Global Lifetime and 12-Month Prevalence of Suicidal Behavior, Deliberate Self-Harm and Non-Suicidal Self-Injury in Children and Adolescents between 1989 and 2018: A Meta-Analysis

Objective: This meta-analysis aimed to estimate the global lifetime and 12-month prevalence of suicidal behavior, deliberate self-harm and non-suicidal self-injury in children and adolescents. Methods: A systematic search for relevant articles published between 1989 to 2018 was performed in multiple electronic databases. The aggregate 12-month and lifetime prevalence of suicidal behavior, deliberate self-harm, and non-suicidal self-injury were calculated based on the random-effects model. Subgroup analyses were performed to compare the prevalence according to school attendance and geographical regions. Results: A total of 686,672 children and adolescents were included. The aggregate lifetime and 12-month prevalence of suicide attempts was 6% (95% CI: 4.7–7.7%) and 4.5% (95% CI: 3.4–5.9%) respectively. The aggregate lifetime and 12-month prevalence of suicidal plan was 9.9% (95% CI: 5.5–17%) and 7.5% (95% CI: 4.5–12.1%) respectively. The aggregate lifetime and 12-month prevalence of suicidal ideation was 18% (95% CI: 14.2–22.7%) and 14.2% (95% CI: 11.6–17.3%) respectively. The aggregate lifetime and 12-month prevalence of non-suicidal self-injury was 22.1% (95% CI: 16.9–28.4%) and 19.5% (95% CI: 13.3–27.6%) respectively. The aggregate lifetime and 12-month prevalence of deliberate self-harm was 13.7% (95% CI: 11.0–17.0%) and 14.2% (95% CI: 10.1–19.5%) respectively. Subgroup analyses showed that full-time school attendance, non-Western countries, low and middle-income countries, and geographical locations might contribute to the higher aggregate prevalence of suicidal behaviors, deliberate self-harm, and non-suicidal self-injury. Conclusions: This meta-analysis found that non-suicidal self-injury, suicidal ideation, and deliberate self-harm were the three most common suicidal and self-harm behaviors in children and adolescents.


Introduction
During the past 30 years, suicide has become a severe cause of mortality across all ages in the world. In 2015, the number of suicide deaths worldwide was estimated to be 788,000 [1], with a global average of 10.7 per 100,000. Suicide was ranked second as a cause of mortality amongst those aged 15-29 years old globally [2], making it a global public health concern. Suicidal behavior, deliberate self-harm and non-suicidal self-injury are important antecedents of suicide in children and adolescents [2]. Suicidal behaviors involve suicidal ideation, planning for suicide and suicide attempts [3]. Self-harm behavior is defined here as an act of intentionally causing harm to own self, irrespective of the type, motive or suicidal intent [2]. Non-suicidal self-injury is defined as deliberate direct destruction or alteration of body tissue without conscious suicidal intent [4]. Deliberate self-harm is an encompassing term for self-injurious behavior, both with and without suicidal intent that has a non-fatal outcome [5]. Non-suicidal self-injury and deliberate self-harm are common in young people who will have borderline personality traits or disorder [6]. Non-suicidal self-injury and deliberate self-harm have been known to predict future suicide attempts [7].
There are potential factors that affect the global prevalence of suicidal and self-harm behavior in children and adolescents. From cross-cultural perspectives, there are ethnic differences in risk factors of suicide attempts [8,9]. In Western countries like Canada, suicide accounts for 10% of deaths in children aged 10 to 14 years and for 23% of deaths in adolescents aged 15 to 19 years [10]. In New Zealand, children and adolescents from the lowest socio-economic status were found to be 31 times more likely to attempt suicide compared to individuals in the higher socio-economic status [11]. In Asia, relationship issues, academic and environmental stressors are common precipitants for suicide attempts among young people [12]. In Singapore, a peak in suicide attempts has been observed in adolescents and young adults aged 15 to 24 years old [12]. Cross-cultural studies found that self-harm behaviors amongst eighth-graders in Hong Kong (23.5%) were less frequent compared to those in the United States (32%) [13]. The lower self-harm rate in Hong Kong adolescents was attributed to cultural differences between the Eastern and Western cultures, with a stronger emphasis on family structures and rules in Asian culture [13]. A meta-analysis is required to study cross-cultural perspectives of suicidal behavior, deliberate self-harm, and non-suicidal self-injury among young people in different countries in a systematic manner.
From a gender perspective, Lewinsohn et al. found female adolescents to have a significantly higher risk of suicide attempts compared to male counterparts, with the differences between genders diminishing as participants increased with age [14]. Furthermore, gender was found to predict lethality in suicide attempts as more males than females made attempts with high perceived lethality and medical lethality [15]. Youth who experienced difficulty in school were at risk for suicide [16]. However, there is little published information specific to the relationship between school attendance, suicidal behaviors, deliberate self-harm, and non-suicidal self-injury in children and adolescents.
Despite the seriousness and scope of the problem, little is known about the global prevalence of suicidal and self-harm behaviors in children and adolescents in the past 30 years. Further research is required to compare the prevalence of suicidal, and self-harm behavior among children and adolescents from different geographical regions as contextual differences (e.g., exposure to adversity) across countries may affect prevalence estimates. Given the above findings and observations, we aimed to conduct a meta-analyze to estimate the global 12-month and lifetime prevalence of adolescents having a history of suicide attempts, suicide plans, suicidal ideation, non-suicidal self-injury, and deliberate self-harm between 1989 to 2018.

