A Systematic Review of School-Based Interventions to Prevent Bullying FREE

Bullying is a form of aggression in which 1 or more children repeatedly and intentionally intimidate, harass, or physically harm a victim.¹ Victims of bullying are perceived by their peers as physically or psychologically weaker than the aggressor(s), and victims perceive themselves as unable to retaliate.² Although bullying, harassment, and victimization can take many forms, the key elements of this behavior are aggression, repetition, and the context of a relationship with an imbalance of power.³

Bullying can impact the physical, emotional, and social health of the children involved. Victims of bullying more often report sleep disturbances, enuresis, abdominal pain, headaches, and feeling sad than children who are not bullied.^{4 - 5} Bullies, their victims, and those who are both bullies and victims have significantly increased risk for depressive symptoms and suicidal ideation.^{6 - 7} Students who report victimization are 3 to 4 times more likely to report anxiety symptoms than uninvolved children.^{8 - 9} The effects of bullying on emotional health may persist over time; 1 study¹⁰ showed that children bullied repeatedly through middle adolescence had lower self-esteem and more depressive symptoms as adults. Victims of bullying are more likely to feel socially rejected or isolated and to experience greater social marginalization and lower social status.¹¹

Bullying impacts a child's experience of school on numerous levels. Bullying creates problems with school adjustment and bonding, affecting the victims' completion of homework or desire to do well at school.^{6 ,12} In 1 study,¹³ 20% of grade-school children reported being frightened through much of the school day. Bullying seems to increase school absenteeism, with victimized children becoming more school avoidant as the victimization increases.¹⁴ Furthermore, involvement in bullying affects academic performance, although studies^{15 - 18} show mixed results regarding which children are most affected. Most bullying takes place at school, particularly at times and places where supervision is minimal.¹⁹ Schools where adults tolerate more bullying may have more severe bullying problems.¹⁰

As school bullying increasingly becomes a topic of public concern and research efforts, a growing number of studies examines school-based interventions targeted to reduce bullying. Although many of these interventions have been rigorously evaluated, the evaluations reveal mixed results.²⁰ For example, evaluations of the Olweus Bullying Prevention Program, a comprehensive “whole-school” intervention on which many subsequent programs have been based, report reductions of 30% to 70% in the student reports of being bullied and bullying others.^{2 ,20 - 22} In contrast, evaluation of a similar comprehensive prevention program implemented in Belgium did not show significant differences in victimization or bullying scores among primary or secondary school students.²³ Although some review articles have described several of these interventions, to our knowledge, no systematic reviews of interventions to reduce bullying have been published in peer-reviewed literature. The objective of this study was to review rigorously evaluated school-based interventions to reduce or prevent bullying with the goal of determining whether these interventions worked.

We searched several bibliographic databases, including MEDLINE (January 1, 1966, through August 23, 2004), PsycINFO, EMBASE, Educational Resources Information Center, the Physical Education Index, Sociology: A SAGE Full-Text Collection, and the Cochrane Clinical Trials Registry (all as of August 23, 2004). We used the search terms bullying or bully as Medical Subject Headings or keywords. We used a keyword search because it was more robust than searches using only Medical Subject Headings. One of us (R.C.V.) reviewed the titles of all returned articles and the bibliographies of all relevant review articles to determine which studies examined a school-based intervention to prevent or reduce bullying. Articles were immediately excluded if they obviously did not include an intervention or did not occur at a school.

After articles that clearly did not meet the inclusion criteria were excluded, both of us (R.C.V. and A.E.C.), blinded to the journal citation and article text other than the “Methods” section, independently reviewed the articles. The 2 reviewers independently decided on trial inclusion using a standard form with predetermined eligibility criteria. Disagreements were resolved by consensus reached after discussion. For inclusion, a study needed to describe an experimental intervention with control and intervention groups and to include a follow-up evaluation with measured outcomes. In addition, the intervention needed to be school based and designed to reduce or prevent bullying. Each article was analyzed to determine the study method, intervention components, outcomes measured, and results. There was no assessment of quality in choosing or evaluating study outcomes beyond the inclusion criteria. We did not exclude or discount studies based on baseline similarities among treatment groups, study power, retention rates, or program intensity because these characteristics are not associated definitively with the strength of treatment effects.²⁴ Duplicate publications or multiple articles that reported identical outcomes measured over the same period on the same population were excluded.

