Key PointsQuestion
What is the effectiveness of anti-bullying interventions, their population impact, and is there an association between moderator variables and the effectiveness of these interventions?
Findings
Across 77 samples from 69 randomized clinical trials (111 659 participants), meta-analyses showed that interventions were statistically significantly effective in reducing bullying and improving mental health problems at study end point. Meta-regression analyses showed that duration of intervention was not statistically significantly associated with effectiveness and that the impact of the anti-bullying programs did not diminish over time during follow-up.
Meaning
Findings of this meta-analysis support the concept that school anti-bullying interventions may have a valuable population impact.
Importance
Bullying is a prevalent and modifiable risk factor for mental health disorders. Although previous studies have supported the effectiveness of anti-bullying programs; their population impact and the association of specific moderators with outcomes are still unclear.
Objective
To assess the effectiveness of school anti-bullying interventions, their population impact, and the association between moderator variables and outcomes.
Data Sources
A search of Ovid MEDLINE, ERIC, and PsycInfo databases was conducted using 3 sets of search terms to identify randomized clinical trials (RCTs) assessing anti-bullying interventions published from database inception through February 2020. A manual search of reference lists of articles included in previous systematic reviews and meta-analyses was also performed.
Study Selection
The initial literature search yielded 34 798 studies. Included in the study were articles that (1) assessed bullying at school; (2) assessed the effectiveness of an anti-bullying program; (3) had an RCT design; (4) reported results; and (5) were published in English. Of 16 707 studies identified, 371 met the criteria for review of full-text articles; 77 RCTs were identified that reported data allowing calculation of effect sizes (ESs). Of these, 69 independent trials were included in the final meta-analysis database.
Data Extraction and Synthesis
Random-effects and meta-regression models were used to derive Cohen d values with pooled 95% CIs as estimates of ES and to test associations between moderator variables and ES estimates. Population impact number (PIN), defined as the number of children in the total population for whom 1 event may be prevented by an intervention, was used as an estimate of the population impact of universal interventions targeting all students, regardless of individual risk.
Main Outcomes and Measures
The main outcomes are the effectiveness (measured by ES) and the population impact (measured by the PIN) of anti-bullying interventions on the following 8 variable categories: overall bullying, bullying perpetration, bullying exposure, cyberbullying, attitudes that discourage bullying, attitudes that encourage bullying, mental health problems (eg, anxiety and depression), and school climate as well as the assessment of potential assocations between trial or intervention characteristics and outcomes.
Results
This study included 77 samples from 69 RCTs (111 659 participants [56 511 in the intervention group and 55 148 in the control group]). The weighted mean (range) age of participants in the intervention group was 11.1 (4-17) years and 10.8 (4-17) years in the control group. The weighted mean (range) proportion of female participants in the intervention group was 49.9% (0%-100%) and 50.5% (0%-100%) in the control group. Anti-bullying interventions were efficacious in reducing bullying (ES, −0.150; 95% CI, −0.191 to −0.109) and improving mental health problems (ES, −0.205; 95% CI, −0.277 to −0.133) at study end point, with PINs for universal interventions that target the total student population of 147 (95% CI, 113-213) and 107 (95% CI, 73-173), respectively. Duration of intervention was not statistically significantly associated with intervention effectiveness (mean [range] duration of interventions, 29.4 [1 to 144] weeks). The effectiveness of anti-bullying programs did not diminish over time during follow-up (mean [range] follow-up, 30.9 [2-104] weeks).
Conclusions and Relevance
Despite the small ESs and some regional differences in effectiveness, the population impact of school anti-bullying interventions appeared to be substantial. Better designed trials that assess optimal intervention timing and duration are warranted.
A preventive approach to mental health and well-being may be essential to reduce the burden associated with mental health disorders, especially in young people.1 Traditional bullying is defined as deliberate aggressive behaviors by another youth or group of youths who are not siblings or current dating partners that are repeated and involve a power imbalance favoring the perpetrators.2-4 Bullying is a major target for universal prevention given the high prevalence rates, association with increased lifetime prevalence of mental health disorders,5 and converging evidence supporting the feasibility and cost-effectiveness of anti-bullying interventions.6,7 Based on a 2009-2010 survey, the Health Behaviour in School-aged Children study reported a bullying exposure prevalence of more than 10% among a school aged population,8 but other studies9,10 have reported school bullying rates of 20% to 30% or even higher. There is greater uncertainty regarding estimated prevalence rates for cyberbullying, defined as intentional and repeated harm inflicted through electronic devices and social media. Available studies suggest that cyberbullying may affect approximately 15% to 25% of youth and that it usually coexists with traditional bullying.11
Bullying exposure has been consistently associated with worse mental health in childhood and adolescence.9,12 The negative consequences of bullying are pervasive, and bullying exposure in childhood is also associated with poor mental and physical health, lack of social relationships, economic hardship, and decreased quality of life in early adulthood and midlife.5,9,13-20 Bullying perpetrators also experience worse physical and mental health both in childhood and adulthood along with social disadvantage during adulthood.21-23
Taken together, the available evidence indicates that bullying is 1 of the most prevalent potentially modifiable risk factors for mental health disorders, thus rendering it a major public health concern,4,24 especially considering the high associated lifetime direct and indirect economic costs.7,25 The growing awareness of bullying has led to the implementation of different school-based anti-bullying programs in the last 20 years.26 Some meta-analyses27-31 have reported small to moderate effectiveness of anti-bullying programs, with a mean decrease of approximately 20% in bullying rates. The results of these meta-analyses support the feasibility of implementing anti-bullying programs in schools and suggest their potential effectiveness. However, the population impact (ie, the number of children in the whole population among whom 1 event of bullying on average may be prevented by anti-bullying interventions) remains unclear.32 Previous meta-analyses27-31 also leave several questions unanswered. What is the association of bullying prevention interventions with mental health? What is the optimal duration of an intervention program? Does the benefit of the intervention diminish over time after the intervention ends? Do differential factors moderate the effectiveness of anti-bullying and anti-cyberbullying programs? To address these questions, we conducted a meta-analysis of randomized clinical trials (RCTs) that assessed school interventions with the aim of reducing bullying or cyberbullying rates or improving school climate to evaluate their short-term and medium-term population impact. In addition, we conducted meta-regression analyses to assess whether moderator variables impacted the effectiveness of the interventions.