Statistical Analysis
All statistical analyses were conducted in Comprehensive Meta-analysis statistical software version 3.0 (BioStat Solutions, Inc, Frederick, MD, USA). The aggregate prevalence was calculated based on the random-effects model. The random-effects model was used as it assumes varying effect sizes between studies, because of differing study design and study population [24,25]. A forest plot was then constructed and reported the aggregate prevalence, 95% confidence interval (CI) and p-value based on the method adopted by previous meta-analysis on prevalence [26,27]. The statistical significance level was set at p < 0.05. The I 2 statistic was used to assess the between-study heterogeneity [28], which describes the percentage of variance on the basis of real differences in study effects. I 2 value of 25% was considered low, 50% moderate and 75% substantial [29].
Publication bias was assessed with the utilization of Egger's regression [25]. A p-values of 0.05 or less was used as the cut off for the presence of statistically significant publication bias [30]. The presence of publication bias was then further investigated using both the standard and Orwin's fail-safe N tests to provide an estimated number of additional studies required to make the eventual effect size insignificant [31]. Meta-regression analyses with a mixed-effect model were performed to identify the effects of potential moderators on the overall heterogeneity. Potential moderators include mean age of sample and proportion of female gender. Subgroup analyses were performed to compare the aggregate prevalence of each study outcome with regards to school attendance and study location. The definitions of developing and developed countries were based on Standard Country or Area Codes for Statistical Use developed by the United Nations [32].

Selection Results and Study Characteristics
A total of 668 potentially relevant citations were gathered after an extensive literature search was performed on the databases listed in Figure 1. A total of 106 studies were found to be duplicated. Of the remaining 562 studies for which titles and abstracts were screened, 400 were excluded. The final 162 studies were then reviewed in full, of which 96 were excluded, leaving 66 studies that met the inclusion criteria to be used in this meta-analysis. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses-chart depicting the detailed process of paper selection can be seen in Figure 1. The 66 studies included in the meta-analysis yields a total population of 686,672 study participants. Table 1 shows the characteristics of the included studies.

Aggregate Prevalence of Suicide Attempts in Children and Adolescents
The aggregate lifetime prevalence of suicide attempts was found to be 6.0% (95% Confidence Interval (CI): 4.7-7.7%). The forest plot is shown in Figure 2. There was a significantly high level of heterogeneity across the included studies (I 2 = 98.60, p <0.001). There was no evidence of publication bias (intercept = 0.16, 95% CI: −5.87-6.2, t = 0.06, df = 27, p = 0.96).

Aggregate Prevalence of Suicide Plans in Children and Adolescents
The aggregate lifetime prevalence of suicide plans was found to be 9.9% (95% CI: 5.5-17.0%). The result is demonstrated using the forest plot, as shown in Figure 4. There was a significantly high level of heterogeneity across the included studies (I 2 = 99.35, p < 0.001). The aggregate 12-month prevalence of suicide plans was found to be 7.5% (95% CI: 4.5-12.1%). There was a significantly high level of heterogeneity across the included studies (I 2 = 99.82, p < 0.001). The result is demonstrated using the forest plot, as shown in Figure 4. There was no evidence of publication bias (intercept = 15.24, 95% CI: −5.06-35.54, t = 1.58, df = 17, p = 0.13).

