Risk and Protective Factors of Self-harm and Suicidality in Adolescents: An Umbrella Review with Meta-Analysis

The most reviewed risk factor for suicidal behavior in the literature was antidepressants, including selective serotonin reuptake inhibitor (SSRI) and the new generation of antidepressant exposure. These studies focused on youth aged between 9–25 years, including participants diagnosed with a depressive disorder. Antidepressant use was measured through randomized control trials and observational studies of Mirtazapine, Fluoxetine, Paroxetine, Sertraline, Citalopram, Fluvoxamine, Venlafaxine, and Escitalopram. In the included randomized control trials (Dubicka et al., 2010; Hetrick et al., 2012), the pooled OR were derived from comparisons between the antidepressant and placebo groups. The suicide behaviors were reported as suicide-related outcomes or adverse events in these studies, including suicide attempts, completed suicide, and suicidal self-harm. In the observational studies (Barbui et al., 2009; Li et al., 2022), antidepressant intake was included as an exposure in cohort and case-control studies. Barbui et al. (2009) found that the OR for the overall relationship between SSRI and suicidal behavior was 1.92 (95% CI, 1.51–2.44). Dubicka et al. (2010) found the OR for antidepressant exposure was 1.70 (95% CI, 1.13–2.54). Hetrick et al. (2012) examined the association between antidepressant use and the combined outcomes of suicidal behavior and ideation, resulting in a pooled RR of 1.58 (95% CI 1.02- 2.45). The latest review on this topic by Li et al. (2022) featured the most significant number of individual participants from 11 studies and found that the pooled RR of antidepressant exposure (including SSRI) was 1.38 (95% CI: 1.16–1.64), and for SSRI exposure only was 1.28 (95% CI: 1.09–1.51). The result indicated that the risk of suicidal behavior in youth exposed to antidepressants was 38% higher (Table 1).

Dubicka et al. (2010) found suicidal thoughts, which occurred in nine of 738 (1.2%) young people with depression treated with antidepressants compared to five in 634 of those treated with a placebo (2.6%). The results indicated a significant association between the risk of youth suicidal thoughts and antidepressant use (OR:1.45). Both Dubicka et al. (2010) and Hetrick et al. (2012) explored the combined outcomes of suicidal behavior and ideation (Table 2).

Dubicka et al. (2010) also reported that self-harm occurred in 19 of 569 (3.3%) young people with depression treated with antidepressants compared to 12 in 469 of those treated with a placebo (2.6%). The results represented a significant association between the risk of youth self-harm and antidepressant use, with an OR of 1.44 (Table 3).

Despite the identified reviews being from different types of studies (i.e., RCT vs. observational) and the corresponding pooled effect size using OR (cross-sectional) or RR (cohort), their results in antidepressants and suicidal behavior are in the similar range between 1.70 and 1.92, which are higher than the effect size of suicidal ideation. The same exposure also had a similar effect size in self-harm- a different outcome from suicide ideation. Review studies of RCT have not explicitly reported results for subgroup populations (different genders or age groups) and included both clinical samples and samples from communities. The observational review (Barbui et al., 2009), including different age groups, suggested a promoting effect of SSRI exposure among adolescents but a protective effect in adults and the elderly. The data synthesis from the current umbrella review, including the recent observational review (Li et al., 2022), suggested that the increased risk of suicidal behavior for SSRIs and other antidepressants was similar in both children and adolescents.

Three studies explored the relationship between bullying/peer victimization and suicidal behavior (Holt et al., 2015; Koyanagi et al., 2019; Van Geel et al., 2014). Koyanagi et al. (2019) included studies of the GSHS from 48 low and middle-income countries; the other two reviews included participants aged between 9–23 years from low-, middle-, and high-income countries. These reviews included studies that measured bullying according to the Centers for Disease Control and Prevention’s uniform definition and included the following components: providing a definition of bullying/peer victimization followed by questions, measuring aggressive acts, measuring power imbalance or differential, or direct asking if the participant was bullied. The OR ranged from 2.78 to 3.26 for the overall relationship between bullying victimization and suicidal behavior. Holt et al. (2015) reviewed 15 studies that assessed the relationship between bullying perpetration and suicidal behavior, measured through behaviorally based questions and directly asking students if they had bullied others. They found a significant OR of 2.62 (95% CI, 1.51–4.55), indicating a 1.62-fold higher risk of suicidal behavior for youth who perpetrated bullying (Table 1).