This meta-analysis of RCTs used the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline. We conducted a systematic 2-step literature search to identify RCTs assessing anti-bullying interventions (eTable 1 in the Supplement).33 We first performed a search of Ovid MEDLINE, ERIC (Eric.ed.gov), and PsycInfo databases from inception through February 2020 (eMethods in the Supplement). Three sets of search terms were used: (1) [“bullying” OR “peer abuse” OR “abuse” OR “aggression” OR “harassment” OR “perpetrator” OR “victim” OR “victimization” OR “peer violence” OR “violence” OR “cyberbullying” OR “anti-bullying”], (2) AND [“school” OR “peer”], and (3) AND [“intervention” OR “curriculum” OR “prevention” OR “program” OR “resilience” OR “school climate” OR “school-based” OR “therapy” OR “treatment” OR “trial”]. We then performed a manual search of the reference lists of articles included in previous systematic reviews and meta-analyses for any RCTs not identified by the literature search.
Study Selection and Data Extraction
eFigure 1 in the Supplement shows a flowchart of the systematic literature search strategy. The initial literature search yielded 34 798 studies. The manual search identified 6 additional records. After removing 18 097 duplicates, we evaluated 16 707 potential studies.
Four of us (M.A., M.D-C., R.A-C., and I.E-B.) double-screened all articles in 3 phases, resolving discrepancies through discussion and consensus. The eMethods in the Supplement describes study selection criteria and procedures in detail. Briefly, in phase 1, inclusion criteria included articles that (1) assessed bullying at school; (2) assessed the effectiveness of an anti-bullying program; (3) had an RCT design; (4) reported results; and (5) were published in English. Of 16 707 studies identified, 371 met the criteria to proceed to phase 2. Phase 2 consisted of a review of the full-text articles; 77 RCTs were identified that reported data that would allow calculation of effect sizes. Of these trials, 69 original independent RCTs met the criteria for inclusion in the final meta-analysis database.
Six of us (D.F., C.M.D-C., M.A., M.D-C., R.A-C., and I.E-B.) extracted data from each eligible study independently and double-checked them by pairs, with discrepancies resolved via discussion. Data extracted included the following: year of publication, region (country and city if available) where the study was conducted, name of the intervention program, date of intervention, duration of intervention, duration of follow-up (when applicable), type of randomization (individual or cluster), type of control group, type of school (public or private), primary (age, ≤11 years) vs secondary (age, 12-18 years) education, sample size, number of randomized groups, mean age, age range, and percentage of females (for both intervention and control groups), type of approach (universal or targeted), type of bullying variable (dichotomous or continuous), and statistics to calculate effect sizes for the meta-analyses and meta-regressions (eMethods in the Supplement).
Classification of Outcome Variables and Quality Assessment
The 69 original independent RCTs used more than 500 different instruments to assess outcome variables. Three of us (D.F., R.A-C., and I.E-B.) independently classified these instruments into a manageable number of outcome variables, with discrepancies resolved by discussion. This classification allowed us to consolidate outcome variables into the following 8 categories based on previous meta-analyses28,29: (1) overall bullying (pooled measure, including data on bullying perpetration, bullying exposure, and cyberbullying), (2) bullying perpetration, (3) bullying exposure, (4) cyberbullying (both perpetration and exposure), (5) attitudes that discourage bullying, (6) attitudes that encourage bullying, (7) mental health problems (eg, anxiety and depression), and (8) school climate.
We performed meta-analyses for each of the categories at end point of intervention and follow-up (interval between end of intervention and further assessment). Details regarding classification of outcome variables are described in the eMethods and eTable 2 in the Supplement.
We assessed the quality of the 69 selected RCTs using an item checklist constructed for this meta-analysis based on the Cochrane Collaboration’s tool for assessing risk of bias.34 Details on the quality assessment are described in eTable 3 in the Supplement.