Aggregate Prevalence of Suicide Plans in Children and Adolescents
The aggregate lifetime prevalence of suicide plans was found to be 9.9% (95% CI: 5.5-17.0%). The result is demonstrated using the forest plot, as shown in Figure 4. There was a significantly high level of heterogeneity across the included studies (I 2 = 99.35, p < 0.001). The aggregate 12-month prevalence of suicide plans was found to be 7.5% (95% CI: 4.5-12.1%). There was a significantly high level of heterogeneity across the included studies (I 2 = 99.82, p < 0.001). The result is demonstrated using the forest plot, as shown in

Aggregate Prevalence of Suicide Plans in Children and Adolescents
The aggregate lifetime prevalence of suicide plans was found to be 9.9% (95% CI: 5.5-17.0%). The result is demonstrated using the forest plot, as shown in Figure 4. There was a significantly high level of heterogeneity across the included studies (I 2 = 99.35, p < 0.001). The aggregate 12-month prevalence of suicide plans was found to be 7.5% (95% CI: 4.5-12.1%). There was a significantly high level of heterogeneity across the included studies (I 2 = 99.82, p < 0.001). The result is demonstrated using the forest plot, as shown in Figure 4. There was no evidence of publication bias (intercept = 15.24, 95% CI: −5.06-35.54, t = 1.58, df = 17, p = 0.13).

Aggregate Prevalence of Suicide Ideation in Children and Adolescents
The aggregate lifetime prevalence of suicidal ideation was found to be 18% (95% CI: 14.2-22.7%). The result is demonstrated using the forest plot, as shown in Figure 5. There was a significantly high level of heterogeneity across the included studies (I 2 = 99.68, p <0.001). There was no evidence of publication bias (intercept = −11.18, 95% CI: −21.49-0.88, t = 2.21, df = 31, p = 0.03).

Aggregate Prevalence of Suicide Ideation in Children and Adolescents
The aggregate lifetime prevalence of suicidal ideation was found to be 18% (95% CI: 14.2-22.7%). The result is demonstrated using the forest plot, as shown in Figure 5. There was a significantly high level of heterogeneity across the included studies (I 2 = 99.68, p <0.001). There was no evidence of publication bias (intercept = −11.18, 95% CI: −21.49-0.88, t = 2.21, df = 31, p = 0.03). The aggregate 12-month prevalence of suicidal ideation was found to be 14.2% (95% CI: 11.6-17.3%). The result is demonstrated using the forest plot as shown in Figure 6. There was a significant high level of heterogeneity across the included studies (I 2 = 99.82, p < 0.001). There was no evidence of publication bias (intercept = −5. 18  The aggregate 12-month prevalence of suicidal ideation was found to be 14.2% (95% CI: 11.6-17.3%). The result is demonstrated using the forest plot as shown in Figure 6. There was a significant high level of heterogeneity across the included studies (I 2 = 99.82, p < 0.001). There was no evidence of publication bias (intercept = −5.

Aggregate Prevalence of Deliberate Self-Harm in Children and Adolescents
The aggregate lifetime prevalence of deliberate self-harm was 13.7% (95% CI: 10.9-17.1%). The result is demonstrated using the forest plot, as shown in Figure 8.

Aggregate Prevalence of Deliberate Self-Harm in Children and Adolescents
The aggregate lifetime prevalence of deliberate self-harm was 13.7% (95% CI: 10.9-17.1%). The result is demonstrated using the forest plot, as shown in Figure 8.

Subgroup Analyses Based on School Attendance
A higher aggregate lifetime prevalence of suicide attempts is found amongst young people who attended school full-time as compared to young people from the mixed group of education consisting partial and non-school attendees (6.7% (95% CI: 5.3-8.4%) vs. 4.3% (95% CI: 2.7-6.7%)). The aggregate prevalence of suicide attempts in the past 12 months for full-time school attendees (5.6%, 95% CI: 4.2-7.3%) was found to be higher than partial and non-school attendees (2.1%, 95% CI: 1.3-3.6%).

