Summary
Background
Childhood suicidal ideation and behaviours are poorly understood. We examined correlates of suicidality in a US population-based sample of children participating in the Adolescent Brain and Cognitive Development (ABCD) study. The ABCD study aims to examine trajectories of mental health from childhood to adulthood and collects information on multiple domains, including mental and physical wellbeing, brain imaging, behavioural and cognitive characteristics, and social and family environment. We sought to identify and rank risk and protective factors for childhood suicidal thoughts and behaviours across these multiple domains and evaluate their association with self-agreement and caregiver agreement in reporting suicidality.
Methods
The ABCD sample comprises a cohort of 11 875 children aged 9–10 years. The sociodemographic factors on which the sample was recruited were age, sex, race, socioeconomic status, and urbanicity. Participants were enrolled at 22 sites, the catchment area of which encompassed over 20% of the entire US population in this age group. Multistage sampling was used to ensure both local randomisation and representativeness of sociodemographic variation of the ABCD sample. The data used in this study were accessed from the ABCD Study Curated Annual Release 2.0. Suicidal thoughts and behaviours (suicidality) in each child were evaluated through independent child and caregiver reports based on the computerized Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5 (KSADS-5). We used bootstrapped logistic regression to quantify the association between suicidal ideation and behaviours, with measures of mental and physical wellbeing, behaviour, cognition, and social and family environment in participants from the ABCD study.
Findings
Our study sample comprised 7994 unrelated children (mean age 9∙9 years [SD 0∙5]; 4234 [53%] male participants) with complete data on child-reported and caregiver-reported suicidal ideas and behaviours. Overall, 673 (8·4%) children reported any past or current suicidal ideation, 75 (0·9%) had any past or current suicidal plans, and 107 (1·3%) had any past or current suicidal attempts. According to caregivers, 650 (8·1%) of the children reported any past or current suicidal ideation, 46 (0·6%) reported any past or current suicidal plans, and 39 (0·5%) reported past or current suicidal attempts. However, inter-informant agreement was low (Cohen’s κ range 0·0–0∙2). Regardless of informant, child psychopathology (odds ratio [OR] 1∙7–4∙8, 95% CI 1∙5–7∙4) and child-reported family conflict (OR 1∙4–1∙8, 95% CI 1∙1–2∙5) were the most robust risk factors for suicidality. The risk of child-reported suicidality increased with higher weekend screen use time (OR 1∙3, 95% CI 1∙2–1∙7) and reduced with greater parental supervision and positive school involvement (for both OR 0∙8, 95% CI 0∙7–0∙9). Additionally, caregiver-reported suicidality was positively associated with caregiver educational level (OR 1∙3, 95% CI 1∙1–1∙5) and male sex in children (1∙5, 1∙1–2·0), and inversely associated with the number of household cohabitants (0∙8, 0∙7–1·0).
Interpretation
We identified risk and protective factors that show robust and generalisable associations with childhood suicidality. These factors provide actionable targets for optimising prevention and intervention strategies, support the need to identify and treat psychopathology in school-age children, and underscore the importance of school and family interventions for childhood suicidality.
Funding
National Institutes of Health.
Introduction
Suicide is a major worldwide public health concern1 but is also a particular public health crisis in the USA, where suicide rates have been steadily increasing, even among children and adolescents.2 In the USA, suicide is currently the second leading cause of death in those aged 10–14 years,2 and the number of emergency clinic visits and hospitalisations for suicidal ideation and suicide attempts in this age group has doubled over the last decade.3,4 Suicidal ideas and behaviours in childhood predict adult psychiatric morbidity and mortality and thus serve as a marker of lifelong vulnerability to poor mental health.5,6 Prompt identification of childhood suicidal ideas and behaviours and interventions to minimise childhood suicidal risk are crucial steps in shifting developmental trajectories to better functional outcomes.
Suicidality is currently understood as a spectrum that spans suicidal ideas, plans, and attempts, some of which may be fatal.1 Suicidal attempts and non-suicidal self-injury sometimes involve similar acts, but differ specifically in that non-suicidal self-injury does not involve any desire to die.7,8 Nevertheless, both behaviours are associated with personal and family distress and raise professional concerns. Although non-suicidal self-injury is associated with an increased risk of subsequent suicide attempts,7 the two phenomena are not equivalent but equally complex and understudied in children. Here, we focus on identifying personal, family, and environmental factors associated with suicidal thoughts and behaviours, given their contribution to childhood mortality.
Previous research in adults has identified multiple risk factors for suicidal ideas and behaviours that generally reflect higher levels of psychosocial adversity, psychopathology, and medical morbidity.9 Similar risk factors apply to adolescents in whom epidemiological studies have implicated higher levels of psychopathology (distress, disruptive, or impulsive behaviour), psychiatric morbidity (mainly anxiety, depression, and substance disorders), and social stressors (negative life events and interpersonal conflict).10–16 However, there is a paucity of large-scale epidemiological studies in younger children. This knowledge gap is important because age could substantially influence how suicidal ideas and behaviours are reported and the nature of risk and protective factors involved. For example, age seems to influence the association between cognition and suicidal ideation and suicide attempts.17,18 Moreover, the agreement between self-reported and parent-reported suicidality is generally low in younger children, but increases with age during adolescence.19–21 Over the past few years, new putative risk factors have emerged relating to the use of digital media by young children. In the USA, an estimated almost 98% of children aged 8 years or younger have access to internet-connected devices and have over 2 h of screen time per day.22 Evidence from meta-analyses and population-based cohorts23–25 suggests a positive correlation between screen time and youth psychopathology that could theoretically extend to suicidality. Finally, it is essential that protective factors that might mitigate suicide risk in school-age children are identified, as these are as important as risk factors in formulating early intervention strategies.