We conducted random-effects meta-analyses using Comprehensive Meta-Analysis, version 2.0 (Biostat Inc).35 Cohen d values with pooled 95% CIs were used as estimates of the effect size of each anti-bullying intervention compared with control groups. For purposes of this work, a positive Cohen d value indicates that a specific variable increases more in the intervention group than in the control group during the assessed period, whereas a negative Cohen d value indicates that a specific variable increases more in the control group than in the intervention group during the assessed period. Forest plots were generated using DistillerSR Forest Plot Generator (Evidence Partners).36
We assessed statistical heterogeneity through visual inspection of forest plots and using the Q statistic (a magnitude of heterogeneity) and the I2 statistic (a measure of the proportion of variance in summary effect sizes attributable to heterogeneity).37 I2 values less than 30% were considered an insignificant amount of heterogeneity.38 We assessed publication bias by visually inspecting funnel plots and using the fail-safe N described by Orwin,39 with a criterion for a trivial standardized difference in means of 0.1 and a mean standardized difference in means in missing studies of 0. Furthermore, we used the linear regression method described by Egger et al40 to quantify the bias captured by the funnel plot.
We used meta-regressions with a random-effects model with unrestricted maximum likelihood to test associations of potential moderators with effect size estimates for statistically significant meta-analyses. Statistically significant meta-regression values were confirmed by excluding 1 study at a time, and only meta-regressions for which P values remained statistically significant after this process were considered statistically significant. The threshold for statistical significance was set at .05.
Because recent meta-analyses of the effectiveness of anti-bullying interventions have reported a statistically significant association with geographic location,27 we performed a meta-analytic subgroup analysis by region. We conducted additional subgroup meta-analyses of universal interventions (targeting the total student population, regardless of individual risk) and targeted (nonuniversal) interventions.
Cohen d values were converted to number needed to treat (NNT) as recommended in the method by Furukawa and Leucht.41 The NNT was used to obtain the population impact number (PIN) as an estimated measure of the population impact of the intervention. The PIN is defined as children in the total population for whom 1 event may be prevented by an intervention32,42 or, the number needed to participate in an anti-bullying program to prevent 1 case of bullying.43 The PIN values were calculated using RCTs that assessed a universal intervention.
We used false discovery rate correction for multiple comparisons.44 The percentage of tolerated false-positive results was 5% (Q <.05). The Q value is the adjusted P value calculated using a false discovery rate approach. The threshold for statistical significance was set at .05 (2 sided). Details on statistical analyses are described in the eMethods in the Supplement.
Characteristics of the Selected RCTs and Samples
This meta-analysis included 77 samples from 69 RCTs, constituting an overall sample of 111 659 participants. These participants included 56 511 (in 609 randomized clusters) in the intervention group and 55 148 (in 601 randomized clusters) in the control group. Among all 69 RCTs, 5 tested interventions targeting cyberbullying, and 15 reported results at follow-up, with a mean (range) follow-up of 30.9 (2-104) weeks.
The weighted mean (range) age of participants in the intervention group was 11.1 (4-17) years and 10.8 (4-17) years in the control group. The weighted mean (range) proportion of female participants was 49.9% (0%-100%) in the intervention group and 50.5% (0%-100%) in the control group. The mean (range) duration of interventions was 29.4 (1-144) weeks (95% CI, 21.5-37.3 weeks). Characteristics of the selected RCTs are listed in eTable 4 and eTable 5 in the Supplement.
Effectiveness of Anti-Bullying Interventions at Study End Point
Table 1 summarizes the results. Anti-bullying interventions were effective in reducing overall bullying (as a pooled measure, including bullying perpetration, bullying exposure, and cyberbullying) at study end point (number of samples [k] = 45; effect size, −0.150; 95% CI, −0.191 to −0.109; P < .001) (Figure 1).45-86 Anti-bullying interventions showed statistically significant effectiveness compared with control groups on all assessed bullying-related outcomes after the intervention. The effect sizes were mostly statistically significant and small (mean range, 0.070-0.205), with high statistical heterogeneity and risk of publication bias. Anti-bullying interventions also showed statistically significant effectiveness in improving mental health problems (eg, anxiety and depression) at study end point, with small effect size (k = 20; effect size, −0.205; 95% CI, −0.277 to −0.133; P < .001) (Figure 2).47,51,52,55,58,59,61,64,66,73,77,86-92
Among all 69 RCTs, 31 (45.0%) were conducted in Europe, 19 (27.5%) in North America (United States and Canada), and 19 (27.5%) elsewhere. Meta-analyses by region showed that anti-bullying interventions had comparable effect sizes in overall bullying and bullying perpetration in Europe and North America. However, interventions were statistically significantly effective in decreasing bullying exposure and in decreasing attitudes that encourage bullying at study end point in Europe (k = 18; effect size, −0.142; 95% CI, −0.193 to −0.091; P < .001 and k = 10; effect size, −0.155; 95% CI, −0.242 to −0.068; P < .001, respectively) but not in North America (k = 8; effect size, −0.209; 95% CI, −0.563 to 0.145; P = .28 and k = 4; effect size, −0.016; 95% CI, −0.122 to 0.090; P = .78; respectively). Greater effect sizes were found for the effectiveness of interventions in increasing attitudes that discourage bullying at study end point in European trials (k = 15; effect size, 0.243; 95% CI, 0.164-0.323) than in North American trials (k = 6; effect size, 0.110; 95% CI, 0.063-0.157). The other meta-analyses found no statistically significant differences between regions. eTable 6 and eTable 7 in the Supplement provide additional details.
Of the 69 RCTs, 58 (84.1%) assessed a universal anti-bullying intervention. Subgroup meta-analyses of trials testing a targeted intervention showed results comparable to those found for universal interventions in terms of the direction of the associations, albeit with smaller effect sizes for most outcome measures (eTable 8 in the Supplement).
