The effects of interventions preventing self-harm and suicide in children and adolescents: an overview of systematic reviews

Background: Self-harm and suicide in children and adolescents are of serious consequence and increase during the adolescent years. Consequently, there is need for interventions that prevent such behaviour. The objective of this paper: to evaluate the effects of interventions preventing self-harm and suicide in children and adolescents in an overview of systematic reviews. Methods: We conducted an overview of systematic reviews (OoO). We included reviews evaluating any preventive or therapeutic intervention. The methodological quality of the included reviews was assessed independently, and data was extracted by two reviewers. We report the review findings descriptively. The certainty of evidence was assessed using Grading of Recommendations Assessment, Development and Evaluation (GRADE). Results: Moderate certainty evidence suggests that school-based interventions prevent suicidal ideation and attempts short term, and possibly suicide attempts long term. The effects of community-based interventions following suicide clusters and local suicide plans are unknown, as are the benefits and harms of screening young people for suicide risk. The effects of most interventions targeting children and adolescents with known self-harm are unknown. However, low certainty evidence suggests that dialectical behavioural therapy and developmental group therapy are equally as effective on repetition of self-harm as enhanced treatment as usual. Conclusions: Research on several recommended practices, such as local suicide plans, prevention of suicide clusters and approaches to risk assessment, is lacking. When such interventions are implemented, the effects should be closely evaluated. There is also need for more research on treatment of repeated self-harm. Further research should include long term follow-up, and investigate possible adverse effects. In prevention of self-harm and suicide in children and adolescents, policy makers and health providers should consider evidence from population-based studies with mixed-age samples, adult samples, and studies on conditions associated with self-harm and/or suicidality, such as depression and psychosis. PROSPERO registration: CRD42019117942 08/02/19


Amendments from Version 1
We wish to thank the reviewers for valuable comments on the manuscript. The main differences compared with the previous version are: • We describe the reason for the five-year cut-off in our search for systematic reviews. This cut-off is pragmatic but similarly practiced by others, e.g. the Cochrane Library (https://community. cochrane.org/editorial-and-publishing-policy-resource/cochranereview-development/cochrane-review-updates), in considering that older reviews are in need of being updated. It takes time before a review is published, and it may be published one to three years after the search for primary studies. Thus, a review published earlier than 2012 may not include primary studies published the last >10 years. This has improved with time and new publication standards for reviews. • We have made it explicit that the search-words are included in the IN SUM Search Strategy (extended data, reference 25) and that we screened providing all the references in IN SUM. We also provide examples of search words. IN SUM is a database of systematic reviews on effects of child mental health and welfare interventions from the following databases: Cochrane Database of Systematic Reviews, Campbell Library, PsycINFO, MEDLINE, Embase, Web of Science, Database of Abstracts of Reviews of Effects (DARE) and Evidence Based Mental Health. • We have tried to make the introduction and discussion more coherent and on point about the main message, and we comment the implications of TAU versus other active intervention control groups. • We have added information on the importance of biological factors and bereavement as risk factors in the introduction and the summary. • We have added information about why including reviews in English, Norwegian, Danish or Swedish: we have included languages available to us, and guidelines developed in Sweden, Denmark and UK carry out extensive evidence reviews.

Introduction
Self-harm involves intentional self-poisoning or self-injury, irrespective of type of motive or the extent of suicidal intent 1,2 . It is often a coping mechanism used to solve a difficult situation, and can serve functions such as affect regulation, communicating the extent of pain, or self-punishment 3 . While self-harm is rare in children younger than 12 years 4 , it is prevalent amongst adolescents 5 : across international studies, 18% between the ages of 12 and 18 report a history of one or several episodes of intentional self-harm. Prevalence is highest amongst adolescent girls, typically done by cutting, but self-harm is also a problem amongst boys, more often hitting themselves 6,7 . It may be temporary or more long-lasting in nature 6 , and one episode of self-harm is a strong predictor of repetition 8,9 . When repeated, the person often advances to a combination of different methods, increasing the medical severity 10 . Completed suicide is on the other hand defined as the act of intentionally ending one's own life 11 . Suicide is rare before the age of 15 but increases in prevalence through adolescence 5 , and is somewhat most prevalent amongst males 12 . It is the most common cause of death in female adolescents, and the third most common cause of death in male adolescents (after road-traffic accidence and violence) 5 . Self-harm and suicide result from underlying riskand maitaining factors, spanning from other mental health problems such as depression, biological factors, exposure to traumatic events or other difficult circumstances in the young person's environment 4,13 . Furthermore, there are repercussions to being exposed to family and/or friends' self-harm and suicide. Such exposure may contribute to self-harm and suicide in adolescents, a phenomenon referred to as "social contagion" 4 . Related, the bereavement process of survivors after losing a significant other may last a long time and increase the risk of suicide 14 and suicidal thoughts 15 .
Evidently, self-harm and suicide in children and adolescents are complex and multifaceted phenomena. As prevention likely warrants a variation of measures, clinicians and policy makers are in need of knowledge the effects of different types of preventive interventions.
There are several reviews with summarized evidence on effects of interventions aimed at preventing (re)occurance of self-harm and suicide. However, many reviews are of variable quality, or outdated [16][17][18][19][20][21] . Furthermore, there is a large overlap of interventions covered in the different reviews, making it difficult for professionals to sort out the best available evidence needed to make informed decisions 22 . Consequently, we wanted to provide an up-to-date overview of the best quality summarized evidence on effects of all types of interventions aimed at preventing self-harm and suicide.