The Adolescent Brain and Cognitive Development (ABCD) study provides a unique opportunity to study childhood suicidal thoughts and behaviours in the largest currently available nationally-representative population-based sample of 9–10 year-olds living in the USA. The ABCD study aims to examine trajectories of mental health from childhood to adulthood and collects information on multiple domains, including mental and physical wellbeing, brain imaging, behavioural and cognitive characteristics, and social and family environment. Harnessing the power of the ABCD cohort, we sought to identify and rank risk and protective factors for childhood suicidal thoughts and behaviours across these multiple domains and evaluate their association with self-agreement and caregiver agreement in reporting suicidality. We were specifically interested in identifying factors that are reproducible and robust to demographic and site variation to maximise the relevance of the study findings to national or international public health initiatives that aim to improve mental health in children. We hypothesised that suicidal thoughts and behaviours in children would generally be associated with adverse clinical, behavioural, and social characteristics, but that factors relevant to emotional wellbeing in childhood, such as the quality of family life, might be of greater predictive value in this age group.
Methods
The ABCD study
The ABCD sample comprises a cohort of 11 875 children aged 9–10 years, recruited to match national socioeconomic variation as captured by the American Community Survey, which is done annually by the US Census Bureau, and annual 3rd and 4th grade school enrolment data, which is maintained by the National Center for Education Statistics. The sociodemographic factors on which the sample was recruited were age, sex, race, socioeconomic status, and urbanicity. Participants were enrolled at 22 sites, the catchment area of which encompassed over 20% of the entire US population in this age group. Multistage sampling was used to ensure both local randomisation and representativeness of sociodemographic variation of the ABCD sample (appendix p 2).26 The data used in this study were accessed from the ABCD Study Curated Annual Release 2.0 and are available on request from the National Institute of Mental Health data archive. The University of California at San Diego (San Diego, CA, USA) Institutional Review Board is responsible for the ethical oversight of the ABCD study.
Procedures
Suicidal thoughts and behaviours in each child were evaluated via independent child and caregiver reports based on the computerised Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5 (KSADS-5). The KSADS-5 assesses the following features: passive suicidal ideation, active but non-specific suicidal ideation, suicidal ideation with a specific method, active suicidal ideation with an intent, active suicidal ideation with a plan, preparatory actions toward imminent suicidal behaviour, interrupted suicidal attempts, aborted suicidal attempts, and suicidal attempts (appendix p 3–4). Items relating to self-injurious behaviour without suicidal intent were not included.
Based on previous literature regarding risk factors for suicidal ideation and behaviours,11–19,25,27 we selected variables pertaining to neighbourhood environment, parental characteristics, and family function and to the prenatal and obstetric history, physical and mental wellbeing, cognitive ability, and behavioural traits of the children. Details of the procedures and instruments used in the ABCD study have been previously published.28–30 Briefly, these comprise study-specific questionnaires and widely used instruments including, in addition to KSADS-5, the Child Behavior Checklist (CBCL), PhenX, and the Children’s Report of Parental Behavioral Inventory. A detailed definition of each variable and the instrument used for its assessment is provided in the appendix (pp 5–8). The total problem score of the CBCL was highly correlated with the externalising (rho = 0∙84) and internalising (rho = 0∙82) subscales, as well as with the other CBCL-derived subscales (anxious/depressed, withdrawn/depressed, somatic complaints, aggressive behaviour, rule-breaking behaviour, attention, social, and thought problems subscales; rho range 0·56–0·83).
Statistical analysis
Complete data were available for all suicidality features. Missing data for personal, family, and social characteristics (range 0∙1%–9·0%) were imputed using expectation maximisation implemented using SPSS version 23. The expectation maximisation algorithm maximises the log-likelihood of the available data, with missing data marginalised so that the log-likelihood for the full data (available plus missing) is greater than that for the available data alone.
We assessed agreement between child and caregiver reports for each feature of suicidality using Cohen’s κ (implement in R-cran with the kappa2 function). Following convention, κ values below 0∙20 indicated poor child–caregiver agreement.
We tested group differences in personal, family, and social characteristics between children who were never suicidal and those who endorsed suicidality with Welch’s t tests, implemented using the t test function in R, which allows for unequal sample sizes. Results are reported as significant at p<0∙001 following Bonferroni correction.
We used logistic regression analyses (implemented in R-cran, version 3.5.3 using the glm function) to quantify statistically significant associations between participants’ personal, family, and social characteristics and suicidal thoughts and behaviours. We standardised continuous variables to create z-scores with a mean of 0 and a SD of 1. We excluded collinear variables using a conservative variation inflation factor cutoff greater than 2∙5. Although the CBCL total problems score was used in the main regression analyses, we also tested each subscale of the CBCL separately. We used a bootstrapping approach to establish generalisability across sites, address the unequal sample size of the groups, and correct for multiple comparisons. Each regression analysis was thus repeated 10 000 times and in each reiteration the never-suicidal group was resampled to create 10 000 random subgroups, each matching the size of groups reporting suicidality. For each reiteration, we extracted the Nagelkerke R2 and p value of the overall model and the β, Wald statistic, p value, and odds ratio (OR) and corresponding 95% CI for each variable. Adjustments for multiple comparisons were implemented first at the level of the overall model and then at the level of the individual predictor variables to ensure that the results were robust and generalisable across sites. The overall model of each reiteration was considered significant at p<0·05 following Bonferroni correction for multiple comparisons. We reported individual predictor variables if they made statistically significant contributions to more than 60% of the significant models (appendix p 8). The statistical significance of a variable was defined by the number of iterations that reported p<0∙05 for this variable, and thus was explicitly corrected for multiple comparison against the 10 000 random subsamples. We recomputed all analyses in the complete sample and separately for boys and girls.
Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Results
The study sample comprised 7994 unrelated children (mean age 9∙9 years [SD 0∙5]; 4234 [53%] male participants) with complete data on child-reported and caregiver-reported suicidal ideas and behaviours. A flowchart of participant selection is shown in the appendix (p 2). The study sample was similar to the full ABCD sample (appendix p 9).
Most caregivers were mothers. Agreement between child and caregiver reports for each feature of suicidality was generally low (mean Cohen’s κ 0∙05 [SD 0∙06]; range 0·0–0∙2; table 1; and appendix p 10).