We extracted data from the selected articles regarding direct outcome measures of bullying, including bullying, victimization, aggressive behavior, violence, school responses to violence, and violent injuries. Data were also extracted for outcomes thought to be indirectly related to bullying, such as school achievement, perception of school safety, self-esteem, or knowledge about or attitudes toward bullying.

The systematic literature search identified 2090 articles. The online search of MEDLINE yielded 353 articles, and the search of EMBASE yielded 269 articles, 9 of which were not found by the MEDLINE search. The search of PsycINFO yielded 897 articles, Educational Resources Information Center yielded 552 articles, the Physical Education Index yielded 16 articles, and Sociology: A SAGE Full-Text Collection yielded 3 articles. An additional 4 potential studies^{25 - 28} were identified through searches of bibliographies and were also reviewed. Once articles that obviously did not address school-based interventions were excluded, 321 relevant articles remained. Reviewing the abstracts of these articles allowed for the further exclusion of articles that did not address school-based interventions. Fifty-six articles were then assessed by both of the reviewers. Articles were most commonly excluded at this stage because they were not evaluations of interventions, they did not have control groups, or they did not address bullying.

The 26 studies that met the selection criteria varied substantially in intervention type, study population, study design, and outcome measures. The detailed characteristics of the studies are reported in Table 1. The interventions could be divided into 5 categories: curriculum interventions, multidisciplinary or whole-school interventions, targeted social and behavioral skills groups, mentoring, and increased social work support. To maximize clarity and clinical usefulness, we present the subsequent results of the review based on the type of intervention. All 26 studies investigated interventions for some group of primary school students, but the primary grade levels varied from first to eighth grade. Six studies^{22 - 23 ,30 ,39 ,41 ,48} included secondary school students (older than eighth grade) in their interventions and outcomes. The selected studies reported a range of outcomes that were subsequently categorized into direct and indirect outcomes. The direct and indirect outcomes of all of the studies are reported in Table 2.

Ten studies^{23 ,26 ,29 - 36} evaluated the implementation of new curriculum. The curriculum interventions included videotapes, lectures, and written curriculum, and varied in intensity from a single videotape followed by classroom discussion to 15 weeks of classroom modules. The details of the study designs, participants, intervention type, and important outcomes of the curriculum interventions are all described in Table 1. The studies all used a pretest, posttest, control group design; 6 of the 10 studies randomized the assignment of the groups.^{23 ,29 - 31 ,35 - 36}

The curriculum interventions did not consistently decrease bullying, and several actually suggested that the bullying within the intervention group increased (Table 2). Of the 10 studies of curriculum interventions, 6 showed no significant improvements in bullying.^{23 ,26 ,30 ,32 - 33 ,36} Although bullying and victimization did not change significantly, Boulton and Flemington³⁰ did find that the students in the intervention group broadened their definition of bullying slightly, and Englert³² found that the teachers reported a significant decrease in observed physical and verbal violence (P<.01).

Of the 4 studies that did show less bullying after a curriculum intervention, 3 also showed more bullying or victimization in certain populations or with certain measurement tools.^{29 ,34 - 35} The study by Baldry and Farrington²⁹ showed a decrease in self-reported victimization among older children (P<.05), but younger children actually reported more victimization (P<.01), and there were no significant differences in either victimization or bullying overall. Teglasi and Rothman³⁵ found that teachers reported decreased antisocial behavior for children not identified as aggressive and increased aggressive behavior for the children previously identified as aggressive (P<.01 for both). The individual self-reports for aggression did not reveal any significant effects from the intervention. A study by Rican et al³⁴ found significant decreases in peer nominations of bullying (P=.02) and victims (P=.03) using unspecified “broad criteria,” but no change in victimization using “narrower criteria.” Only 1 curriculum intervention showed unequivocal improvements, and this was in an indirect outcome. The randomized trial of Elliott and Faupel³¹ of a group problem-solving curriculum resulted in increased generation of responses to a simulated bullying situation by the intervention group.

Ten studies^{22 ,25 ,28 ,37 - 43} evaluated interventions using a multidisciplinary whole-school approach that included some combination of schoolwide rules and sanctions, teacher training, classroom curriculum, conflict resolution training, and individual counseling. Table 1 describes the components of each of these multidisciplinary studies in detail. The whole-school studies involved more subjects than the curriculum interventions, with up to 42 schools in a single study. Only 2 of the studies^{39 ,41} evaluated interventions among secondary school students, and the rest looked at primary schools. In contrast to the curriculum studies, only 2 of the whole-school studies incorporated randomization in their study design. Two of the studies^{41 ,50} used a quasi-experimental design with time-lagged age cohorts.