Holt et al. (2015) and Van Geel et al. (2014) explored bullying victimization as a risk for suicidal ideation from 23 and 24 individual studies, respectively; the pooled OR was 2.23 and 2.34 for the overall relationship between bullying victimization and suicide ideation. These results indicated a 1.2–1.3-fold higher risk of suicidal ideation for youth exposed to bullying. Holt et al. (2015) reviewed studies that assessed the relationship between bullying perpetration and suicidal ideation and found a pooled OR of 2.12. Kowalski et al. (2014) investigated cyberbullying specifically. They found that youth who reported high levels of cyberbullying victimization were also likely to report high levels of suicidal ideation with a correlation coefficient (r: 0.27) (Table 2).

Heerde and Hemphill (2019) and Wang et al. (2022) reviewed the association between general bullying victimization (both traditional and cyberbullying) and youth self-harm from three and seven studies, respectively. The two reviews found that OR ranged from 1.98 to 2.34 for the relationship between bullying victimization and youth self-harm. Heerde and Hemphill (2019) explored traditional bullying and cyberbullying separately; they reported the association between cyberbullying victimization and risk of self-harm with a pooled OR of 3.55 (Table 3).

The synthesis of these reviews and meta-analysis identified both bullying victims and perpetration as common risk factors for suicide behavior, ideation, and self-harm. Risk is higher in suicidal behavior than ideation from bully victimization, and there is a similarity between suicidal ideation and self-harm. The cross-review comparison suggested that bully victimization has a slightly higher risk of suicide behavior and ideation compared to bully penetration; cyberbullying has a higher risk than traditional bullying in self-harm. All reviews included samples from schools and communities; one included mental health clinics and another setting. Most studies included in these reviews are cross-sectional, with a few longitudinal design studies included (Holt et al., 2015).

One review (Holt et al., 2015) found a significant moderating effect from the country of origin of the study in both bullying perpetration and victimization. Gender was reported in individual studies as a moderator to associations between bullying and suicidality (Klomek et al., 2009) but not significant in Holt’s meta-analysis. The other moderators with non-significant results included the sampling methods (cluster, stratified, simple random sample or census) and measurements of bullying (single vs multiple) (Van Geel et al., 2014). The meta-analysis of Koyanagi et al. (2019) also tested different forms of bullying in the order of their effect sizes, which are religion, race/nationality/color, physical bullying, sexual bullying, exclusion, and being made fun of for physical appearance.

Liu et al. (2019) conducted a review including 34,933 participants to explore sleep disturbances defined as difficulty sleeping, insomnia symptoms, and poor sleep, measured by various standardized scales and questionnaires. Their review included ten cross-sectional and four longitudinal studies and found that sleep disturbances were associated with a higher risk of adolescent suicide attempts (OR: 1.92) and a higher risk of adolescent suicidal ideation (OR: 2.35) as well as suicidal ideation with a plan (OR: 1.58). The significant moderators on the association between sleep disturbances and suicide ideation were using insomnia symptoms in measurement, age, and reliable sleep measures. The female percentage in studies was a significant positive moderator on the association for suicide attempts.

Miranda-Mendizabal et al. (2019) and Van Meter et al. (2022) all identified that suicidal attempts were more prevalent in female youth than male youth. These two reviews comprised youth from multiple countries. Miranda-Mendizabal et al. (2019) found that female youth were almost twice as likely to experience suicidal behavior (OR: 1.96). Van Meter et al. (2022) found an 8.5% prevalence in female youth reporting suicidal attempts compared to a 4.9% prevalence in males (Table 1). In addition, both reviews found that suicidal ideation with a plan was higher in female youth than male youth, with the pooled prevalence for females ranging from 11.4 to 19.8% (Table 2). The most reported risk factor for self-harm was gender; Gillies et al. (2018), Xiao et al. (2022), and Wang et al. (2022), all identified that the risk of self-harm was more prevalent in female youth compared to male youth. Gillies et al. (2018) reviewed 261 studies, including 597,548 participants, and reported that the female gender had a pooled RR of 1.72. In a review of 43 studies with 107,285 male and 102,473 female participants, Xiao et al. (2022) found a 25.4% NSSI prevalence for female adolescents compared to a 22% prevalence for males. In a review of 8 studies, Wang et al. (2022) found a significant association between female gender and a significant risk of self-harm (OR: 2.89) (Table 3).