Objective
The objective of this review is to summarize the effects of interventions aimed at preventing self-harm and suicide in children and adolescents.

Methods
This review was registered with the international prospective register of systematic reviews (PROSPERO; CRD42019117942) on February 8 2019.

Inclusion and exclusion criteria
We included systematic reviews published in 2012 and later (last date searched August 2018), and fulfilling the DARE-criteria 23 . This five-year cut-off is pragmatic in considering that older reviews are no longer a reliable basis for updated evidence. A review published earlier than 2012 may not include primary studies published the last >10 years. Furthermore, to include the broadest possible evidence base, we included reviews in all the languages available to us: English, Norwegian, Danish or Swedish. The other inclusion criteria (PICO) are presented in Box 1.

Box 1.
Population: Children and adolescents under 18 with or without an identified risk of developing problems involving self-harm and/or suicide, or those who have already developed these problems. Intervention: Any intervention aimed at preventing or reducing self-harm and suicide, including psychological therapy, pharmaceutical interventions, psychosocial interventions, physical activity or nutrition.

Control:
Other relevant interventions, treatment as usual (TAU) or wait list. Outcome: All outcomes evaluated in children and youth, including (but not restricted to) self-harm, completed suicide, other health outcomes, quality of life, function, use of health care, attitudes and unwanted effects of interventions.
We excluded systematic reviews that did not meet the criteria for the above-mentioned PICO: • Children and adolescents with other main-diagnosis, e.g. children admitted to hospitals because of somatic illness at the same time as experiencing depressive symptoms.
• Interventions with the main objective to prevent other mental health problems, such as depression.
• Interventions preventing other behaviours with no direct association with mental health, e.g. interventions targeting smoking cessation.
• Pharmaceutical interventions compared to placebo. This review was conducted to inform decision-making in Norway, and for this purpose only direct comparisons between pharmaceutical treatments were judged to be relevant.

Literature search
The literature search for this review was completed in August 2018 and is largely based on IN SUM: a database of systematic reviews on effects of child mental health and welfare interventions 24  All publications judged to meet the inclusion criteria were retrieved in full text. Two researchers (ISM, AA) independently screened and assessed all full text reviews for potential inclusion.
In cases of disagreement, we consulted a third person.

Assessment of overlap between reviews and methodological quality
We sorted all included reviews by population and intervention comparisons (the PICOs). In cases were more than one review addressed the same comparison for the same population, we included the review with the newest search date (and completeness of this search by considering the included primary studies) and the best quality. In considering overlap, the first author (ISM) extracted this information from the reviews, and the second author (AA) double-checked the information. Further, we assessed the methodological quality of the included reviews based on a checklist for systematic reviews (AMSTAR: A MeaSurement Tool to Assess systematic Reviews) 27 . Two people (ISM, IB) considered each publication independently and decided on the quality through discussions until consensus.
The final decision on which reviews to include was done through agreement between two of the authors (ISM and AA). Table 1 contains documentation on characteristics of the included reviews, including methodological quality.
Data extraction and analyses ISM extracted data from the systematic reviews and AA checked its accuracy. As this was an overview of systematic reviews, we extracted information as it was reported in the systematic reviews, including any supplementary tables or appendixes. We did not retrieve primary studies to provide additional information about interventions or results.
From the systematic reviews, we extracted information about the primary studies' populations, characteristics of the interventions and comparison groups, duration of the interventions, follow-up periods, outcome measures and pooled effect estimates for each outcome. In cases were the effect estimates were not pooled in a meta-analysis, we reported the results of each individual study for each outcome.
We did not attempt any reanalysis, but present results as reported in the systematic reviews. For reviews including studies on both children/adolescents and adult populations,

All types of interventions
Interventions for existing self-harm: therapeutic assessment versus treatment as usual (TAU) Population: Adolescents, 12-18-year olds, referred for a psychosocial assessment following an episode of self-injury or self-poisoning, irrespective of intent Intervention: Standard psychosocial history and suicide assessment, a review of this information, identification of target problems, considering ways to change them and motivations to do so, and alternative problem-solving strategies Control: Treatment as usual comprised of standard psychosocial history and suicide risk assessment Length