Two studies, both evaluating the seminal Olweus Bullying Prevention Program, revealed disparate results. The Olweus Bullying Prevention Program pioneered the whole-school approach to preventing and reducing bullying with an intervention program in Bergen, Norway, that included training for school personnel, materials for parents, a videotaped classroom curriculum, and evaluation through the bullying questionnaire developed by Olweus.⁵⁰ By using unspecified composite measures involving student questionnaires and teacher ratings, the follow-up evaluation found decreased bullying, decreased victimization, decreased antisocial behavior, and improved school climate after the intervention.^{22 ,51} Evaluation of the nationwide Olweus Bullying Prevention Program in Rogaland, Norway, revealed strikingly different results. Roland^{27 ,41} reported increased victimization and social exclusion for boys, and increased bullying for boys and girls based on student self-reports. Unlike in Olweus' protocol, the schools in this sample did not interact with the researchers during the intervention. The schools' degree of involvement in the program was directly related to positive effects from the antibullying program, particularly for girls. Although the evaluations apparently involved the same nationwide campaign and evaluation tools, Olweus states that they “were completely different in terms of planning, data quality, times of measurement, and contact with the schools.”^{3 (p39)}

Since the publication of the study by Olweus,⁵¹ interventions targeting the whole school have been implemented in several other countries. Overall, these whole- school studies had positive effects on bullying. Of the additional 8 studies, 7 revealed positive outcomes.^{25 ,37 - 40 ,42 - 43} Five of these studies^{25 ,37 - 38 ,40 ,43} reported decreases in bullying or victimization. Among kindergarteners, Alsaker and Valkanover²⁵ found decreased victimization on teacher and student reports, although there was no significant change in bullying on either student nominations or teacher ratings. An Italian schoolwide peer support intervention prevented some of the increased negative behaviors and attitudes reported in the control group on student reports.³⁷ Examining administrative office records, Metzler et al³⁸ found decreased discipline referrals (P=.04) and harassment (P=.02) in select populations after 2 years of implementing a schoolwide behavioral management program. However, student reports of physical and verbal attacks did not significantly change. Twemlow et al⁴³ found decreased disciplinary referral rates, decreased suspension rates, and increased achievement test scores after a schoolwide intervention. In their evaluation of a 12-week schoolwide program, Rahey and Criag⁴⁰ found mixed results. On peer and self-reports, the students in grades 5 through 8 had decreased victimization (P<.05) and decreased peer isolation (P<.01) compared with the controls.⁴⁰ In contrast, the younger students reported increased levels of victimization (P<.05) and increased exclusion (P<.01).⁴⁰ In addition, neither student nor teacher reports showed a significant decrease in bullying for any age group.

Two of the studies^{39 ,42} that revealed improvements after a multidisciplinary intervention only measured outcomes indirectly related to bullying. Some of the indirect outcomes were positively affected, but others remained unchanged. The evaluation by Sanchez et al⁴² indicated no significant increase in knowledge of bullying, but the intervention students were more likely to report seeing bullying and to express readiness to intervene personally (P<.05 for both). A randomized controlled trial³⁹ evaluating an intervention to develop “health-promoting schools” revealed an increased awareness of health-related policies and practices among school administrators after the intervention (P=.04). No significant changes in health-related policies or practices occurred in the intervention group, including those related to bullying.

The other evaluation of a whole-school intervention that we identified revealed no significant decreases in bullying. Melton et al²⁸ implemented an intervention based on the Olweus Bullying Prevention Program in 6 rural school districts in South Carolina. After 2 years, they found no significant differences in student self-reports of bullying, victimization, general antisocial behaviors, or attitudes toward bullying between the intervention and control students.

Four studies^{44 - 47} looked at targeted interventions involving social and behavioral skills groups for children involved in bullying. Two of these interventions specifically targeted children with high levels of aggression,^{45 - 46} while the other 2 targeted children who were themselves victims.^{44 ,47} Of the 4 studies, 3 focused on older students, in sixth through eighth grades,^{45 - 47} while the fourth looked at third-grade students.⁴⁴ The most positive outcomes occurred for the youngest students.⁴⁴ DeRosier⁴⁴ tested the efficacy of social skills group training for third-grade students with peer relationship difficulties in 11 public primary schools in North Carolina. The intervention resulted in decreased aggression on peer reports (P<.001), decreased bullying on self-reports (P<.05), and fewer antisocial affiliations on self-reports (P<.05) for the previously aggressive children. This was the only social skills training intervention that showed clear reductions in bullying from the intervention.