Table 1:
Suicidal thoughts, plans, and behaviours in the child-reported suicidality group and caregiver- reported suicidality group
| Child-reported suicidality (n=684) | Caregiver-reported suicidality (n=654) | Cohen’s k | |
|---|---|---|---|
| Past | |||
| Passive suicidal ideation | 469 (5·8%) | 550 (6·9%) | 0·05 |
| Active but non-specific suicidal ideation | 294 (3·7%) | 270 (3·4%) | 0·2 |
| Suicidal ideation with a specific method | 75 (0·9%) | 70 (0·9%) | 0·06 |
| Active suicidal ideation with an intent | 47 (0·6%) | 30 (0·4%) | 0·05 |
| Active suicidal ideation with a plan | 29 (0·4%) | 10 (0·1%) | 0 |
| Preparatory actions toward imminent suicidal behaviour | 34 (0·4%) | 38 (0·5%) | 0 |
| Interrupted suicidal attempts | 10 (0·1%) | 5 (<0·1%) | 0 |
| Aborted suicidal attempts | 24 (0·3%) | 8 (0·1%) | 0 |
| Suicidal attempts | 59 (0·7%) | 23 (0·3%) | 0·2 |
| Current | |||
| Passive suicidal ideation | 98 (1·2%) | 48 (0·6%) | 0·05 |
| Active but non-specific suicidal ideation | 126 (1·6%) | 93 (1·2%) | 0·08 |
| Suicidal ideation with a specific method | 47 (0·6%) | 14 (0·2%) | 0·03 |
| Active suicidal ideation with an intent | 25 (0·3%) | 7 (0·1%) | 0·06 |
| Active suicidal ideation with a plan | 20 (0·3%) | 3 (<0·1%) | 0 |
| Preparatory actions toward imminent suicidal behaviour | 17 (0·2%) | 6 (0·1%) | 0 |
| Interrupted suicidal attempts | 4 (0·1%) | 0 | 0 |
| Aborted suicidal attempts | 16 (0·2%) | 1 (<0·1%) | 0 |
| Suicidal attempts | 15 (0·2%) | 5 (0·1%) | 0·1 |
| Any of the above past or current suicidality features | 684 (8·6%) | 654 (8·2%) | 0·2 |
Data are n (%) unless otherwise indicated. Percentages use the total analysis group as the denominator (n=7994). More than one suicidality feature was present in some children.
Overall, 673 (8·4%) children reported any past or current suicidal ideation, 75 (0·9%) had any past or current suicidal plans, and 107 (1·3%) had any past or current child-reported suicidal attempts (table 1). According to caregivers, 650 (8·1%) children reported any past or current suicidal ideation, 46 (0·6%) reported any past or current suicidal plans, and 39 (0·5%) reported past or current suicidal attempts (table 1). At the time of writing, no deaths by suicide had occurred in participants in the ABCD study. Since the proportions of suicidal plans and attempts were low, we did not analyse these features separately but grouped them together with suicidal ideation. For the sake of brevity, we refer collectively to suicidal ideas and behaviours as suicidality.
The divergence of self-reports and caregiver reports prompted us to examine child-reported and caregiver-reported suicidality separately. The child-reported suicidality group comprised 684 children who self-reported any suicidal feature either in the past or currently. The caregiver-reported suicidality group comprised 654 children who had any suicidal feature either in the past or currently, according to their caregiver. Individuals that overlapped between the previous two groups comprised the concordantly-reported suicidality group. The never-suicidal group comprised 6854 children who had never expressed any suicidal feature based on concordant self-reports and caregiver reports (table 2). Groups reporting suicidality generally had higher levels of personal and parental psychopathology and social adversity compared with the never-suicidal group (table 2).
Table 2:
Descriptive statistics of the study sample
| Child-reported suicidality (n=684) | Caregiver-reported suicidality (n=654) | Concordantly- reported suicidality (n=198) | Never suicidal (n=6854) | |
|---|---|---|---|---|
| Child characteristics | ||||
| Age (months) | 118·6 (7·3) | 118·8 (7·4) | 120·3 (7·3)*† | 118·5 (7·2) |
| Sex | ||||
| Male | 407 (59·5%)*† | 421 (64·4%)*† | 131 (66·2%)*† | 3538 (51·6%) |
| Female | 277 (40·5%) | 233 (35·6%) | 67 (33·8%) | 3316 (48·4%) |
| Race | ||||
| White | 325 (47·5%) | 332 (50·8%) | 102 (51·5%) | 3367 (49·1%) |
| Black | 112 (16·4%) | 76 (11·6%)* | 24 (12·1%) | 1058 (15·4%) |
| Hispanic | 142 (20·8%) | 137 (20·9%) | 41 (20·7%) | 1560 (22·8%) |
| Asian | 21 (3·1%) | 19 (2·9%) | 5 (2·5%) | 171 (2·5%) |
| Other | 84 (12·3%) | 90 (13·8%)* | 26 (13·1%) | 698 (10·2%) |
| Fluid intelligence | 94·9 (18·1) | 93·9 (17·5)* | 93·0 (16·9)* | 95·6 (17·2) |
| Crystallised intelligence | 106·7 (18·2) | 106·4 (17·8) | 107·8 (18) | 106·6 (18·7) |
| Child Behavior Checklist total problems (t score) | 52·0 (11·6) *† | 57·8 (11·0)*† | 59·7 (10·7)*† | 45·3 (10·8) |
| Positive school involvement | 12·1 (2·7) *† | 12·3 (2·7)*† | 11·7 (3)*† | 13·2 (2·3) |
| School disengagement | 4·1 (1·6)*† | 4·1 (1·6)*† | 4·2 (1·7)*† | 3·7 (1·3)*† |
| Positive view of school environment | 18·8 (3·3) *† | 19·0 (3·4)*† | 18·3 (3·6)*† | 20·0 (2·8) |
| Screen time during weekdays (h per day) | 4·2 (3·5)*† | 3·9 (3·4)*† | 4·2 (3·6)*† | 3·4 (3·1) *† |