Table 2 lists NNT and PIN values for universal school anti-bullying interventions based on different estimated prevalence rates of school bullying. The PIN values of anti-bullying interventions for overall bullying and mental health problems at study end point were 147 (95% CI, 113-213) and 107 (95% CI, 73-173), respectively, assuming a bullying prevalence of 15%. Figure 3 shows PIN values for each outcome variable. For example, PIN values of anti-bullying interventions at study end point were 207 (95% CI, 153-307) for bullying perpetration, 140 (95% CI, 93-260) for bullying exposure, and 167 (95% CI, 100-360) for cyberbullying, assuming a bullying prevalence of 15%.
Effectiveness of Anti-Bullying Interventions Over Time
Anti-bullying interventions were effective in reducing overall bullying during a mean 44.0-week follow-up (k = 21; effect size, −0.171; 95% CI, −0.243 to −0.099; P < .001). Details are shown in eFigure 2 in the Supplement.
At a mean 30.9-week follow-up, the effectiveness of anti-bullying interventions remained statistically significant for all bullying-related outcomes except for cyberbullying, for which only 1 RCT was available. The effect sizes were small (range, 0.122-0.202) (Table 1). The effectiveness of anti-bullying interventions was also statistically significant in decreasing mental health problems at a mean 27.3-week follow-up, with a small effect size (k = 6; effect size, −0.202; 95% CI, −0.347 to −0.056; P = .01) (eFigure 3 in the Supplement).
The population impact of universal anti-bullying interventions for overall bullying (mean, 44.0-week follow-up) and mental health problems (mean, 27.3-week follow-up) was 127 (95% CI, 87-247) and 107 (95% CI, 53-787), respectively, assuming a bullying prevalence of 15%. Table 2 and Figure 3 summarize details on NNT and PIN values for universal anti-bullying interventions at follow-up.
Meta-regression analyses found that (1) duration of anti-bullying programs was not statistically significantly associated with effectiveness, (2) there was no statistically significant association between the length of follow-up and the effectiveness of anti-bullying programs at follow-up (mean [range] follow-up, 30.9 [2-104] weeks), and (3) no additional factor moderated the effectiveness of anti-bullying or anti-cyberbullying programs. The results of the meta-regression analyses are summarized in eTable 9 in the Supplement.
Bullying is a major public health problem worldwide.4,24 This meta-analysis shows that school anti-bullying interventions are statistically significantly effective not only in reducing bullying rates but also in improving mental health problems in young people. Despite the small effect sizes and some regional differences in effectiveness, the findings suggest that universal anti-bullying interventions have a substantial population impact.
Universally delivered psychosocial interventions in adolescence have proven to be effective in improving mental health and reducing risk behavior, including bullying.93 The results of the present study add to previous meta-analyses27-30,93 on this topic, which found that anti-bullying interventions had statistically significant effectiveness by specifically assessing the population impact of the interventions. For an estimated bullying prevalence of 15% (a conservative estimate considering prevalence rates reported in previous studies10,94), an average anti-bullying intervention needs to include 207 (95% CI, 153-307) people to prevent 1 case of bullying perpetration or 140 (95% CI, 93-260) people to prevent 1 case of bullying exposure and 107 (95% CI, 73-173) people to improve mental health problems. We also found a substantial population impact (PIN, 167; 95% CI, 100-360) for interventions targeting cyberbullying (ie, 167 young people on average need to receive the intervention to prevent 1 case of cyberbullying perpetration or exposure). To put these results into context, the PIN is 35 450 for taking aspirin to avoid 1 death during the 6 months after a first nonhemorrhagic stroke,95 and the PIN is 324 for human papillomavirus vaccination in girls to prevent cervical cancer.96
Therefore, the findings of the present study have implications for public health given the adverse effects of bullying and cyberbullying on the mental and physical health of those involved9,14,17-20,97 and the cost-effectiveness of bullying prevention programs.7,25 In this regard, a recent cost-effectiveness analysis of KiVA, a Finnish school-based anti-bullying program delivered by teaching staff, estimated net savings of $66 172 for a cohort of 200 pupils through age 50 years.7 This finding is further evidence that bullying prevention interventions should become a priority for universal primary prevention in mental health.1
Universal anti-bullying interventions were found to be at least as effective as targeted interventions, if not more so, thus providing additional support for a universal approach as a first tier in school bullying prevention. Further research should address whether combining universal and targeted anti-bullying interventions is associated with greater effectiveness, especially in specific subgroups of children with greater needs, who may benefit less from universal school interventions.98
We found that duration of intervention (mean duration of interventions, 29.4 weeks; range, 1 session to 144 weeks) was not statistically significantly associated with effectiveness of anti-bullying interventions. This result is consistent with a previous meta-analysis28 that found no differences in the effectiveness of anti-bullying programs for prevention of school violence and bullying exposure between programs with a duration longer vs shorter than 1 year. Because most of the interventions included in the present meta-analysis lasted less than 1 year and some were as short as a few weeks, our findings suggest that short school interventions spanning less than 1 school year may be sufficient to substantially reduce bullying rates and improve mental health in young people, thus supporting their applicability in wider clinical contexts.