Date of search
The authors' defined study population Interventions for existing self-harm: developmental group therapy versus TAU Population: Adolescents, 12 to 17-year olds, referred to child and adolescent services following an episode of intentional self-injury or self-poisoning, irrespective of intent Intervention: Manualized developmental group psychotherapy involving elements of cognitive behavioural therapy, social skills training, interpersonal psychotherapy, dialectical behavioural therapy, and group psychotherapy with or without addition to treatment as usual Control: Treatment as usual (i.e. individual counselling, family individual-based interventions such as counselling, family sessions, pharmaceutical treatment) Length of intervention: Acute treatment phase weekly sessions over 6 weeks, followed by weekly or biweekly booster sessions as long as required Follow-up period: 6 and 12 months Interventions for existing self-harm: other psychotherapeutic approaches (no primary studies identified) Interventions for existing self-harm: nutrition No primary studies identified Interventions for existing self-harm: pharmacological treatment No primary studies identified Interventions for existing self-harm: compliance enhancement versus TAU Population: Children and adolescents, 10 to 19-year olds, admitted to the emergency department of a general hospital following an episode of self-injury irrespective of intent, and/or increased risk for suicidality Intervention: a single, one-hour session that reviewed expectations for outpatient treatment as well as addressing factors likely to impede attendance and treatment misconceptions and encouraged both the adolescent and parent to make verbal contract and to attend all treatment sessions. Follow-up phone-calls 1, 2, 4 and 8 weeks after disposition. Control: TAU Length of intervention: 8 weeks Follow-up period: 3 months Interventions for existing self-harm: home-based family intervention versus TAU Population: Adolescents aged 16 or younger referred to child and adolescent mental health services following an episode of selfpoisoning irrespective of intent Intervention: manualized home-based family therapy intervention involving one assessment session and 4 home visits in addition to treatment as usual Control: Treatment as usual Length of treatment: Not stated Follow-up period: 6 months Interventions for existing self-harm: emergency cards plus TAU versus TAU Population: adolescents in the ages of 12 to 16 admitted to hospital after an episode of self-injury or self-poisoning Intervention: emergency green card in addition to usual care. The green card acted as a passport to re-admission into a paediatric ward at the local hospital Control: standard follow-up including treatment from a clinic or child psychiatry department as required Length of intervention: 12 months Follow-up period: 12 months Interventions for existing self-harm: postcards versus TAU Population: Adolescents and young adults between the ages of 15 to 24 with a history of suicidal threats, ideation, attempts and/or selfinjury who did not meet entry criteria for service because they either were not well enough or were receiving treatment elsewhere Intervention: Postcards mailed out monthly over 12 months expressing interest for that person's well-being, remining him or her about previously identified sources of help and describing one of six rotating self-help strategies (e.g. physical activity, books, Web-sites) Control: Treatment as usual Length of intervention: 12 months Follow-up period: Post-intervention  we only extracted information from studies on children and adolescents. When reported, the effect estimates were presented with relevant measures of uncertainty.

Assessing the certainty of evidence and reporting of results
We assessed our confidence in the evidence of effect for each outcome using the GRADE methodology (the Grading of Recommendations Assessment, Development and Evaluation) 28 . If the systematic review authors already had completed a GRADE assessment, we reviewed this. We describe our confidence in the effect estimates as high, moderate, low or very low for each outcome.

Results of the literature search
All 1259 references in the INSUM database was reviewed for potential relevance (see Figure 1). Additionally, we identified 12 records through hand-searches. Of the all together 1271 references, we excluded 1242 based on title or summary, mainly because they focused on other diagnosis or problem-areas than self-harm and/or suicide. Overall, 29 full texts were retrieved, 12 were excluded because they did not fulfil the inclusion criteria. Out of 18 potentially included reviews, 9 were excluded because of overlap (see Table 2 for excluded studies). Figure 1 describes the search-process and the number of articles excluded in each step. Eight systematic reviews 1,16,17,29-33 , including summary of new evidence of two of them 34,35 , were consequently included in the analysis. One review was identified after we had completed the analysis 36 and is therefore not included in the present review of systematic reviews.
Although the initial cut-off for age in our population was 18, two of the reviews included studies with young people up to 24 29,30 . These were included because the upper age limit used to define adolescence in research on self-harm and suicides varies between 18 and 25 5 .

Description of interventions
The reviews included a broad range of interventions. Most of the studies included adolescent populations in the age-range 12 to 18, with some exceptions of samples including younger children or young adults up to the age of 24. Preventive interventions were either focused on primary prevention for mixed-age population based samples (suicide awareness campaigns and other school-based prevention programs, screening for suicide risk) or secondary prevention (local approaches following Table 2. Systematic reviews excluded after full text assessment.   suicide clusters, suicide prevention in residential custodial and detention settings, interventions to support children and adolescents bereaved or affected by a suspected suicide) 17,29,30 . The reviews also included psychosocial or psychological intervention in cases of existing self-harm (defined as a history of at least one episode of self-harm) (therapeutic assessment, mentalization based therapy, dialectic behaviour therapy, cognitive behaviour therapy, developmental group therapy, compliance enhancement, home-based family intervention, emergency green cards, digital interventions for self-management of suicidal ideation and self-harm, postcards) 16, 30,31,34 .