| Screen time during weekends (h per day) | 5·9 (4·2) *† | 5·3 (4·1)*† | 6·0 (4·6)*† | 4·6 (3·6) |
| Experienced stressful events | 286 (41·8%)*† | 321 (49·1%)*† | 98 (49·5%)*† | 2355 (34·4%) |
| Any caregiver reported common childhood medical condition | 311 (45·5%) | 328 (50·2%)*† | 97 (49·0%) | 2898 (42·3%) |
| Any caregiver reported lifetime psychiatric diagnosis | 530 (77·5%)*† | 588 (89·9%)*† | 185 (93·4%)*† | 4308 (62·9%) |
| Pregnancy and birth | ||||
| Pregnancy was planned | 367 (53·7%)* | 349 (53·4%)* | 104 (52·6%) | 4037 (58·9%) |
| Maternal medical conditions during pregnancy | 0·7 (1) | 0·7 (1·1) | 0·6 (1) | 0·6 (1) |
| Mother used medication during pregnancy | 85 (12·4%) | 94 (14·4%)* | 31 (15·7%)* | 756 (11·0%) |
| Mother used alcohol or substances during pregnancy | 89 (13·0%)*† | 90 (13·8%)*† | 28 (14·1%)* | 574 (8·4%) |
| Born prematurely | 75 (11·0%) | 75 (11·5%) | 15 (7·6%) | 697 (10·2%) |
| Obstetric complications | 0·4(1)* | 0·4 (0·8)*† | 0·5 (1)*† | 0·3 (0·7) |
| Birth weight (ounces) | 117 (20·7) | 118·1(19·4) | 118·8(19) | 117·9 (19·7) |
| Parent or caregiver characteristics | ||||
| Primary caregiver education level | 16·5 (2·7) | 16·8 (2·5)* | 17·0 (2·4)* | 16·5 (2·9) |
| Primary caregiver in full employment | 466 (68·1%) | 414 (63·3%) | 130 (65·7%) | 4736 (69·1%) |
| Maternal age when child was born (years) | 28·8 (6·3)* | 29·0 (6·5) | 28·7 (6·4) | 29·3 (6·3) |
| Paternal age when child was born (years) | 31·2 (7·1) | 31·2 (7·2) | 30·7 (7·1) | 31·7 (7·1) |
| Parental help-seeking for mental health problems (one parent only) | 204 (29·8%)* | 230 (35·2%)*† | 69 (34·8%)* | 1723 (25·1%) |
| Parental help-seeking for mental health problems (both parents) | 111 (16·2%)* | 144 (22·0%)*† | 50 (25·3%)*† | 814 (11·9%) |
| Family environment | ||||
| Family income level in the preceding 12 months | 7·1 (2·3) | 7·1 (2·3) | 7·2 (2·1) | 7·2 (2·4) |
| Number of household cohabitants | 4·3 (1·4)* | 4·2 (1·4)*† | 4·2 (1·4)* | 4·5 (1·5) |
| Primary caregiver cohabiting with partner | 476 (69·6%)* | 445 (68·0%) | 134 (67·7%)* | 5053 (73·7%) |
| Family conflict (caregiver report) | 2·8 (2·1)*† | 3·4 (2·1)*† | 3·3 (2·2)*† | 2·4 (1·9) |
| Family conflict (child report) | 2·8 (2·2)*† | 2·6 (2·1)*† | 3·1 (2·3)*† | 1·9 (1·9) |
| Parental supervision | 4·2 (0·6)*† | 4·3 (0·6)*† | 4·1 (0·6)*† | 4·4 (0·5) |
| Perceived primary caregiver’s warmth | 2·7 (0·4)*† | 2·7 (0·4)*† | 2·7 (0·4)*† | 2·8 (0·3) |
| Perceived secondary caregiver’s warmth | 2·6 (0·4)*† | 2·6 (0·4)*† | 2·5 (0·5)*† | 2·7 (0·4) |
| Neighbourhood environment | ||||
| Area Deprivation Index | 41·1 (27·3) | 40·5 (27) | 40·0 (27) | 39·4 (27·6) |
| Area level of pollution | 7·6 (2·7) | 7·5 (2·6) | 7·4 (2·6) | 75 (2.5) |
| Neighbourhood security (caregiver report) | 3·7 (1)* | 3·7 (1)* | 3·7 (1)* | 4·0 (1·1) |
| Neighbourhood security (child report) | 3·8 (1·2)*† | 3·9 (1·2)* | 3·8 (1·2)* | 4·0 (1·1) |
Data are mean (SD) or n (%). Definitions of all variables are in the appendix (pp 5·8).
Denotes group differences at p<0·05 (uncorrected) between the never-suicidal group and each of the groups that endorsed suicidality.
Denotes group differences at p<0·001 following Bonferroni correction between the never-suicidal group and each of the groups that endorsed suicide.
Because of low inter-informant agreement, we estimated separate regression models with child-reported suicidality, caregiver-reported suicidality, and concordantly-reported suicidality as dependent variables. All models were significant (p<0·0001 for all), explaining a mean of 24% of the variance in child-reported suicidality, 38% of the variance in caregiver-reported suicidality, and 57% of the variance in concordantly-reported suicidality. Figure 1 highlights the results for child-reported suicidality as shown in table 3. Figure 2 is a visual representation of the reproducibility of the statistically significant associations between each predictor variable with child-reported, caregiver-reported, and concordantly-reported suicidality across all bootstrapped regression models. Regardless of informant, child psychopathology—as captured by the CBCL total score—and child-reported family conflict were the most robust risk factors for suicidality (table 3). Similar results were obtained when substituting the CBCL total score with any CBCL subscale in the regression models. Additionally, the risk of child-reported suicidality increased with higher weekend screen use time (OR 1∙3, 95% CI 1∙2–1∙7) and reduced with greater parental supervision and school involvement (for both OR 0∙8, 95% CI 0∙7–0∙9; table 3). Furthermore, caregiver-reported suicidality was positively associated with caregiver educational level (OR 1∙3, 95% CI 1∙1–1∙5) and male sex in children (1∙5, 1∙1–2·2), and inversely associated with the number of household cohabitants (0∙8, 0∙7–1·0; table 3).
Figure 1: Risk and protective factors for child-reported suicidality.