One of the key issues in the field of intervention in psychiatry and psychology, including preventive interventions like anti-bullying programs, is the durability of treatment effects.99-101 This meta-analysis shows that the effectiveness of the intervention does not diminish over time after the end of the intervention (mean [range] follow-up, 30.9 [2-104] weeks), thus suggesting that there is long-term effectiveness, with a relevant population impact (PIN, 127; 95% CI, 87-247), and that there are no statistically significant differences between effect sizes at the end of the intervention and after follow-up ranging from 2 to 104 weeks. It remains unclear whether treatment effectiveness can be convincingly maintained over longer periods, whether longer-term interventions or booster sessions might provide additional benefits, and the specific factors that might be associated with a sustained response.
Despite the favorable results of anti-bullying interventions and the fact that these programs are traditionally considered to be low-risk interventions, it should not be presumed that school interventions targeting bullying are always safe.102 Along these lines, the findings offer encouraging data on the safety of anti-bullying programs. None of the selected RCTs reported an increase in either bullying perpetration or bullying exposure at study end point or follow-up. Furthermore, mental health improved in all trials that assessed this outcome, both at study end point and follow-up. Therefore, our results suggest that anti-bullying interventions are not only efficacious but also safe.
This meta-analysis found no differential factors moderating the effectiveness of anti-bullying and anti-cyberbullying programs. There is still ongoing debate about whether cyberbullying is categorically distinct from traditional bullying and about the role of common and differential factors associated with both types of bullying.103-105 Although bullying and cyberbullying may be mediated by some differential factors, such as emotional problems and the personality of bullies,106 there is no clear cutoff between moderators of bullying and cyberbullying. This observation could be at least partially owing to the fact that a high proportion of cyberbullying recipients also experience traditional bullying.107,108 However, this finding could also be a consequence of the difficulty of researching the online world and comparing it with the off-line world.105,109
Consistent with a previous meta-analysis,27 we found some differences in the effectiveness of interventions in trials conducted in Europe vs North America. In both regions, anti-bullying programs showed comparable effectiveness in decreasing overall bullying and bullying perpetration at study end point and were statistically significantly efficacious in decreasing attitudes that encourage bullying and improving mental health problems, albeit with smaller effect size in the subgroup of trials conducted in North America. However, the effectiveness in decreasing bullying exposure and promoting attitudes that discourage bullying at study end point was statistically significant only in the subgroup of trials conducted in Europe. Reduced effectiveness for some outcome measures in trials performed in North America vs Europe may be the result of methodological heterogeneity (eg, differences in specific programs or study design) or statistical heterogeneity among trials, which was high for RCTs in both regions. However, these findings may also reflect differences in social, educational, or cultural context that could be incorporated into the design of anti-bullying programs. Further studies should also try to clarify the potential reasons underlying differing effectiveness for some outcome measures in North America.
This study had several limitations. First, there was substantial methodological heterogeneity among selected RCTs in terms of intervention characteristics, measures of bullying exposure and other outcome variables, sample characteristics, and social context. Additional sources of heterogeneity included differences in study quality and statistical heterogeneity among RCTs, which was high for most outcome variables. Second, few RCTs assessed the same specific anti-bullying program. Therefore, we could only assess the effectiveness of anti-bullying programs as a whole. It is known that not all anti-bullying programs are efficacious and that effectiveness may vary in different settings.29,99 Meta-regression analyses were conducted to investigate the association of study-related factors. However, the results did not allow us to identify which program works in a specific context. Third, most outcome variables were not clearly defined in the original RCTs, and they were highly heterogeneous, with few RCTs assessing the same outcome variable. Therefore, it was necessary to classify and subsume specific outcome measures into outcome groups, and there may be some degree of overlap and heterogeneity in terms of internal validity among the resulting categories. Fourth, the original RCTs did not report the presence (or rates) of mental, intellectual, or physical disabilities in sample populations. This limitation may have altered both rates of bullying and bullying exposure and intervention effectiveness.110,111 Fifth, despite our comprehensive search in databases covering the main scientific fields relevant to this meta-analysis, we restricted our search to RCTs published in peer-reviewed journals. Although this strategy may have limited the representativeness of our search to some extent, we made this decision to ensure a minimum quality of the included trials in an attempt to increase the validity of our results. Sixth, none of the RCTs reported concomitant individual interventions, which may have altered the effectiveness of the anti-bullying programs.
Despite the small effect sizes and some regional differences in effectiveness, with greater effectiveness for some outcomes in trials conducted in Europe vs North America, this meta-analysis suggests that school anti-bullying interventions have valuable population impact. This meta-analysis also highlights the need for better designed trials that assess the factors associated with the effectiveness of anti-bullying programs, including optimal timing and duration of interventions, their essential components, and the mediating association between bullying prevention and improvement in mental health problems. Trials should also specifically test targeted interventions in vulnerable populations at higher risk for bullying exposure, such as those living with disabilities and LGBTQ (lesbian, gay, bisexual, transgender, queer; lesbian, gay, bisexual, transgender, questioning) youth.112,113 These studies could inform more efficacious bullying prevention programs that promote a reduction in bullying rates and improve global and mental health.
Accepted for Publication: June 26, 2020.