Summary of findings
The effects of interventions are presented by type population (young people with or without an identified risk, or with existing self-harm, e.g. a history of at least one episode of self-harm) and by treatment comparison. Our assessment of certainty of evidence corresponds to GRADE-tables in Table 3-Table 16. For comparisons with many outcomes, we report the main outcomes in the present results section. See GRADE-assessments in Table 3-Table 16 for the remaining outcomes.
The review authors also searched for research on effects of the following interventions (versus treatment as usual (TAU) or alternative interventions), but studies on children and adolescents under the age of 18 were not identified. These were primary and secondary preventive interventions (reducing access to means, local suicide plans, local media reporting of suicides in newspapers, Internet or other digital channels, suicide prevention in residential custodial and detention settings) 29 and interventions targeting existing self-harm (assessment in children and adolescents at the emergency department, psychoeducation, pharmacological treatment or a combination of pharmacological treatment and psychotherapy, nutrition, other psychotherapeutic approaches such as problem-solving therapy, psychodynamic therapy, multi-systemic therapy, supportive therapy, or other psychosocial approaches such as counselling, self-management, respite care, assertive outreach) 1,31-35 .

School-based suicide prevention programs versus TAU, alternative interventions, wait list or no intervention.
The evidence includes 13 studies with <337 221 children and adolescents aged 10 to 23, as well as personnel in different local arenas working with young people 17,29 . In one of the studies, the participants (n=320 500) were habitants in a county in which county-based prevention programs were implemented. These participants included school students and personnel in schools and other local arenas. School-based prevention programs probably reduce suicidal ideation (RR 0.67, 95% KI 0.48 to 0.93, moderate certainty⊕⊕⊕⊖) and suicide attempts (RR 0.53, 95% KI 0.36 to 0.80, moderate certainty⊕⊕⊕⊖) at three to 12 months. Regarding suicide attempts, three studies conclude accordingly at six-and 12-month follow-up period. This effect possibly holds at ≥twoand 15-year follow-up (low certainty⊕⊕⊖⊖). Further, school-based interventions possibly reduce the rate of completed suicides at three-year follow-up (low certainty⊕⊕⊖⊖). Effects on help-seeking and unwanted effects are unclear since the evidence for these outcomes is of very low certainty⊕⊖⊖⊖. See Table 3.
Primary prevention: local approaches following suicide clusters versus historical control. The evidence includes three studies with children and adolescents between the ages of 10 and 24 29 . Follow-up period was up to four years. The evidence of effects of local approaches following suicide clusters is of very low certainty⊕⊖⊖⊖. See Table 4.

Secondary prevention: interventions to support children and adolescents bereaved or affected by a suspected suicide compared to TAU or historical control.
The evidence includes two studies 29 . However, the evidence of effects of interventions to support children and adolescents bereaved or affected by a suspected suicide is of very low certainty⊕⊖⊖⊖. See Table 5.
Primary prevention: screening for suicide risk versus no screening. The evidence is based on one review 30 . The review authors did not identify studies evaluating beneficial effects of screening as a preventive strategy in children or adolescents. They did however identify two studies evaluating adverse effects associated with screening for psychological distress and a history of deliberate self-harm and suicidal ideation in primary care settings. The studies comprised of 2650 adolescents between 13 and 19 years old, and the evidence is of very low certainty⊕⊖⊖⊖. See Table 6.
Interventions for existing self-harm: therapeutic assessment versus TAU. The evidence includes one study with 70 adolescents, 12 to 18-year olds referred for a psychosocial assessment following an episode of self-injury or self-poisoning, irrespective of intent 31 . Length of intervention was one hour and 40 minutes. Follow up was 12 and 24 months. The evidence of effects of therapeutic assessment is of very low certainty ⊕⊖⊖⊖. See Table 7.

Interventions for existing self-harm: mentalization based therapy (MBT-A) versus TAU.
The evidence includes one study with 80 adolescents, 12 to 17-year olds, diagnosed with depression and presenting to emergency departments or community psychiatric services following an episode of selfinjury or self-poisoning, irrespective of whether suicidal intent was present 31 . Length of treatment was 12 months, and follow-up period was also 12 months. The evidence of effects of therapeutic assessment is of very low certainty⊕⊖⊖⊖. See Table 8.