Mean odds ratios and 95% CIs of bootstrapped logistic regression models for child-reported suicidality
Table 3:
Significant and reproducible risk and protective factors for suicidality in mid-childhood
| Odds ratio | 95% CIs |
β | Wald statistic | Proportions of regression models when variable was significant | ||
|---|---|---|---|---|---|---|
| Lower | Upper | |||||
| Child-reported suicidality | ||||||
| Child Behavior Checklist total problems | 1·7 (0·1) | 1·5 (0·1) | 2·0 (0·1) | 0·6 (0·1) | 7·6 (0·6) | 100·0% |
| Family conflict (child report) | 1·4 (0·1) | 1·3 (0·1) | 1·7 (0·1) | 0·4 (0·1) | 5·4 (0·7) | 100·0% |
| Screen use during the weekends | 1·3 (0·1) | 1·2 (0·1) | 1·7 (0·1) | 0·3 (0·1) | 3·3 (0·6) | 96·8% |
| Parental supervision | 0·8 (0·1) | 0·7 (0·03) | 0·9 (0·04) | −0·2 (0·1) | −3·1 (0·6) | 93·4% |
| Positive school involvement | 0·8 (0·04) | 0·7 (0·03) | 0·9 (0·04) | −0·2 (0·1) | −2·7 (0·5) | 74·1% |
| Caregiver-reported suicidality | ||||||
| Child Behavior Checklist total problems | 3·3 (0·2) | 2·8 (0·2) | 4 (0·3) | 1·2 (0·1) | 13 (0·4) | 100·0% |
| Family conflict (child report) | 1·3 (0·1) | 1·1 (0·1) | 1·6 (0·1) | 0·3 (0·1) | 3·5 (0·7) | 98·1% |
| Male sex (child) | 1·5 (0·1) | 1·1 (0·1) | 2 (0·1) | 0·4 (0·1) | 2·8 (0·5) | 80·7% |
| Number of household cohabitants | 0·8 (0·03) | 0·7 (0·03) | 1 (0·04) | −0·2 (0·04) | −2·7 (0·5) | 73·5% |
| Primary caregiver education level | 1·3 (0·1) | 1·1 (0·04) | 1·5 (0·1) | 0·2 (0·04) | 2·6 (0·4) | 61·0% |
| Concordantly-reported suicidality | ||||||
| Child Behavior Checklist total problems | 4·8 (0·8) | 3·2 (0·5) | 7·4 (1·5) | 1·6 (0·2) | 7·4 (0·4) | 100·0% |
| Family conflict (child report) | 1·8 (0·2) | 1·2 (0·1) | 2·5 (0·4) | 0·5 (0·1) | 3 (0·6) | 88·9% |
Data are mean (SD) across all significant regression models.
Figure 2: Stability of predictor variables for suicidality.
Heatmap showing the percentage of times (columns) that the contribution of each predictor variable (rows) was statistically significant in bootstrapped logistic regression models for child-reported, caregiver-reported, and concordantly-reported suicidality. Further details in the appendix (pp 16–18).
The results of our analyses in the complete sample recapitulated our findings in the imputed sample (appendix pp 11–18). We observed the same pattern in sex-specific analyses (appendix pp 19–20). Childhood psychopathology and child-reported family conflict remained the most robust risk factors for suicidality for boys and girls, regardless of informant. We noted sex-related variation in the consistency of some associations. Screen time and parental supervision were more consistently associated with suicidality in boys than in girls.
Discussion
To our knowledge, this is the largest study to date to investigate risk and protective factors associated with suicidal thoughts and behaviours in a sample comprised exclusively of children aged 9–10 years. The key findings are that around eight in every 100 children in this age group experience suicidality; child–caregiver concordance in suicidality reporting is low across the spectrum of suicidal features and indicates that suicidality in children cannot be reliably assessed by parental report alone; and after modelling a wide range of factors, child psychopathology and child-reported family conflict showed the most consistent and robust associations with suicidality. Furthermore, child-reported suicidality was associated with greater screen time. Greater parental supervision and positive school engagement emerged as protective factors as they appeared to mitigate the likelihood of child-reported suicidality.
The results of this study provide additional support for societal concerns of suicidality in people as young as 9–10 years. The rates of suicidal ideation, suicidal plans, and suicide attempts in the ABCD sample indicate that suicidality in this age group is dominated by suicidal thoughts, but suicide attempts are rare. The rates of self-reported suicidal ideation observed in this study are similar to those found in the Philadelphia Neurodevelopmental Cohort (PNC), a community US sample of 6151 individuals aged 11–21 years, in which 8% of pre-pubertal participants endorsed suicidal thoughts.17 However, suicidality rates reported in community samples of older adolescents, when suicidality is known to begin peaking, are generally higher.11,12,14,16,17 For example, in the National Comorbidity Survey Replication Adolescent Supplement the rates of suicidal ideation, suicidal plans, and suicide attempts in 6483 adolescents aged 13–18 years, were 12∙1%, 4∙0%, and 4∙1%, respectively.12
Agreement between child-reported and caregiver-reported suicidality was low, which has major implications regarding the assessment of suicide risk in this age group. Child–caregiver agreement ranged between a Cohen’s κ of almost 0 for the different types of suicidal ideation to 0∙2 for previous suicide attempts. These values are substantially lower than those reported in older adolescents in the PNC, in which κ for youth–parent concordance for suicidal thoughts was 0∙17 and 0∙46 for suicide attempts.19 Child–caregiver divergence is not uncommon, and the reasons might vary depending on the issue being assessed. Studies in adolescents and young adults suggest that non-disclosure of suicidality mainly involves fears of stigmatisation, communication difficulties, and inadequate social and family support.31 Collectively, this evidence underscores the need for separate and independent assessment of suicidality in parents and children.
The pattern emerging from simple comparisons between never-suicidal children and children who experience suicidality (regardless of informant) suggests that suicidality is associated with various factors reflecting personal, family, and social adversity. Child psychopathology and family conflict emerged as the most robust risk factors for suicidality, regardless of informant. The link between higher psychopathology and suicidality is well documented in young people11,12,14,16,17 and adults.9 Psychopathology in this study was associated with up to an almost five-times increase in the likelihood of suicidality, regardless of whether the CBCL scores for total, internalising, or externalising problems were used in the analyses. It could be the degree rather than the precise nature of psychopathology that is important for suicidality in this age group. Given the rarity of suicide attempts in this sample, the present results primarily inform about the correlates of suicidal ideation in 9–10-year-olds. Previous research in adolescents found that individuals who had attempted suicide differed from those who only experienced suicidal ideation in two respects: those attempting suicide were more likely to have been diagnosed with depressive, behavioural, or anxiety disorders and were more likely to have been exposed to self-harming behaviour in a friend or family member.11 Personal psychopathology (at the syndromal and symptom level) could be a powerful common predictor of suicidal thoughts and attempts, the latter being more likely with increasing severity and persistence of the mental health problems.