Corresponding Author: David Fraguas, MD, PhD, Institute of Psychiatry and Mental Health, Hospital Clínico San Carlos, Instituto de Investigación Sanitaria del Hospital Clínico San Carlos (IdISSC), Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), School of Medicine, Universidad Complutense, Calle de la Cabeza 4, 28012 Madrid, Spain (davidfraguas@gmail.com).
Published Online: November 2, 2020. doi:10.1001/jamapediatrics.2020.3541
Author Contributions: Drs Fraguas and Díaz-Caneja contributed equally to this work. Dr Fraguas had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Fraguas, Díaz-Caneja, Ayora, Abregú-Crespo, Arango.
Acquisition, analysis, or interpretation of data: Fraguas, Díaz-Caneja, Durán-Cutilla, Ezquiaga-Bravo, Martín-Babarro, Arango.
Drafting of the manuscript: Fraguas, Díaz-Caneja, Abregú-Crespo.
Critical revision of the manuscript for important intellectual content: Fraguas, Díaz-Caneja, Ayora, Durán-Cutilla, Ezquiaga-Bravo, Martín-Babarro, Arango.
Statistical analysis: Fraguas.
Obtained funding: Fraguas, Díaz-Caneja, Arango.
Administrative, technical, or material support: Fraguas, Ayora, Durán-Cutilla, Abregú-Crespo.
Supervision: Fraguas, Díaz-Caneja, Ayora, Martín-Babarro, Arango.
Conflict of Interest Disclosures: Dr Fraguas reported being a consultant to and/or receiving fees from Angelini, Eisai, IE4Lab, Janssen, Lundbeck, and Otsuka and receiving grant support from Fundación Alicia Koplowitz and Instituto de Salud Carlos III (Spanish Ministry of Science and Innovation). Dr Díaz-Caneja reported receiving honoraria from AbbVie, Exeltis, and Sanofi. Drs Díaz-Caneja, Ayora, and Durán-Cutilla and Mr Abregú-Crespo reported receiving grant support from Instituto de Salud Carlos III. Dr Arango reported being a consultant to or receiving honoraria or grants from Acadia, Angelini, Gedeon Richter, Janssen-Cilag, Lundbeck, Otsuka, Roche, Sage Therapeutics, Servier, Shire, Schering-Plough, Sumitomo Dainippon Pharma, Sunovion, and Takeda. No other disclosures were reported.
Funding/Support: This work was supported by Instituto de Salud Carlos III (grants PI14/00397 [principal investigator, Dr Fraguas], PI16/02012 [principal investigator, Dr Arango], and PI17/00481 [principal investigator, Dr Fraguas]), cofinanced by the European Regional Development Fund from the European Commission (“A way of making Europe”). It was also supported by Centro de Investigación Biomédica en Red de Salud Mental, Madrid Regional Government (grant B2017/BMD-3740 AGES-CM-2 [principal investigator, Dr Arango]), European Union Structural Funds, the European Union’s Seventh Framework Program under grant FP7-HEALTH-2013-2.2.1-2-603196 (Project PSYSCAN [principal investigator, Dr Arango]), the European Union’s H2020 Program under the Innovative Medicines Initiative 2 Joint Undertaking (grant 115916 [Project PRISM {principal investigator, Dr Arango}] and grant 777394 [Project AIMS-2-TRIALS] {principal investigator, Dr Arango}), Fundación Familia Alonso (principal investigator, Dr Arango), Fundación Alicia Koplowitz (principal investigators, Drs Arango and Fraguas), and Fundación Mutua Madrileña (principal investigator, Dr Arango).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We acknowledge the contributions of the following individuals at the Observatorio para la Convivencia Escolar (Observatory for School Coexistence), Madrid Regional Government, to this study: Rafael van Grieken Salvador, Cristina Álvarez Sánchez, Alicia Aguado Abril, Ángel Antolínez Cuesta, Rafael Carbonell Peris, José Maximino García González, Juan Carlos García González, Fernando Gómez Rivas, María de los Ángeles González Martín, Camilo Jené Perez, Epifanio Martínez Nohales, Juan José Nieto Romero, Teresa Pacheco Tabuenca, Ofelia Ramírez Fernández, José Luis Riesco Conde, Graciela Sánchez Ramos, Ismael Sanz Labrador, Concepción Vereterra Gutiérrez-Maturana, and José Manuel Arribas Álvarez. They were not compensated for their contributions.
2.Olweus
D. Bullying at School: What We Know and What We Can Do. Blackwell; 1993.
3.Olweus
D. Bully at school: long-term outcomes for the victims and an effective school-based intervention program. In: Huesmann LR, ed.
Aggressive Behavior: Current Perspectives. Plenum Press; 1994:97-129. Plenum Series in Social/Clinical Psychology. doi:
10.1007/978-1-4757-9116-7_5
6.Beecham
J, Byford
S, Kwok
C, Parsonage
M. School-based interventions to reduce bullying. In: Knapp M, McDaid D, Parsonage M, eds. Mental Health Promotion and Prevention: The Economic Case. London Dept of Health, London School of Economics and Political Science; 2011.
8.Chester
KL, Callaghan
M, Cosma
A,
et al. Cross-national time trends in bullying victimization in 33 countries among children aged 11, 13 and 15 from 2002 to 2010.
Eur J Public Health. 2015;25(suppl 2):61-64. doi:
10.1093/eurpub/ckv029
PubMedGoogle ScholarCrossref 10.National Center for Education Statistics. Student Reports of Bullying: Results From the 2015 School Crime Supplement to the National Crime Victimization Survey. National Center for Education Statistics, US Dept of Education; 2016.