Interventions for existing self-harm: dialectical behaviour therapy (DBT-A) versus TAU or enhanced TAU.
The evidence includes two studies with 106 adolescents between the age of 12 and 19 years old with a history of multiple episodes self-harm 31,34 . Length of treatment was 19 weeks. Follow-up period was 16 weeks and six months. Based on the available evidence, DBT-A has little or no additional effect on repetition or frequency of self-harm (OR 0.72, 95% KI 0.12 to 4.40, low certainty⊕⊕⊖⊖) compared to (enhanced) treatment as usual.     However, DBT-A may have a moderate effect on reduction of suicidal ideation (SMD -0.62, 95% KI -1.07 to -0.16, low certainty⊕⊕⊖⊖). The certainty of evidence for other outcomes is very low⊕⊖⊖⊖. See Table 9.

Interventions for existing self-harm: cognitive behaviour therapy (CBT) versus non-directive psychotherapy.
The evidence contains one study with 39 adolescents between the age of 12 and 17 presenting to a paediatric general or psychiatric facility following self-injury in which an intent to die was indicated 31 . Length of treatment was six months. Follow-up period was three, six and 12 months. The certainty of evidence for effects of CBT compared to non-directive psychotherapy is very low⊕⊖⊖⊖. See Table 10.

Interventions for existing self-harm: developmental group therapy versus TAU.
The evidence contains three studies with 487 adolescents, 12 to 17-year olds, referred to child and adolescent services following an episode of intentional self-injury or self-poisoning, irrespective of intent 31 . The acute treatment phase was six weekly sessions, followed by weekly or biweekly booster sessions for as long as required. Follow-up period was between six and 12 months. Based on the available evidence, the effects of developmental group therapy compared to TAU are uncertain on the following outcomes: repetition of self-harm (six months: OR  Table 11.
Interventions for existing self-harm: compliance enhancement versus TAU. The evidence contains one study of 76 adolescents, 12 to 19-year olds, admitted to the emergency department of a general hospital following an episode of self-injury, irrespective of intent, and/or with an increased risk for suicidality 31 . Length of treatment was eight weeks. Follow-up period was three months. The evidence of effects of compliance enhancement is of very low certainty⊕⊖⊖⊖. See Table 12.

Interventions for existing self-harm: home-based family intervention versus TAU.
The evidence contains one study in a sample of adolescents aged 16 years or younger referred to child and adolescent mental health services following an episode of self-poisoning irrespective of intent 31 . The intervention was a manualized home-based family therapy intervention. Follow-up period was six months. The evidence of effects of home-based family intervention is of very low certainty⊕⊖⊖⊖. See Table 13.

Interventions for existing self-harm: emergency green cards plus TAU versus TAU.
The evidence contains one study with 105 adolescents between the ages of 12 and 16 who were admitted to hospital following an episode of self-injury or selfpoisoning 31 . The intervention was emergency green cards in addition to usual care. The green card acted as a passport to re-admission into a paediatric ward at the local hospital. Length of treatment was 12 months. Follow-up period was 12 months. The evidence of effects of emergency green cards is of very low certainty⊕⊖⊖⊖. See Table 14.

Interventions for existing self-harm: digital interventions for self-management of suicidal ideation and self-harm versus psychoeducation or historical control.
The evidence contains three studies with 184 adolescents reporting suicidal thoughts and/or receiving treatment for depression 16 . The interventions spanned from two to 12 weeks and follow-up was post treatment. The evidence of effects of digital interventions for self-management is of very low certainty⊕⊖⊖⊖. See Table 15.

Interventions for existing self-harm: postcards versus TAU.
The evidence is based on two systematic reviews 30,34 . One of the reviews 34 included one study with 2300 adolescents and young adults over the age of 12 previously admitted to a specialist poisons hospital after self-poisoning. The other review 30 included one study of 165 adolescents and young adults of 15 to 24 years old with a history of suicidal threats, ideation, attempts and/or self-injury who did not meet entry criteria for service because they either were not unwell enough or were receiving treatment elsewhere. Follow-up was post study. The evidence of effects of postcards is of very low certainty⊕⊖⊖⊖. See Table 16.

Discussion
The present paper gives a comprehensive overview of effects of interventions aimed at preventing self-harm and suicide in children and adolescents. We found evidence to suggest that school-based interventions probably prevent suicidal ideation and suicide attempts short term, and possibly suicide attempts long term. The effects of community-based interventions following suicide clusters and local suicide plans are unknown, as are

Adverse effects
Not reported 1. Downgraded by 1 level due to risk of bias (lack of blinding).

Downgraded by 1 level due to imprecision (few incidences).
the benefits and harms of screening young people for suicide risk. The effects of most interventions targeting children and adolescents with known self-harm are also unknown. However, low certainty evidence suggests that dialectical behavioural therapy and developmental group therapy are equally as effective on repetition of self-harm as enhanced treatment as usual. In general, the populations are adolescents in the age-range of 12 to 18 years.