The other social skills group interventions, all of which involved older children, did not result in clear changes. Meyer and Lesch⁴⁶ evaluated a behavioral skills modification program for boys identified as bullies in South Africa. This intervention did not produce any statistically significant decrease in bullying behaviors by peer report or self-report. Fast et al⁴⁵ examined whether group training in peer mediation for aggressive seventh graders would reduce their level of aggression. The aggressive students in the intervention group had a significant decrease in their problem behaviors as measured by their teachers (P<.05); however, no significant changes in disciplinary referrals for aggressive behavior or impulsivity scores occurred. Tierney and Dowd⁴⁷ used social skills group training for eighth-grade girls with emotional and behavioral concerns. Although the teacher data indicated statistically significant progress in the areas of friendships, behavior, interactions with peers, and level of teacher concern, the data and analysis were not given within the article. Data from the pupil questionnaires were not analyzed, but in student self-reports, 8 reported no difference in victimization and 7 reported improvement.

A single study, done by Bagley and Pritchard,⁴⁸ examined the effects of an increase in the number of school social workers focused on problem behaviors, including bullying. Compared with matched control schools, they found a significant decrease in self-reported bullying within the primary school (P<.05), but worsening bullying in the secondary school. For self-reports of theft, truancy, fighting, and drug use, the primary and secondary intervention schools had significant improvements (P<.05). A study by King et al⁴⁹ investigated the effects of a mentoring program for “at-risk” children. The mentored students were significantly less likely than their nonmentored age-matched peers to report bullying (P<.002), physically fighting (P<.001), and feeling depressed (P=.006) in the past 30 days.

As governments, schools, and educators invest increasing amounts of money and time into antibullying interventions, the findings of this review provide evidence for how to best achieve the desired outcome of decreased bullying within schools. By systematically gathering and compiling the growing number of studies evaluating these interventions, it becomes clear that some of the antibullying interventions actually decrease bullying, while others have no effect or even seem to increase the amount of bullying. Grouping the studies by the type of intervention seems to offer the most insight into what leads to success.

The curriculum interventions were generally designed to promote an antibullying attitude within the classroom and to help children develop prosocial conflict resolution skills. Most of these interventions drew on the social cognitive principles of behavioral change,^{52 - 53} with focus on changing students' attitudes, altering group norms, and increasing self-efficacy. Curriculum changes are often attractive because they usually require a smaller commitment of resources, personnel, and effort. Nevertheless, the interventions that consisted only of classroom-level curriculum seldom improved bullying. The basis in social, cognitive, behavioral change may explain part of the problem; previous work^{54 - 55} suggests that younger children benefit less from these techniques. However, the failure of classroom-level interventions for older and younger students points to the systemic nature of bullying and supports the theory of bullying as a sociocultural phenomenon. If bullying is a systemic group process involving bullies, victims, peers, adults, parents, school environments, and home environments,⁵⁶ an intervention on only 1 level is unlikely to have a significant consistent impact. Furthermore, if bullying is, as some propose, a sociocultural phenomenon springing from the existence of specified social groups with different levels of power,⁵⁷ then curriculum aimed at altering the attitudes and behaviors of only a small subset of those groups is unlikely to have an effect.

Similarly, the targeted interventions providing training in social skills did not clearly improve bullying or victimization.^{45 - 47} The failure of these interventions, also based largely on social, cognitive, behavioral changes, points again to the inability of a single-level intervention to combat bullying effectively. Interestingly, the older children had worse outcomes from the social skills training groups than the younger children.⁴⁴ The 1 study looking at younger children found decreased aggression, bullying, and antisocial affiliations. Although one cannot generalize from a single study, it is possible that addressing social skills changes in the context of a small targeted group during a particular developmental window could be effective. Overall, the studies of social skills group interventions suggest again that failing to address the systemic issues and social environment related to bullying undermines success.

The whole-school interventions, which included multiple disciplines and complementary components directed at different levels of the school organization, more often reduced victimization and bullying than the interventions that only included classroom-level curricula or social skills groups. The whole-school interventions address bullying as a systemic problem meriting a systemic solution. They seek to alter the school's entire environment and to involve individuals, peer groups, classrooms, teachers, and administration. The success of the whole-school interventions suggests that bullying does, indeed, spring from factors external to individual children's psychosocial problems, including a complex process of social interactions. An evaluation of whole-school approaches by Smith et al^{56 (p557)} in 2004 suggests that these interventions may “reflect a reasonable rate of return on the investment inherent in low-cost, nonstigmatizing primary prevention programs.” Our review offers further support for using whole-school interventions to reduce or prevent bullying.