15.Varese
F, Smeets
F, Drukker
M,
et al. Childhood adversities increase the risk of psychosis: a meta-analysis of patient-control, prospective- and cross-sectional cohort studies.
Schizophr Bull. 2012;38(4):661-671. doi:
10.1093/schbul/sbs050
PubMedGoogle ScholarCrossref 21.Ttofi
MM, Farrington
DP, Lösel
F, Loeber
R. The predictive efficiency of school bullying versus later offending: a systematic/meta-analytic review of longitudinal studies.
Crim Behav Ment Health. 2011;21(2):80-89. doi:
10.1002/cbm.808
PubMedGoogle ScholarCrossref 32.Heller
RF, Dobson
AJ, Attia
J, Page
J. Impact numbers: measures of risk factor impact on the whole population from case-control and cohort studies.
J Epidemiol Community Health. 2002;56(8):606-610. doi:
10.1136/jech.56.8.606
PubMedGoogle ScholarCrossref 33.Shamseer
L, Moher
D, Clarke
M,
et al; PRISMA-P Group. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation.
BMJ. 2015;350:g7647. doi:
10.1136/bmj.g7647
PubMedGoogle ScholarCrossref 34.Higgins
JP, Altman
DG, Gøtzsche
PC,
et al; Cochrane Bias Methods Group; Cochrane Statistical Methods Group. The Cochrane Collaboration’s tool for assessing risk of bias in randomised trials.
BMJ. 2011;343:d5928. doi:
10.1136/bmj.d5928
PubMedGoogle ScholarCrossref 35.Borenstein
M, Hedges
L, Higgins
J, Rothstein
H. Comprehensive Meta Analysis Version 2.0. Biostat Inc; 2005.
37.Lipsey
M, Wilson
D. Practical Meta-analysis. Sage Publications; 2000.
45.Athanasiades
C, Kamariotis
H, Psalti
A, Baldry
AC, Sorrentino
A. Internet use and cyberbullying among adolescent students in Greece: The “TABBY” project.
Hell J Psychol. 2015;12:14-39.
Google Scholar 48.Boulton
MJ, Flemington
I. The effects of a short video intervention on secondary school pupils’ involvement in definitions of and attitudes towards bullying.
Sch Psychol Int. 1996;17:331. doi:
10.1177/0143034396174003Google ScholarCrossref 51.Connolly
J, Josephson
W, Schnoll
J,
et al. Evaluation of a youth-led program for preventing bullying, sexual harassment, and dating aggression in middle schools.
J Early Adolesc. 2015;35(3):403-434. doi:
10.1177/0272431614535090Google ScholarCrossref 52.Conduct Problems Prevention Research Group. The effects of a multiyear universal social-emotional learning program: The role of student and school characteristics.
J Consult Clin Psychol. 2010;78(2):156-168. doi:
10.1037/a0018607PubMedGoogle ScholarCrossref 57.Espelage
DL, Rose
CA, Polanin
JR. Social-Emotional Learning Program to reduce bullying, fighting, and victimization among middle school students with disabilities.
Remedial Spec Educ. 2015;36(5):299-311. doi:
10.1177/0741932514564564Google ScholarCrossref 58.Farmer
VL, Williams
SM, Mann
JI, Schofield
G, McPhee
JC, Taylor
RW. Change of school playground environment on bullying: a randomized controlled trial.
Pediatrics. 2017;139(5):e20163072. doi:
10.1542/peds.2016-3072PubMedGoogle Scholar 61.Frey
KS, Hirschstein
MK, Snell
JL, Edstrom
LV, MacKenzie
EP, Broderick
CJ. Reducing playground bullying and supporting beliefs: an experimental trial of the steps to respect program.
Dev Psychol. 2005;41(3):479-490. doi:
10.1037/0012-1649.41.3.479PubMedGoogle ScholarCrossref 66.Holen
S, Waaktaar
T, Lervåg
A, Ystgaard
M. Implementing a universal stress management program for young school children: are there classroom climate or academic effects?
Scand J Educ Res. 2013;57(4):420-444. doi:
10.1080/00313831.2012.656320Google ScholarCrossref 67.Hormazábal-Aguayo
I, Fernández-Vergara
O, González-Calderón
N,
et al. Can a before-school physical activity program decrease bullying victimization in disadvantaged children? The Active-Start Study.
Int J Clin Health Psychol. 2019;19(3):237-242. doi:
10.1016/j.ijchp.2019.05.001PubMedGoogle ScholarCrossref 72.Kärnä
A, Voeten
M, Little
TD, Alanen
E, Poskiparta
E, Salmivalli
C. Effectiveness of the KiVa antibullying program: grades 1–3 and 7–9.
J Educ Psychol. 2013;105(2):535-551. doi:
10.1037/a0030417Google ScholarCrossref 77.Nocentini
A, Menesini
E, Pluess
M. The personality trait of environmental sensitivity predicts children’s positive response to school-based antibullying intervention.