Discussion
To the best of our knowledge, this is the first meta-analysis that analyzed suicidal and self-harm phenomena based on 686,672 young people worldwide. The key findings are summarized as follows. non-suicidal self-injury was most frequent with aggregate lifetime and 12-month prevalence of 22.1% and 19.5% respectively. Suicidal ideation was second most frequent with aggregate lifetime and 12-month prevalence of 18% and 14.2% respectively. Deliberate self-harm was third most frequent with aggregate lifetime and 12-montnh prevalence of 13.7% and 14.2% respectively. Suicidal plan ranked fourth with aggregate lifetime and 12-month prevalence 9.9% and 7.5% respectively. Suicide attempt was least frequent with aggregate lifetime and 12-month prevalence of 6.0% and 4.5% respectively.
This meta-analysis found that the aggregate lifetime prevalence of suicide attempts was higher in Western (6.5%, 95% CI: 4.7-9.0%) than non-Western countries. There are several reasons to explain higher prevalence of suicide attempts among young people in western countries. First, substance abuse appeared to have affected suicide rates of young males in Western countries [129]. Second, high suicide rates among young indigenous people in Western countries have been attributed to internalised anger and despair related to social disruption and disempowerment [130]. Third, young people in Western countries could have more access to suicide means, including firearms. In contrast, the aggregate lifetime prevalence of suicide plans, suicide ideation and non-suicidal self-injury were higher in non-Western countries than Western countries. This finding suggests that young people in non-Western countries could have thought about suicide but did not attempt suicide. Attempted suicide is illegal in some of the non-Western countries including Bangladesh, Hungary, India and Japan, though in Japan it is not punishable [129]. The legal implication could deter suicide attempts in some of the non-Western countries.
This meta-analysis found that non-suicidal self-injury had the highest aggregate lifetime and 12-month prevalence worldwide. Non-suicidal self-injury is defined as the intentional destruction of one's own body tissue without suicidal intent [27]. Examples of non-suicidal self-injury, including self-laceration, skin scratching, burning and hitting. Klonsky et al. proposed functional theories that explain the reasons for non-suicidal self-injury in young people [131]. The reasons include alleviation of negative emotion, self-punishment, self-directed anger, and expression of distress. Klonsky et al. highlighted the misconception that non-suicidal self-injury is always a symptom of borderline personality disorder [132]. For young people with engaging non-suicidal self-injury, a psychological intervention which aims at building positive emotion, reducing self-directed anger and promoting more adaptive way to express distress may reduce the prevalence of non-suicidal self-injury.
A recent study found that the 12-month prevalence rates of youth self-harm in low and middleincome countries were comparable to high-income countries. This meta-analysis with a larger sample size showed that children and adolescents from low-and middle-income countries with lower income had a higher aggregate 12-month prevalence of deliberate self-harm than children and adolescents from developed countries with higher income. Previous research reported that non-suicidal self-injury appeared to be more common among Caucasians than non-Caucasians [133]. Our meta-analysis found that the 12-month prevalence of non-suicidal self-injury was highest in Australia which has 74.3% of the population who are Caucasians [134].
The subgroup analysis yielded several interesting findings. The aggregate lifetime and 12-month prevalence of suicidal and self-harm behavior were higher in full-time school attendees as compared with partial and non-attendees. School attendees are more likely to be exposed to risk factors, including academic stress and school bullying. Academic stress leads to anger, anxiety, helplessness, shame, and boredom [135]. A previous study found that skin picking, which causes skin damage was positively correlated with academic stress and trait anxiety was a predisposing factor [136]. The other risk factors faced by school attendees are peer victimization, which reflects the experience of overt (e.g., hitting, pushing), reputational (e.g., spreading rumors), or relational aggression from peers (e.g., being excluded, gossiped about) [137,138]. Vergara et al. (2019) found that peer victimization was associated with the frequency of past month non-suicidal self-injury thoughts and past month non-suicidal self-injury behaviors [139]. The aggregate lifetime and 12-month prevalence of suicidal and self-harm behavior were higher in developing and non-western countries.
We found that the lifetime prevalence of suicide ideation, the 12-month prevalence of suicide attempts and suicide ideation were highest in Africa. This could be due to the fact that large numbers of African children and adolescents were exposed to adverse childhood experiences [140]. Young people in developing and non-western countries are more likely to be exposed to adverse childhood experiences including alcohol abuse [141], lack of access to care for mental health problems, orphanage, and early parental death [142], human immunodeficiency virus (HIV) infection [142] and violence against children and adolescents [143]. Interventions to reduce African children and adolescent suicidality include those improving family functioning, reducing poverty, mitigating the impacts of HIV and the provision of effective mental health services for adversity-exposed children and adolescent [140]. As the infrastructure for mental health service is still developing, the main challenge is to reach out to children and adolescents in low and middle-income countries and educate them to handle suicidal and self-harm behaviors. Electronic health (E-health) was found to provide a cost-effective solution in mental health [144]. As the use of smartphones becomes increasingly prevalent and affordable, more children and adolescents in low and middle-income countries can own a smartphone device and download health-related applications [17]. Proof of concept feasibility studies and randomized trials should be conducted in low and middle-income countries to determine that smartphone applications are efficacious to reduce suicidal and self-harm behavior in children and adolescents before their actual implementation [145].
The lifetime and 12-month prevalence of suicide plans and DSH were found to be highest in Asia. Asian children and adolescents face more academic-related stress due to the competitiveness in the education system, and getting poor grades in the examination have been found to bear two major significant sources of anxiety and depression amongst Asian children and adolescents [146]. Willingness to seek help was found to be a protective factor against suicidal and self-harm behavior for Asian children and adolescents [15]. Nevertheless, help-seeking from peers may not be beneficial [147]. Children and adolescents may not receive the help that they require, as peers often might be poorly equipped to provide appropriate advice.
Meta-regression found that age was a critical moderator that explains for heterogeneity for a lifetime and 12-month prevalence of suicide attempts and suicide plans. Children and young adolescents were less exposed to suicide risk factors as compared with older adolescents [148]. Older adolescents are predisposed to specific risk factors associated with suicidal behaviors, including baseline interpersonal problems in one's social circle [149], psychiatric disorders [148], and STD-related risk [150]. Meta-regression also found that gender was not a vital moderator that explains for heterogeneity of the prevalence of suicidal and self-harm behavior. This finding may challenge the gender paradox, which suggests significant epidemiological differences in suicidal and self-harm behaviors between adolescent females and males [151]. A previous study found that there were no gender differences in family problems and school problems which are well-known risk factors associated with suicidal and self-harm behavior in young people [152].
The strengths of this meta-analysis include an extensive search in identifying a large number of articles on suicidal and self-harm behaviour in 686,672 children and adolescents, adherence to the guidelines, the inclusion of meta-regression and subgroup analysis as well as lack of publication bias [28]. Nevertheless, this meta-analysis has several limitations. First, this meta-analysis classified suicidal and self-harm phenomena into five sub-categories and not able to study their inter-relationship. Klonsky et al. proposed that non-suicidal self-injury may be an essential risk factor for suicidal behaviour [132]. Future study is required to study the inter-relationship between suicidal behaviour, deliberate self-harm, and non-suicidal self-injury in young people. Second, we could not classify dliberate self-harm as suicidal and non-suicidal deliberate self-harm. Future research is required to understand the differences between young people who attempt suicidal and non-suicidal deliberate self-harm.