Miranda-Mendizábal et al. (2017) studied the association between sexual orientation and youth suicidal behavior, including six studies with 22,117 participants. The pooled OR for the overall relationship between sexual orientation and youth suicidal behavior is 2.26. Marshal et al. (2011) studied the association between sexual orientation and youth suicidality (behavior and suicidal ideation) through 19 studies with 122,955 participants, revealing a pooled OR of 2.92. The review also investigated specifically bisexual youth, including four studies with 42,413 participants, and found a significant association between bisexual youth and suicidal behaviors and ideations (OR: 4.92) (Tables 1, 2).

Six reviews explored the associations between various mental health disorders (including anxiety disorders and major depressive disorders) and suicidal behavior. Gili et al. (2019) found that any presence of mental health disorders increased the risk of suicidal attempts (OR: 3.57). Miranda-Mendizabal et al. (2019) investigated mental health disorders and suicidal behavior with male and female gender separately. The mental health issues associated with female suicidal behavior included anxiety disorder, drug abuse disorder, major depressive disorder, depressive symptoms, and any mental disorder or abuse. The OR ranged from 1.15 to 4.49 for the overall relationship between mental health disorders and female suicidal behavior. Mental health issues associated with male suicidal behavior identified in this review included anxiety disorder, major depressive disorder, and any mental disorder or abuse, with the OR ranging from 3.79 to 6.07. Major Depressive Disorder had the highest association with suicidal behavior for female youth (OR: 4.49) and male youth (OR: 6.07) (Table 1). Hauser et al. (2013) reviewed 11 studies that specifically explored the link between bipolar disorder and suicidality. Findings from the review showed a 21.3% prevalence of past suicide attempts and a 25.5% prevalence of suicide attempts recorded at the time of the study. The two latest review studies on mental health and suicidal behaviors included Leigh et al. (2023) on the association between social anxiety and risk of suicidality and O’halloran et al. (2022) on the prevalence of suicidal behaviors in youth with Autism. Leigh et al. (2023) suggested a positive linear relationship between social anxiety and suicidal attempts (correlation coefficient r: 0.10) and current suicide risk (pooled correlation coefficient r: 0.24). O’Halloran et al. (2022) found an 8.3% prevalence (95% CI: 3.6–18.2%) of suicidal behavior in autistic youth. Miranda-Mendizabal et al. (2019) also found an association between a family history of mental disorders and abuse and the risk of male youth suicide attempts (OR: 2.63).

O’halloran et al. (2022), Das Neves Peixoto et al. (2017) and Hauser et al. (2013) explored the association between a range of mental health disorders, including social anxiety, bipolar disorder, Autism, and suicidal ideation. O’halloran et al. (2022)‘s review of 22 studies found a pooled 25.2% (95% CI 18.2–33.8) prevalence of suicidal ideation in autistic youth. Das Neves Peixoto et al. (2017) and Hauser et al. (2013) both explored bipolar disorder as a risk factor for youth suicidal ideation. Das Neves Peixoto et al. (2017) obtained a pooled RR of 2.94 (95% CI: 2.30, 3.78), showing that youth with bipolar disorder were more vulnerable to suicidal ideation (Table 2).

Wang et al. (2022) explored the association between self-harm risk and mental disorders, including depression symptoms, anxiety symptoms, personality disorders, adaption disorders, emotional scale scores, and psychological symptoms. The review of 21 studies found a pooled OR of 1.89 (Table 3).

A review of five studies by Miranda-Mendizabal et al. (2019) identified previous suicidal ideation as a significant risk factor for suicidal behavior in both female youth (OR: 4.39) and male youth (OR: 3.97); female youth were almost six times more at risk to attempt suicide if they have had previous suicide attempts (OR: 6.96). A review including 15 datasets and 37,784 participants by Gillies et al. (2018) found that suicide attempts were significantly higher in adolescents who self-harmed (RR: 9.14). From 41 comparisons, Castellví et al. (2017) also reported an over-3-fold risk for those with any previous self-injurious thoughts and behaviors (OR: 3.48, 95% CI: 2.71–4.43), suicide ideation history (OR: 3.26), previous NSSI (OR: 2.26), and previous suicide attempts had the largest pooled OR of 5.56. Miranda-Mendizabal et al. (2019) also found an association between a family’s previous suicidal behavior and the risk of female youth suicide attempts (OR: 2.84).