Young people rely on their family for general support and for the development of healthy patterns of attachment and emotional self-regulation.32 Even after accounting for psychopathology, children who reported family conflict were 30–75% more likely to experience suicidality. This observation reinforces previous reports that showed consistent associations between family dysfunction, suicide risk, and death by suicide in children.33 We did not observe a robust association between suicidality and any ethnic origin despite previous reports of increased risk of death by suicide in black youth.33,34 Therefore, it is possible that, unlike death by suicide, suicidal ideation and non-fatal attempts might not be influenced by race. Further studies could be helpful to investigate whether higher levels of suicide in black young people are mediated by factors relating primarily to help-seeking behaviour and access to support.
We identified additional risk factors that were dependent on informant. Child-reported suicidality was associated with greater time using a computer during weekends, which extends previous reports of an association between screen time and psychopathology in youth.23–25 For example, in the Avon Longitudinal Study of Parents and Children, a UK-based prospective cohort study of 14 666 young people, the risk for depression was positively associated with screen use (watching television, using a computer, and texting).23 The data available through the ABCD study do not allow us to comment on the underlying mechanisms, which may be causal (eg, increased exposure to cyberbullying or negative social comparisons) or correlative (eg, associated with social withdrawal or avoidance). Caregivers with higher educational levels were more likely to report suicidality; this effect was not found for child-reported suicidality, suggesting that caregiver education could be linked to parental awareness. Univariate analyses showed an over-representation of boys in both child-reported and caregiver-reported suicidality groups. Further studies are needed to clarify whether this finding represents true male disadvantage or increased disclosure by boys. The association between male sex and suicidality was significant for caregiver reports, which accords with similar findings in the PNC study, showing that parental reports of suicidality were more likely for young boys than for young girls.19
The richness of the ABCD dataset enabled us to identify factors that mitigate the risk of reported suicidality. Higher levels of parental supervision and more positive school engagement were associated with reduced likelihood of a child endorsing suicidality. The measure of parental supervision used in the ABCD study captured parental oversight (ie, knowing where the children are, what they are doing, and with whom) and the frequency of parent–child contact.30 Positive school environment, as measured in the ABCD study, reflects a child’s perception of achievement and involvement in school activities.30 These results, together with similar findings in adolescents,35–38 confirm that positive parental and school-based experiences are robust sources of protection against youth suicidality, as they promote development of identity, self-esteem, and resilience in school-age children and adolescents.32 Additionally, living in households with more cohabitants reduced the likelihood of caregiver-reported suicidality. This association might reflect greater opportunities for support and positive social interactions within the child’s household. Alternatively, the higher number of cohabitants might dilute caregivers’ awareness of a child’s emotional states, including suicidality.
This study has many strengths and limitations. The size and representativeness of the sample and range of domains assessed in the ABCD study are unique strengths, which enabled us to quantify the effect of an array of predictors and identify those with robust and reliable positive and negative influences on childhood suicidal thoughts and behaviours. Limitations include the exclusion of measures regarding non-suicidal self-injury, possible recall bias in reporting previous suicidal thoughts or behaviours, and insufficient detail regarding the nature, timing, and frequency of life events. Additionally, the rarity of suicide attempts in the ABCD sample precluded the meaningful examination of correlates of suicidality in those who have attempted suicide compared with those who only endorsed suicidal ideation. The current study is based on cross-sectional data and thus we cannot assume that the associations reported are causal. Nevertheless, the ABCD sample presents an unprecedented resource for improving understanding of the evolution of suicidality and its correlates. The present study sets the foundation for future analyses using data from the planned follow-up assessments. Such analyses will inform about the temporal stability of the observed associations and their prognostic value for emerging suicide risk and death by suicide as the sample enters adolescence.
In conclusion, the findings reported here have immediate and practical implications as risk factors (childhood psychopathology, family conflict, and screen use) and protective influences (higher parental supervision and positive school engagement) are actionable and modifiable. The optimal way of offering support to children at risk is unclear.30 Nevertheless, meta-analytic evidence suggests that school-based awareness programmes could be the most successful public health interventions for reducing both suicidal behaviours and suicidal ideation.40 Therefore, this study provides further support for initiatives such as the Mental Health Packs for Schools from the Royal College of Psychiatrists in the UK, which aim to increase mental health awareness in young people. Increasing school and parental awareness of child psychopathology and providing parenting education and family support could be clear and attainable targets for early intervention and prevention strategies.
Supplementary Material
Research in context.
Evidence before this study
We searched PubMed for studies published in English from inception to Dec 31, 2019, with the search terms (“suicidality” or “suicidal thoughts” OR “suicidal ideation” OR “suicidal behaviour” OR “suicide” OR “suicide attempt” AND either “children” or “childhood”), and examined citations and reference lists of relevant publications. This search confirmed the paucity of population-based studies on suicidal ideation and behaviours in childhood compared with older adolescents and adults. Psychopathology and psychosocial adversity emerged as consistent risk factors, but the evidence for protective factors was scarce or inconclusive.
Added value of this study
The strengths of this study are fourfold. First, to our knowledge, it is the largest population-based investigation to focus specifically on late childhood. Previous studies in this age group have relied on smaller or clinical samples. Second, our results highlight the poor agreement between a child’s self-report and their caregiver reports of suicidal ideation and behaviours. Third, the comprehensive characterisation of participants enabled the examination of a range of potential risk and protective factors for childhood suicidal ideation and behaviours. Previous research typically examined a single or a restricted number of factors. Fourth, our investigation identified child psychopathology and child-reported family conflict as the most robust risk factors for suicidal ideation, whereas children reporting higher levels of parental supervision and more school engagement were less likely to ever have engaged in suicidal ideas or behaviours.