Clin Psychol Sci. 2018;6(6)1-12. doi:
10.1177/2167702618782194Google ScholarCrossref 78.Ostrov
JM, Godleski
SA, Kamper-DeMarco
KE, Blakely-McClure
SJ, Celenza
L. Replication and extension of the Early Childhood Friendship Project: effects on physical and relational bullying.
School Psych Rev. 2015;44(4):445-463. doi:
10.17105/spr-15-0048.1Google ScholarCrossref 79.Sanchez
E, Robertson
TR, Lewis
CM, Rosenbluth
B, Bohman
T, Casey
DM. Preventing bullying and sexual harassment in elementary schools the Expect Respect Model.
J Emotional Abuse. 2001;2(2-3):157-180. doi:
10.1300/J135v02n02_10Google ScholarCrossref 81.Sorrentino
A, Baldry
AC, Farrington
DP. The efficacy of the Tabby improved prevention and intervention program in reducing cyberbullying and cybervictimization among students.
Int J Environ Res Public Health. 2018;15(11):2536. doi:
10.3390/ijerph15112536PubMedGoogle ScholarCrossref 83.Trip
S, Bora
C, Sipos-Gug
S,
et al. Bullying prevention in schools by targeting cognitions, emotions, and behavior: evaluating the effectiveness of the REBE-ViSC program.
J Couns Psychol. 2015;62(4):732-740. doi:
10.1037/cou0000084PubMedGoogle ScholarCrossref 84.Tsiantis
ACJ, Beratis
IN, Syngelaki
EM, Stefanakou
AA, Asimopoulos
C, Sideridis
GD. The effects of a clinical prevention program on bullying, victimization, and attitudes toward school of elementary school students.
Behav Disord. 2013;38(4):243-257. doi:
10.1177/019874291303800406Google ScholarCrossref 88.Espelage
DL, Rose
CA, Polanin
JR. Social-emotional learning program to promote prosocial and academic skills among middle school students with disabilities.
Remedial Spec Educ. 2016;37(6)1-10. doi:
10.1177/0741932515627475Google ScholarCrossref 89.Mallick
R, Kathard
H, Borhan
ASM, Pillay
M, Thabane
L. A cluster randomised trial of a classroom communication resource program to change peer attitudes towards children who stutter among grade 7 students.
Trials. 2018;19(1):664. doi:
10.1186/s13063-018-3043-3PubMedGoogle ScholarCrossref 90.Moore
B, Woodcock
S, Dudley
D. Developing wellbeing through a randomised controlled trial of a martial arts based intervention: an alternative to the anti-bullying approach.
Int J Environ Res Public Health. 2018;16(1):81-99. doi:
10.3390/ijerph16010081PubMedGoogle ScholarCrossref 92.Yeager
DS, Trzesniewski
KH, Dweck
CS. An implicit theories of personality intervention reduces adolescent aggression in response to victimization and exclusion.
Child Dev. 2013;84(3):970-988. doi:
10.1111/cdev.12003PubMedGoogle ScholarCrossref 94.Jadambaa
A, Thomas
HJ, Scott
JG, Graves
N, Brain
D, Pacella
R. Prevalence of traditional bullying and cyberbullying among children and adolescents in Australia: a systematic review and meta-analysis.
Aust N Z J Psychiatry. 2019;53(9):878-888. doi:
10.1177/0004867419846393
PubMedGoogle ScholarCrossref 100.Rith-Najarian
LR, Mesri
B, Park
AL, Sun
M, Chavira
DA, Chorpita
BF. Durability of cognitive behavioral therapy effects for youth and adolescents with anxiety, depression, or traumatic stress: a meta-analysis on long-term follow-ups.
Behav Ther. 2019;50(1):225-240. doi:
10.1016/j.beth.2018.05.006
PubMedGoogle ScholarCrossref 101.Weisz
JR, Kuppens
S, Ng
MY,
et al. What five decades of research tells us about the effects of youth psychological therapy: a multilevel meta-analysis and implications for science and practice.
Am Psychol. 2017;72(2):79-117. doi:
10.1037/a0040360
PubMedGoogle ScholarCrossref 102.Jeong
S, Lee
BH. A multilevel examination of peer victimization and bullying preventions in schools.
J Criminology. 2013:Article ID 735397. doi:
10.1155/2013/735397Google Scholar 105.Seiler
SJ, Navarro
JN. Bullying on the pixel playground: investigating risk factors of cyberbullying at the intersection of children’s online-offline social lives.
Cyberpsychology: J Psychosocial Res Cyberspace. 2014;8(4):Article 6. doi:
10.5817/CP2014-4-6Google Scholar 110.Pinquart
M. Systematic review: bullying involvement of children with and without chronic physical illness and/or physical/sensory disability—a meta-analytic comparison with healthy/nondisabled peers.
J Pediatr Psychol. 2017;42(3):245-259.
PubMedGoogle Scholar 111.King
T, Aitken
Z, Milner
A,
et al. To what extent is the association between disability and mental health in adolescents mediated by bullying? a causal mediation analysis.
Int J Epidemiol. 2018;47(5):1402-1413. doi:
10.1093/ije/dyy154
PubMedGoogle ScholarCrossref 112.Earnshaw
VA, Reisner
SL, Juvonen
J, Hatzenbuehler
ML, Perrotti
J, Schuster
MA. LGBTQ bullying: translating research to action in pediatrics.
Pediatrics. 2017;140(4):e20170432. doi:
10.1542/peds.2017-0432
PubMedGoogle Scholar