disorders, and familial risk. PE were nondifferentially associated with outcomes across FHR-groups.

To the best of our knowledge, this is the first study to prospectively examine SI and NSSI following early childhood PE and to examine associations between PE and SI and NSSI in children at FHR-SZ or FHR-BP. Our findings corroborate existing evidence that the increased risk of SI and self-injury found in psychotic disorders extends to the far more commonly occurring subclinical phenotype,^30–41,58 while showing that this encompasses early childhood PE in children at FHR. In the current cohort, associations between PE and SI and NSSI were, in most cases, attenuated when accounting for mental disorders, which aligns with findings from previous studies adjusting for presence of other psychopathology.^{30,33,37,39,40} This was expected given the evidence that mental disorders often cooccur with behaviors across the suicidality continuum^{7,20,27,28,59} and that higher dimensional psychopathology is a risk factor for suicidal behaviors in children.⁶⁰ These findings suggest that common risk factors for PE and these behaviors, such as other psychopathology, could partially underlie their associations as previously proposed.³³ In the current cohort, associations between PE and SI were further attenuated when adjusting for familial risk. As both PE and SI were more common in children at FHR-SZ, this could suggest familial risk of schizophrenia as another shared risk factor. Nonetheless, although other factors may partially explain their cooccurrence, our findings and those of others point to PE as important markers of risk for these behaviors.

Studies in older populations have documented that higher severity of PE confers increased risk of suicidality. Some studies have reported a dose–response relationship between a higher number of types of PE and subsequent risk of NSSI, SI, and suicidal behaviors^30,39,58 and some also show that having only one PE type does not substantially increase risk of NSSI or suicidal behavior.^30,58 We found that 2 types of PE increased risk of these behaviors while having one PE did not; in fact, children with one PE had decreased risk of NSSI, supporting the notion that PE are in many cases normal developmental phenomena. One study found that only PE classified as strong at age 11 were associated with elevated risk of attempted or completed suicide by age 38.³⁶ Although, in the current cohort, roughly twice as many children with 3 or more PE reported SI and NSSI as children without PE, while of potential clinical relevance, these differences were not statistically significant, likely due to the broad confidence intervals resulting from relatively small groups sizes. Interestingly, percentages of children reporting SI and NSSI were lower among those with 3 or more PE than among those with 2 PE, suggesting that this does not add further to the risk, contrasting with the abovementioned findings of a dose–response relationship. The small group sizes could be one potential explanation for this somewhat surprising finding. It might also be partially attributable to the young age of the current sample, as PE become more clinically relevant with advancing age,⁶¹ where a dose–response effect may emerge. We found that only persistent and incident but not remittent PE increased risk of SI, aligning with findings from older samples. One study showed that persistent and incident, but not remittent, PE from ages 11 to 16 increased risk of NSSI, SI, and suicidal behaviors at age 16³⁰ whereas another study in youth, ages 12–17, documented that only persistent PE were associated with NSSI, while incident and persistent PE were associated with suicide attempts at 1-year follow-up.³¹ These findings highlight the continuum of PE and the relevance of considering their severity, including their development over time, to assess risk of detrimental outcomes.

Our exploratory findings point to PE cooccurring with early childhood mental disorders as markers of increased risk of suicidality in keeping with the previously mentioned studies of adolescents with other psychopathology documenting that PE are associated with suicidality across the severity continuum.^35,42,43 These associations may reflect more severe underlying psychopathology in those reporting PE.^43,61–63 However, PE predict suicidal behaviors after accounting for symptom burden and multimorbidity^35,43 making it likely that these factors only partially explain this relationship. Another potential explanation includes presence of shared risk factors for PE and suicidality such as adverse life events in childhood^7,64–66 as previously suggested.^35,42,43

Finally, the nondifferential associations between PE and suicidality across FHR-groups in the current cohort suggest similar clinical significance of PE in children at varying familial risk, ie, in children at FHR-SZ, FHR-BP, and controls, aligning with our previous findings.⁴⁷

The current findings, along with our previous reports on PE and mental disorders,^46,47 indicate that children with early childhood PE represent an at-risk group pertinent for monitoring of mental health outcomes. Additionally, our findings indicate that this is the case irrespective of degree of familial risk. Furthermore, the current study along with those in older samples, suggests that inquiring about PE should form part of the assessment of future suicide risk. Presence of more than one type of PE and persistent PE should elicit particular attention and the increased risk conferred by persistent PE suggests a need for repeated, longitudinal follow-up when inquiring about PE. In addition, these findings underline the importance of routinely assessing PE in children and adolescents with other psychopathology, not only when suspecting a psychotic disorder. Of note, although a high-risk group for suicide, only a minority of children and adolescents with SI and NSSI transition to suicide attempts^5,7 and evidence on longitudinal predictors to differentiate those who do is scarce⁵ with inconsistent findings as to whether PE predict excess risk of suicide attempts in those with SI.^39,42 Rates of SI, NSSI, and suicide attempts are likely to increase from childhood to adolescence^2,7,67 and future follow-ups in the current cohort should examine predictors of SI and NSSI in adolescence as well as predictors of transition to more severe suicidal behaviors.

Strengths and Limitations

Strengths of the study include its prospective nature, the high retention rate (86%), and the use of the same gold-standard interview including information from child and caregiver at both assessments. Furthermore, the design allowed for examining outcomes of early childhood PE in same-aged children while adjusting for mental disorders and previous SI and NSSI. Limitations include the relatively small group sizes which may not have been fully conducive to detecting interaction effects or differences for those with 3 or more PE. Therefore, no firm conclusions should be drawn regarding these findings. The group sizes precluded stratification by mental disorder categories and severity of SI and NSSI, which may have yielded further insights. Ideally, future studies should replicate these findings in larger samples. Although our findings point to nondifferential clinical significance of PE irrespective of familial risk and although the rates of mental disorders, SI, and NSSI among controls were comparable to those in preadolescent and early adolescent general population samples,^7,68,69 due to the relatively rarer occurrence of these outcomes among controls compared with those at FHR, large scale general population studies are needed to expand knowledge on the value of early childhood PE in predicting mental health outcomes. The K-SADS-PL does not allow for distinction between different types of NSSI which may have provided further insights. Additionally, we are unable to determine subtype of parental bipolar disorder which precludes identification of potentially homogeneous subgroups of children at FHR-BP. Finally, children with early childhood NSSI were more often lost to follow-up; thus, expecting the same direction of results, associations between PE and NSSI may have been stronger in case of a more even dropout.

Conclusions

This study adds to the limited knowledge on the pathological significance of early childhood PE by showing that 2 types of PE and persistent PE index risk of later SI and NSSI. The findings strengthen the notion that PE are important markers of risk in children at FHR of schizophrenia and bipolar disorder. Additionally, the findings highlight the relevance of including PE in mental health screenings and clinical assessments suggesting that this should extend to early childhood, providing potential for identifying at-risk groups for future suicidality and other detrimental outcomes with ensuing opportunities for early prevention. Future studies should examine transition to more severe suicidal behaviors following early childhood PE.

Funding

This work was supported by the Capital Region of Denmark, the Mental Health Services of the Capital Region of Denmark, Aarhus University, the Central Denmark Region, the TRYG Foundation, the Lundbeck Foundation Initiative for Integrative Psychiatric Research – iPSYCH (grant number R248-2017-2003), The Innovation Fund (grant number 6152-00002B), and the Beatrice Surovell Haskell Fund for Child Mental Health Research of Copenhagen (grant number 11531).