Despite the evidence pointing toward the value of whole-school approaches, significant barriers may still limit their effectiveness. The implementation of the intervention can vary significantly, and this clearly alters the results. The original antibullying whole-school approach studied in Bergen by Olweus⁵¹ and the evaluation of the same program in Rogaland by Roland^{27 ,41} produced the most strikingly disparate results. The contrast may have been the result of decreased school staff participation at the Rogaland schools.²⁷ In addition, the Olweus program does not include detailed instructions for replicating an identical program in another school setting. Difficulty in replicating this program may contribute to the lack of success when used in other settings, such as South Carolina.²⁸ Although the adaptation of the interventions in different settings may create more culturally appropriate interventions, these modifications may produce some of the variance in success. Unfortunately, the specific components of a given intervention are generally not described sufficiently to enable faithful replication. The specific school environment could also significantly impact effectiveness. The small class size, excellent teacher training, and tradition of social welfare intervention in some settings could enable better effects. The suggestion that whole-school interventions may not work as well for younger children, seen in 2 of the studies,^{25 ,40} also merits consideration. This, albeit limited, evidence may support a developmental theory, whereby bullying begins in early childhood as individuals assert themselves to gain dominance and then gradually evolves as children use less socially reprehensive ways to dominate others.⁵⁸ Schoolwide rules and changes in the school's overall responses to bullying may not be as effective in the younger population before they follow their natural developmental progression into conformity with social norms.

There are several limitations to this systematic review that warrant consideration. We only included studies in the English language. Although we may have, therefore, overlooked some relevant studies, we located few non–English-language studies that required exclusion. Some of the included studies did not have ideal methodological strength; however, many of the studies were reasonably well done and offered important counterbalances to the findings. The study results may be overestimated because, in many cases, schools or districts were randomized to treatment conditions, but the students were evaluated as the unit of analysis. The unit of analysis problem could result in a higher type I error if intraclass correlation is not taken into account.⁵⁹ Even so, many of the studies that did this still found no treatment effects. The use of variable outcome measures may further limit the ability to measure accurately the effects of these interventions. The most common outcome measures were self-reports of victimization and bullying that may not wholly correspond with information obtained from peers or teachers or from observations.⁶⁰ Still, self-reports are the standard measure used in most studies evaluating behavioral interventions. Despite the diversity of the evidence reviewed, the studies were primarily performed in Europe and the United States, which may limit the generalizability of the conclusions. In addition, several interventions with positive results, including interventions using mentoring, increased social workers by 2.5 workers, and social skills groups for younger children, were only studied on a single occasion, thus limiting their generalizability. Finally, while we attempted to separate out the most effective components or intervention strategies, many of the studies involved numerous complementary components that were not evaluated individually.

In conclusion, fairly consistent evidence suggests that children's bullying behavior can be significantly reduced by well-planned interventions. The chance of success is greater if the intervention incorporates a whole-school approach involving multiple disciplines and the whole school community. The school staff's commitment to implementing the intervention also may play a crucial role in its success. The use of curriculum or targeted social skills groups alone less often results in any decrease in bullying and sometimes worsens bullying and victimization. Caution should be exercised in supposing that antibullying interventions invariably produce the intended results. This review reveals that not all programs have proved effective. Most reductions in bullying tend to be relatively small and related more to the proportion of children being victimized rather than the proportion engaging in bullying. Additional research to evaluate bullying behaviors and antibullying interventions is clearly needed.

Correspondence: Rachel C. Vreeman, MD, Children's Health Services Research, Indiana University School of Medicine, 699 West Dr, Riley Research Room 330, Indianapolis, IN 46202 (rvreeman@iupui.edu).

Accepted for Publication: August 3, 2006.

Author Contributions: Dr Vreeman had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Vreeman and Carroll. Acquisition of data: Vreeman. Analysis and interpretation of data: Vreeman and Carroll. Drafting of the manuscript: Vreeman. Critical revision of the manuscript for important intellectual content: Vreeman and Carroll. Statistical analysis: Vreeman and Carroll. Administrative, technical, and material support: Vreeman and Carroll. Study supervision: Vreeman and Carroll.

Financial Disclosure: None reported.

Disclaimer: The views expressed herein are those of the authors and do not necessarily represent the views of Indiana University School of Medicine.