Implications of all the available evidence
Suicidal ideas and behaviours are experienced by eight in every 100 children aged 9·10 years. Low caregiver and child agreement in reporting indicates problems with disclosure and awareness that should be considered when assessing suicide risk in young children. This also underscores the importance of considering the informant in future studies of childhood suicidal ideas and behaviours. The identification of robust risk and protective factors indicates that optimizing interventions that aim to treat childhood psychopathology, increase school engagement, and provide family support might assist in mitigating the risk of suicide in this age group.
For more on the Adolescent Brain and Cognitive Development (ABCD) study see https://abcdstudy.org/
For more on the American Community Survey see https://www.census.gov/programs-surveys/acs
To request ABCD study data from the National Institute of Mental Health data archive see https://data-archive.nimh.nih.gov/abcd
For more on Mental Health Packs for Schools see www.rcpsych.ac.uk/mental-health/order-mental-health-packs-for-schools
For the Adolescent Brain Cognitive Development (ABCD) data used in this report see https://nda.nih.gov/study.html?id=721
For the full list of ABCD study supporters see https://abcdstudy.org/federal-partners.html
For a list of participating sites and study investigators in the ABCD study see https://abcdstudy.org/Consortium_Members.pdf
Acknowledgments
SF and GED were supported by grants from the US National Institutes of Mental Health (NIMH; R01MH113619 and R01MH116147). BL was supported by R01MH067924 and R01MH08024 from the NIMH, the Staunton Farm Foundation, and U01 DA041028 from the National Institute of Drug Abuse. BL is site co-investigator for the ABCD study at University of Pittsburgh. DAB was supported by R01 MH100155, R01 MH116652, P50 MH115838, U01 MH116923, and R01 MH11358 from the National Institute of Drug Abuse. Data used in the preparation of this Article were obtained from the Adolescent Brain Cognitive Development (ABCD) study, held in the NIMH data archive. The ABCD Study is supported by the National Institutes of Health (NIH) 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 online. A list of participating sites and a complete list of the study investigators can be found online. ABCD consortium investigators designed and implemented the study and provided data but did not necessarily participate in analysis or writing of this report. This manuscript reflects the views of the authors and does not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time.
Footnotes
Declaration of interests
DAB reports grants from the US National Institutes of Mental Health (NIMH) during the conduct of the study, grants from the NIMH, the American Foundation for Suicide Prevention, the Once Upon a Time Foundation, and the Beckwith Foundation, personal fees from ERT, UptoDate, Guilford Press, and Healthwise, and Klingenstein Third Generation Foundation, outside the submitted work. MP reports personal fees from Janssen, Otsuka, Angelini, Lundbeck, and Allergan, outside the submitted work. All other authors declare no competing interests.
Data sharing
The Adolescent Brain Cognitive Development data used in this report are available online on request. We used open source functions from the Comprehensive R Archive Network (version 3.5.3) and SPSS version 23.
See Online/Comment https://doi.org/10.1016/S2215–0366(20)30094–8
References
- 1.Global Naghavi M., regional, and national burden of suicide mortality 1990 to 2016: systematic analysis for the Global Burden of Disease Study 2016. BMJ 2019; 364: 194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.National Institute of Mental Health. Mental health information. Statistics. Suicide https://www.nimh.nih.gov/health/statistics/suicide.shtml (accessed Feb 5, 2020).
- 3.Burstein B, Agostino H, Greenfield B. Suicidal attempts and ideation among children and adolescents in US emergency departments, 2007–2015. JAMA Pediatr 2019; 173: 598–600. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Plemmons G, Hall M, Doupnik S, et al. Hospitalization for suicide ideation or attempt: 2008–2015. Pediatrics 2018; 141: e20172426. [DOI] [PubMed] [Google Scholar]
- 5.Copeland WE, Goldston DB, Costello EJ. Adult associations of childhood suicidal thoughts and behaviors: a prospective, longitudinal analysis. J Am Acad Child Adolesc Psychiatry 2017; 56: 958–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Herba CM, Ferdinand RF, van der Ende J, Verhulst FC. Long-term associations of childhood suicide ideation. J Am Acad Child Adolesc Psychiatry 2007; 46: 1473–81. [DOI] [PubMed] [Google Scholar]
- 7.Wilkinson P, Goodyer I. Non-suicidal self-injury. Eur Child Adolesc Psychiatry 2011; 20: 103–08. [DOI] [PubMed] [Google Scholar]
- 8.Barrocas AL, Hankin BL, Young JF, Abela JR. Rates of nonsuicidal self-injury in youth: age, sex, and behavioral methods in a community sample. Pediatrics 2012; 130: 39–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Franklin JC, Ribeiro JD, Fox KR, et al. Risk factors for suicidal thoughts and behaviors: a meta-analysis of 50 years of research. Psychol Bull 2017; 143: 187–232. [DOI] [PubMed] [Google Scholar]
- 10.Asarnow JR, Baraff LJ, Berk M, et al. Pediatric emergency department suicidal patients: two-site evaluation of suicide ideators, single attempters, and repeat attempters. J Am Acad Child Adolesc Psychiatry 2008; 47: 958–66. [DOI] [PubMed] [Google Scholar]
- 11.Mars B, Heron J, Klonsky ED, et al. What distinguishes adolescents with suicidal thoughts from those who have attempted suicide? A population-based birth cohort study. J Child Psychol Psychiatry 2019; 60: 91–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Nock MK, Green JG, Hwang I, et al. Prevalence, correlates, and treatment of lifetime suicidal behavior among adolescents: results from the National Comorbidity Survey Replication Adolescent Supplement. JAMA Psychiatry 2013; 70: 300–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.O’Connor RC, Rasmussen S, Hawton K. Distinguishing adolescents who think about self-harm from those who engage in self-harm. Br J Pychiatry 2012; 200: 330–35. [DOI] [PubMed] [Google Scholar]