Effects of preventive interventions: summary of findings and implications
Based on the available research, school-based interventions can prevent suicidal ideation and suicide attempts short term (moderate certainty evidence), and possibly suicide attempts long term (low certainty evidence), which should have obvious implications for policy makers.
As regards other preventive strategies, there is a general a lack of research on effects of recommended practices, such as approaches to risk assessment and local suicide plans. Screening for suicide risk as primary prevention may provide the opportunity of early detection, and if precise, offer the opportunity to provide young people at risk with appropriate treatment. However, it is resource demanding, and based on available research, effects of screening children and young people for symptoms of depression and a history of self-harm or suicidal ideation in the general population are unknown, given very low certainty evidence. Local suicide plans are a recommended strategy in some countries 29,37 . However, the effects of such plans on preventing self-harm and suicide in children and young people is yet to be evaluated in research. Therefore, when implemented, approaches to risk assessment and screening programs, as well as local suicide plans, should be closely evaluated.
We identified no reviews evaluating the effects of reducing access to means from children and young people specifically, or on how media reporting of suicides affects suicide rates in children and young people. In these instances, studies on interventions targeting the general population could be informative. Such studies suggest that reducing access to means may be an effective strategy 29 , and that certain forms of media reporting are associated with an increase in suicides 29 . Guidelines on how to report on suicides is one suggested strategy to address the possible harms of such reporting 29 .
Suicide clusters, although rare, is a phenomenon of major concern. When faced with potential social contagion following suicide, communities are expected to act to prevent contaigon and clustering. However, based on a few studies, the certainty of evidence for community-based interventions following suicide clusters is very low, as is the evidence on effects of support-interventions in young people bereaved or affected by a suicide in their family or other network. Even so, some recommendations are agreed upon, e.g. provision of information to relevant agencies in the community and providing support for those directly affected or other vulnerable individuals 38 . However, given that the above-mentioned research is of very low certainty, we suggest that researchers design appropriate observational studies, allowing for enough observations pre-and post-implementation of preventive measures to inform policy.
The reviews we identified also searched for studies targeting young people in residential custodial and detention settings, but no studies were identified. Therefore, effects of interventions in this high-risk population are uncertain.

Effects of interventions for existing self-harm: summary of findings and implications
Self-harm is a common reason for referral of adolescents in child and adolescent psychiatric services, and often accompanies other psychiatric symptoms presented in such settings. However, based on the available evidence, only two treatment comparisons evaluating psychological therapy provided evidence of their effectiveness (low certainty); dialectical behavioural therapy and developmental group therapy. Both treatments were compared to enhanced TAU (e.g. individual  and family sessions, medication management, and hospital or respite care as required), and there was little or no important difference in effect on repetition of self-harm, nor on symptoms of depression. However, of notice, although not statistically significant, there was a substantial higher degree of repetition of self-harm amongst adolescents participating in group developmental therapy compared to those receiving enhanced TAU at six-month follow-up. At 12-month follow-up, there was little or no important effect on self-harm. Clinicians should be aware of this potential short-term adverse effect, but this should be investigated in future studies. However, the findings on beneficial effects are overall promising. It seems that both  dialectical behavioural therapy and developmental group therapy, or established treatment approaches, are good treatment alternatives.
For remaining interventions targeting self-harm, effects are unknown. It is uncertain which approach to risk assessment of young people after an episode of self-harm is most appropriate given low certainty evidence. Furthermore, the effects of psychoeducation, psychological therapy, psychosocial interventions, digital interventions for self-management and nutrition for treating young people with existing self-harm are unknown, as no studies were identified.
The reviews we included searched for, but did not identify, studies on direct comparisons between different pharmacological treatment alternatives or on the effects of combination therapy (pharmacological treatment plus psychotherapy). The finding that biological factors may be associated with, or even predict, a suicide attempt 13 could have implications for research on pharmacological agents.
The evidence of effects of organization of services, such as home-based treatment and use of emergency green cards, is of very low certainty. New research in this area is pertinent, especially for policy makers.

Limitations
A limitation of overviews of reviews, and consequently of the present paper, is that the analyses are based on secondary reporting of what the review authors interpreted and reported based on the primary studies. It follows that the primary studies may have provided more information than what is reported in the reviews we included. A primary study investigating e.g. treatment attendance would be relevant to a clinician wanting to meet with a client struggling with suicidality regularily in order to build a working alliance. However, if the review authors did not find such an outcome from a primary study relevant, we will have missed this information. Regardless of this limitation, the reader of our overview of reviews could find a particular primary study referenced in the included review, if there is need to check if the primary study investigated other relevant outcomes.
It is also worth noting that the present paper only included reviews of studies where the intervention was to prevent or treat self-harm and suicide in children and adolescents, with exception of a few population-based studies. Self-harm and suicide are associated with other difficulties such as psychosis, depression and anxiety. Therefore, evidence from studies on children and adolescents at risk for or diagnosed with such conditions may provide important direction in decision-making when faces with self-harm and suicide. However, in studies on these conditions, self-harm and suicide are rarely investigated as outcomes [39][40][41] . An exception is research on depression, with low certainty evidence indicating that combination treatment for depression (pharmacological treatment plus psychotherapy) may lead to a reduced risk for suicide 40 .