Two reviews investigated behavior issues such as cannabis use, legal problems, and externalizing symptoms. In one review of three studies, including 13,687 participants, Gobbi et al. (2019) found an association between youth cannabis use and suicide attempts (OR: 3.46). The other review by Soto-Sanz et al. (2019) explored the association between externalizing symptoms, including problems related to aggressiveness, inattentiveness, disobedience, and criminal behavior, and youth suicidal behavior from 21 studies. Their finding suggested a 1.59-fold higher risk of suicidal behavior for youth who experience externalizing symptoms and a 2.36-fold risk for youth who experience legal problems (Table 1).

In reviews of behavior issues and suicide ideation, Gobbi et al. (2019) found an association between youth cannabis use and suicidal ideation (OR:1.50). Epstein et al. (2020) explored the association between school absenteeism and youth suicidal ideation from 34 studies found a significant pooled OR of 1.20, equivalented to 20% increased risk of suicidal ideation (Table 2).

Epstein et al. (2020) also investigated school absenteeism as a risk factor for self-harm and found a significant pooled OR of 1.37. Wang et al. (2022) explored the association between self-harm risk and behavior issues in adolescents, which included internet addiction, alcohol/substance use, smoking, problematic mobile phone use, having run away from home, suicide attempts, internet/mobile phone abuse, intentional misuse of prescription medications, avoidance, opioid misuse, sedative misuse, and gaming disorder. The review included 21 studies, which resulted in a pooled OR of 2.36. Xiao et al. (2022) explored substance use as a risk factor for self-harm. They found a higher prevalence of self-harm in adolescents with a smoking history (24.7%) versus non-smoking adolescents (10.1%), as well as a higher prevalence of self-harm in adolescents with a history of alcohol consumption (24.4%) versus non-drink adolescents (9.3%) (Table 3).

Two of the included reviews discuss school preventative features as protective factors against youth suicidal behaviors. Marraccini and Brier (2017) investigated school connectedness as a protective factor. The included studies used various measurements ranging from single-item questions to multi-construct measuring instruments. The review, which included ten studies with 57,637 participants, found that school connectedness is associated with reduced reports of suicide attempts in general youth (OR: 0.59). Findings were consistent when exploring five studies regarding suicidality in high-risk youth (OR: 0.60) and four studies examining sexual minority youth (OR: 0.61). Gijzen et al. (2022) explored the use of school-based interventions (such as Signs of Suicide, Headstrong, Good Behavior Game, and Mastery Learning) with the primary aims of addressing suicidal thoughts and behaviors (STB) or related mental health outcomes, e.g., aggressive, and disruptive behaviors. The review of 5 studies found that prevention of STBs at school was a significant modifier of effect for suicidal behaviors (Corrected effect size Hedges’ g: 0.30) (Table 1).

Marraccini and Brier (2017) also explored school connectedness as a protective factor for adolescent suicidal ideation. Their review found a pooled OR of 0.53, indicating that adolescents who experience school connectedness are less likely to experience suicidal ideation. Gijzen et al. (2022) investigated the use of school-based STB interventions. The review of 7 studies with 19,803 participants found that prevention of STBs at school was a significant modifier of effect for suicidal ideation (g: 0.15) (Table 2).

Chiu et al. (2018) investigated sleep duration as a protective factor of adolescent suicidality. They found a pooled OR of 0.52, suggesting a lower risk of suicide attempts from adolescents with longer sleep duration; similarly, a pooled OR of 0.55 and 0.50 for suicidal ideation and ideation with a plan. Of 11 studies with 446,033 participants, their review also identified a significant nonlinear dose-response relationship between the risk of adolescent suicide attempts, indicating that the lowest risk of adolescent suicide attempts was observed with a sleep of 8–9 hours (Table 1). For ideation with a plan, the study found that risk decreased by 11% for every 1-hour increase in adolescent sleep duration (Table 2).