- 14.Sampasa-Kanyinga H, Dupuis LC, Ray R. Prevalence and correlates of suicidal ideation and attempts among children and adolescents. Int J Adolesc Med Health 2017; 29: ijamh-2015–0053. [DOI] [PubMed] [Google Scholar]
- 15.Wilkinson P, Kelvin R, Roberts C, Dubicka B, Goodyer I. Clinical and psychosocial predictors of suicide attempts and nonsuicidal self-injury in the Adolescent Depression Antidepressants and Psychotherapy Trial (ADAPT). Am J Psychiatry 2011; 168: 495–501. [DOI] [PubMed] [Google Scholar]
- 16.Mars B, Heron J, Klonsky ED, et al. Predictors of future suicide attempt among adolescents with suicidal thoughts or non-suicidal self-harm: a population-based birth cohort study. Lancet Psychiatry 2019; 6: 327–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Barzilay R, Calkins ME, Moore TM, et al. Neurocognitive functioning in community youth with suicidal ideation: gender and pubertal effects. Br J Psychiatry 2019; 215: 552–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Chang SS, Chen YY, Heron J, Kidger J, Lewis G, Gunnell D. IQ and adolescent self-harm behaviours in the ALSPAC birth cohort. J Affect Disord 2014; 152–154: 175–82. [DOI] [PubMed] [Google Scholar]
- 19.Jones JD, Boyd RC, Calkins ME, et al. Parent-adolescent agreement about adolescents’ suicidal thoughts. Pediatrics 2019; 143: e20181771. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Thompson R, Dubowitz H, English DJ, et al. Parents’ and teachers’ concordance with children’s self-ratings of suicidality: findings from a high-risk sample. Suicide Life Threat Behav 2006; 36: 167–81. [DOI] [PubMed] [Google Scholar]
- 21.Zalsman G, Shoval G, Mansbach-Kleinfeld I, et al. Maternal versus adolescent reports of suicidal behaviors: a nationwide survey in Israel. Eur Child Adolesc Psychiatry 2016; 25: 1349–59. [DOI] [PubMed] [Google Scholar]
- 22.Common Sense Median. The Common Sense Census: media use by kids age zero to eight, 2017. https://www.commonsensemedia.org/research/the-common-sense-census-media-use-by-kids-age-zero-to-eight-2017 (accessed Feb 5, 2020).
- 23.Khouja JN, Munafò MR, Tilling K, et al. Is screen time associated with anxiety or depression in young people? Results from a UK birth cohort. BMC Public Health 2019; 19: 82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Przybylski AK, Weinstein N. Digital screen time limits and young children’s psychological well-being: evidence from a population-based study. Child Dev 2019; 90: e56–65. [DOI] [PubMed] [Google Scholar]
- 25.Suchert V, Hanewinkel R, Isensee B. Sedentary behavior and indicators of mental health in school-aged children and adolescents: a systematic review. Prev Med 2015; 76: 48–57. [DOI] [PubMed] [Google Scholar]
- 26.Garavan H, Bartsch H, Conway K, et al. Recruiting the ABCD sample: design considerations and procedures. Dev Cogn Neurosci 2018; 32: 16–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Alati R, Gunnell D, Najman J, Williams G, Lawlor D. Is IQ in childhood associated with suicidal thoughts and attempts? Findings from the Mater University Study of Pregnancy and its outcomes. Suicide Life Threat Behav 2009; 39: 282–93. [DOI] [PubMed] [Google Scholar]
- 28.Barch DM, Albaugh MD, Avenevoli S, et al. Demographic, physical and mental health assessments in the adolescent brain and cognitive development study: rationale and description. Dev Cogn Neurosci 2018; 32: 55–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Luciana M, Bjork JM, Nagel BJ, et al. Adolescent neurocognitive development and impacts of substance use: overview of the adolescent brain cognitive development (ABCD) baseline neurocognition battery. Dev Cogn Neurosci 2018; 32: 67–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Zucker RA, Gonzalez R, Feldstein Ewing SW, et al. Assessment of culture and environment in the Adolescent Brain and Cognitive Development Study: rationale, description of measures, and early data. Dev Cogn Neurosci 2018; 32: 107–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Levi-Belz Y, Gavish-Marom T, Barzilay S, et al. Psychosocial factors correlated with undisclosed suicide attempts to significant others: findings from the adolescence SEYLE study. Suicide Life Threat Behav 2019; 49: 759–73. [DOI] [PubMed] [Google Scholar]
- 32.Harter S. Emerging self-processes during childhood and adolescence In: Leary M, Tangney J, eds. Handbook of self and identity, 2nd edn New York: Guilford, 2012: 680–715. [Google Scholar]
- 33.Sheftall AH, Asti L, Horowitz LM, et al. Suicide in elementary school-aged children and early adolescents. Pediatrics 2016; 138: e20160436. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Bridge JA, Horowitz LM, Fontanella CA, et al. Age-related racial disparity in suicide rates among US youths from 2001 through 2015. JAMA Pediatr 2018; 172: 697–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kim HH. Parental overprotection and youth suicide behavior in low- and middle-income countries: a multilevel analysis of cross-national data. Int J Public Health 2019; 64: 173–84. [DOI] [PubMed] [Google Scholar]
- 36.Saffer BY, Glenn CR, David Klonsky E. Clarifying the relationship of parental bonding to suicide ideation and attempts. Suicide Life Threat Behav 2015; 45: 518–28. [DOI] [PubMed] [Google Scholar]
- 37.Kaminski JW, Puddy RW, Hall DM, Cashman SY, Crosby AE, Ortega LA. The relative influence of different domains of social connectedness on self-directed violence in adolescence. J Youth Adolesc 2010; 39: 460–73. [DOI] [PubMed] [Google Scholar]
- 38.Stallard P, Spears M, Montgomery AA, Phillips R, Sayal K. Self-harm in young adolescents (12–16 years): onset and short-term continuation in a community sample. BMC Psychiatry 2013; 13: 328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.King CA, Arango A, Kramer A, et al. Association of the youth-nominated support team intervention for suicidal adolescents with 11- to 14-year mortality outcomes: secondary analysis of a randomized clinical trial. JAMA Psychiatry 2019; 76: 492–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Zalsman G, Hawton K, Wasserman D, et al. Suicide prevention strategies revisited: 10-year systematic review. Lancet Psychiatry 2016; 3: 646–59. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.