Conclusions
Overall, evidence of moderate to low certainty suggests that school-based suicide prevention programs can prevent suicide and suicide attempts in young people. The effects of communitybased interventions following suicide clusters and local suicide plans are uncertain. Furthermore, it is not possible to make any conclusions about the benefits or harms of screening in young people with or without known risk of self-harm and suicide.
When it comes to treatment strategies for young people with existing self-harm, evidence of low certainty suggests that dialectical behavioural therapy and developmental group therapy are equally as effective on repetition of self-harm as enhanced treatment as usual (often individual and/or family psychotherapy). The effects of other interventions specifically targeting self-harm are unknown, because of lack of research or evidence of very low certainty, and should be evaluated. These interventions include mentalization-based psychotherapy, cognitive behavioural therapy and psychodynamic therapy.
Collectively, due to a general lack of research, and in some cases very low certainty evidence, the effects of most interventions are unknown. This has several implications. First and foremost, more research is needed, including studies on children younger than 12 years og age, as well as long-term follow up. Second, when implementing recommended practice with unknown effects, such as approaches to risk assessment, practice should be closely evaluated. With all types of interventions, there is a possibility for adverse effects. Hence, it is crucial to be mindful that our own preventive actions or treatment efforts could contribute to an increased risk for self-harm and suicide, and both adverse as well as beneficial effects should be evaluated. Third, policy makers and health providers should consider evidence from other relevant populations in decision-making, such as studies on adults, as well as studies on conditions associated with self-harm and/or suicidality, e.g. depression and psychosis.
A final implication worth mentioning is related to the scope of the present review of systematic reviews: effects of interventions. In decision-making, knowledge on effects of interventions should be supplemented with other relevant research, such as therapeutic processes influencing the outcome, as well as integrated with clinical expertise and the child's or adolescent's and caregiver's values and preferences 42,43 .

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required. Thank you very much for asking me to review the present manuscript. This is, in summary, an interesting paper aimed to evaluate the effects of interventions preventing self-harm and suicide in children and adolescents in an overview of systematic reviews. The authors reported that (moderate certainty evidence) school-based interventions prevent suicidal ideation and attempts short term, and possibly with long term effects on suicide attempts. Overall, the effects of community-based interventions following suicide clusters and local suicide plans resulted uncertain, as are the benefits and harms of screening young people for suicide risk. In addition, the effects of most interventions targeting children and adolescents with known self-harm were uncertain. They added that (low certainty evidence) dialectical behavioural therapy and developmental group therapy are equally as effective on repetition of self-harm as enhanced treatment as usual.