Conclusions
In conclusion, this meta-analysis found that the three most common suicidal and self-harm behaviors were non-suicidal self-injury (aggregate lifetime and 12-month prevalence of 22.1% and 19.5% respectively), suicidal ideation (aggregate lifetime and 12-month prevalence of 18% and 14.2% respectively) and deliberate self-harm (aggregate lifetime and 12-month prevalence of 13.7% and 14.2% respectively). The aggregate lifetime prevalence of suicide attempts was higher in Western than non-Western countries, in contrast, the aggregate lifetime prevalence of suicide plans, suicide ideation and non-suicidal self-injury were higher in non-Western countries than Western countries. Suicidal and self-harm behavior was higher in children and adolescents who were full-time school attendees and those who live in developing countries. Meta-regression analyses showed that the mean age of participants was a significant moderator that contributed to heterogeneity for a lifetime and 12-month prevalence of suicide attempts and suicidal plans. Psychological interventions targeting self-harm and suicidal behavior, social interventions targeting adversities in low-and middle-income countries, and electronic-health interventions to reach out to children and adolescents may reduce the global prevalence of suicidal and self-harm behavior in children and adolescents. Funding: J.W. and Z.Z. declared the receipt of the following financial support for the research, authorship and/or publication of this article. This study was supported in part by National Science Education Planning Project, China (BIA180193).