Extended data
The authors may find as follows my main comments/suggestions: First, as the authors, throughout the Introduction section, correctly stated that self-harm and suicide are associated with relevant psychosocial impairment and result from underlying factors such as other mental health problems, exposure to traumatic events or other difficult circumstances in the young person's environment, they might even mention that the emotional turmoil in the case of suicide survivors of patients died by suicide may last a long time, and in some cases, may end with their own suicide. Thus, together with self-harm and suicide, it is fundamental to understand the bereavement process after the suicide of a significant other to provide a proper care, reduce stigma, and improve the outcomes. In addition, specific biological factors such as prolactin and thyroid hormone levels may be dysregulated and significantly associated with self-harm and suicide attempts and even involved in a complex compensatory mechanism to correct reduced central serotonin activity. The assumption that prolactin and thyroid hormones may be associated or even predict a suicide attempt is of great importance given the availability of such data in everyday clinical practice. Physicians of any kind as well as mental health professionals should be aware of the importance to insert as much information possible in the assessment of suicide and self-harm risk. Thus, given the above mentioned information, the authors could include throughout the manuscript, some published papers regarding the mentioned topics (PMID: 24082246; 31091772; 28843902; 22748186; 12866334) .
In addition, why the authors decided to include all publications in English, Norwegian, Danish or Swedish rather than simply including only studies in English language is a matter of debate and needs to be specified.
Moreover, the authors should immediately present and discuss, in the first lines of the Discussion section, their most relevant study findings. Conversely, they seem to focus with redundancy on the main aims/objectives of the paper which have been already presented elsewhere.
Although the authors reported that the present analyses are based on secondary reporting and the interpretation of the review authors as well as that the present report included only reviews of studies where the population was children and young people with existing self-harm, the most relevant limitations/shortcomings of the present study need to be more carefully described for the general readership.
Finally, what is the take-home message of this manuscript? While the authors stated that practice should be evaluated, and researchers should investigate harmful effects as well as beneficial effects of interventions, they failed, in my opinion, to provide some conclusive remarks about their findings. Here, some further details/information are needed. 12345 turmoil in the case of suicide survivors of patients died by suicide may last a long time, and in some cases, may end with their own suicide. Thus, together with self-harm and suicide, it is fundamental to understand the bereavement process after the suicide of a significant other to provide a proper care, reduce stigma, and improve the outcomes. In addition, specific biological factors such as prolactin and thyroid hormone levels may be dysregulated and significantly associated with self-harm and suicide attempts and even involved in a complex compensatory mechanism to correct reduced central serotonin activity. The assumption that prolactin and thyroid hormones may be associated or even predict a suicide attempt is of great importance given the availability of such data in everyday clinical practice. Physicians of any kind as well as mental health professionals should be aware of the importance to insert as much information possible in the assessment of suicide and self-harm risk. Thus, given the above mentioned information, the authors could include throughout the manuscript, some published papers regarding the mentioned topics (PMID: 24082246; 31091772; 28843902; 22748186; 12866334).
Thank you for interesting suggestions. We have added information and some of the recommended citations in the introduction regarding the importance of bereavement as a risk factor in prevention of suicide, as well as in the section on "Effects of preventive interventions: summary of findings and implications", and about biological factors as possible risk factors in the introduction, as well as in "Effects of interventions for existing self-harm: summary of findings and implications".
In addition, why the authors decided to include all publications in English, Norwegian, Danish or 2) Swedish rather than simply including only studies in English language is a matter of debate and needs to be specified.
We thank reviewer 2 for pointing this out, and have added information about why choosing English, Norwegian, Danish or Swedish rather than simply including only studies in English language. For pragmatic reasons, we have included languages available to us. Furthermore, guidelines developed in Sweden, Denmark and UK carry out extensive evidence reviews. Neglecting to include these would weaken the evidence base.
Moreover, the authors should immediately present and discuss, in the first lines of the 3) Discussion section, their most relevant study findings. Conversely, they seem to focus with redundancy on the main aims/objectives of the paper which have been already presented elsewhere.
We agree, and are now more focused on our main findings in the beginning of the discussion.
Although the authors reported that the present analyses are based on secondary reporting and 4) the interpretation of the review authors as well as that the present report included only reviews of studies where the population was children and young people with existing self-harm, the most relevant limitations/shortcomings of the present study need to be more carefully described for the general readership.
We Finally, what is the take-home message of this manuscript? While the authors stated that 5) practice should be evaluated, and researchers should investigate harmful effects as well as beneficial effects of interventions, they failed, in my opinion, to provide some conclusive remarks about their findings. Here, some further details/information are needed.
Thank you, and we agree. We now have the take-home message as well as some further Author Response 10 Feb 2020 , Regional Centre for Child and Adolescent Mental Health, Eastern and Ida Sund Morken Southern Norway (RBUP), Oslo, Norway The methodology provided is sufficient except it is unclear why only systematic reviews 1) published in 2012 or later are included.
· We thank reviewer 1 for pointing out that this was unclear. This cut-off is pragmatic but similar to that practiced by others, for example, the Cochrane Library (https://community.cochrane.org/editorial-and-publishing-policy-resource/cochrane-review-development/cochrane-r in considering that an older review is obsolete and no longer a reliable basis for evidence and in need for being updated. It takes time before a review is published, consequently a review may be published one to three years later than the search was done for primary studies. This has improved with time and new publication standards for reviews. Thus, a review published earlier than 2012 may not include primary studies published the last >10 years. A sentence to explain this has now been added to the manuscript. It is also unclear what search terms were used for future replications. The authors report a range of outcome measures in the results that may not be directly related 3) to self-harm. Such outcomes (e.g. treatment engagement) were not further commented on or summarised.
· Thank you for pointing this out, as research on other outcomes are often highly relevant. For this review we included all outcomes as reported by the review authors. Effect estimates and judgements of certainty for each such outcome is reported for all pooled estimates. However, in line with the GRADE recommendations, we only make conclusions on outcomes judged to be of low, moderate or high certainty. When evidence is of very low certainty, the effects of these outcomes are considered to be too uncertain as to make any conclusions.
Overall, the systematic review is very comprehensive but the discussion and summary could be 4) more coherent and with a better flow. For instance, the authors can attempt to comment on the implications of TAU control group versus other active intervention control groups.
· Thank you for this feedback. We have now tried to make the discussion and summary more coherent, see introduction, discussion and summary. We have also commented on the implications of TAU control group versus other active intervention control groups, see "Effects of implications of TAU control group versus other active intervention control groups, see "Effects of interventions for existing self-harm: summary of findings and implications".
No competing interests were disclosed.

Competing Interests:
Comments on this article Version 1