Right To Play’s intervention to reduce peer violence among children in public schools in Pakistan: a cluster-randomized controlled trial

ABSTRACT Background Peer violence is common globally, but a little researched topic in low-and middle-income countries. This study presents the evaluation of a two-year randomized controlled trial of a structured play-based life-skills intervention implemented in schools in Hyderabad, Pakistan. Objective To determine the impact of the intervention on school-based peer violence (victimization and perpetration) and depression among school children. Methods 40 single-sex public schools were randomized into two study arms (20 per arm 10 of each sex). A total of 1752 grade 6 students (929 from intervention and 823 from control schools) were enrolled in the trial. The two-year intervention was a biweekly structured game led by a coach followed by critical reflection and discussion for 30 minutes. Primary outcomes (exposure to peer violence exhibited through victimization and perpetration and depression) were evaluated using generalized linear-mixed models. Results Of the enrolled children (N = 1752) 91% provided data for analysis. There were significant decreases in self-reported peer violence victimization, perpetration and depression. For peer violence victimization, the reductions in the intervention and control arms were: 33.3% versus 27.8% for boys and 58.5% versus 21.3% for girls. For peer violence perpetration, the reductions were: 25.3% versus 11.1% for boys and 55.6% versus 27.6% for girls in the intervention and control arms, respectively. There were significant drops in mean depression scores (boys 7.2% versus 4.8% intervention and control and girls 9.5% versus 5.6% intervention and control). Conclusion A well-designed and implemented play-based life-skills intervention delivered in public schools in Pakistan is able to effect a significant reduction in peer violence.

Adolescent mental health; peer violence perpetration; peer violence victimization; school violence; empowerment; Asia; corporal punishment; gender attitudes Background Peer violence is a problem in schools worldwide and is associated with poor school performance [1], absenteeism [2], and school dropout [3,4], as well as aggression in later life [5]. A recent systematic review suggests strong causal relationships between victimization and mental health problems such as depression, anxiety, poor general health, and suicidal ideation [6]. Peer violence is defined as a repeated experience of any act of physical, verbal, or psychological abuse of victims by peer perpetrators (i.e. fellow students, youth in the community or peer relatives at home) with the intention of causing harm [7,8]. Peer violence extends beyond harassment and bullying. There is often overlap between 'victims' and 'perpetrators' in that many of the former also go on to perpetrate violence [9]. Victims are often chosen because they are weaker than perpetrators -smaller, younger, from a minority or less desirable group or less socially connected [7]. Much of the evidence about peer violence comes from the USA [10], where 22.7% of youth reported peer violence victimization in the National Survey of Children's Health in 2016 [11]. Prevalence estimates of peer violence in low-and middleincome countries suggest similar or higher rates [9,12,13].
The overwhelming majority of interventions to reduce peer violence are implemented and evaluated in high-income countries [3,14 -16]. A recent systematic review found only three studies which evaluated interventions in low-and middle-income countries and two of the three studies had a very high risk of bias and major weaknesses in their study design, making their findings hard to interpret. Only one, from Romania, was more robust, and it included a cognitive-behavioural component to enhance students' social and emotional learning. However, the findings showed that intervention was not effective in preventing bullying 9 months after baseline [16].
Evaluations of interventions to reduce peer victimization have been undertaken in some high-income country settings. Only cognitivebehavioural interventions have shown promise in preventing peer violence victimization and there has been little evidence of impact in follow-up assessments. Yet most peer violence interventions have focused on perpetration rather than victimization with social, emotional and cognitive components and peer mentoring showing some reductions in perpetration. Most evaluations have considered change between pre-and post-test, however evidence of the effectiveness of these interventions over the longer term, for example, one to 3 years, has also been shown [14].
Right To Play (https://www.righttoplay.ca/en-ca/) is an international non-governmental organization (NGO) that seeks to use the transformative power of play and sports to educate and empower children to lead healthier, empowered and safer lives. The Positive Child and Youth Development Programme (also called Red Ball Child Play) was developed by a team of educationists, athletes, teacher-trainers, and psychologists. The programme draws on social, cognitive, child development and experiential learning theories, focusing on physical, cognitive, social and emotional development through sports and games [17]. It is premised on the idea that children learn through processes of exploration and reinforcement of new ideas and behaviour, and that in order to achieve enduring change, interventions must be ongoing. Gender equality is a cross-cutting theme which is intended to give girls a voice through play and discussion. Behaviour change is viewed as a complex process that is best achieved when new ideas are explored in groups using empowering, participatory methods [18]. The activities seek to develop and build essential life skills, such as confidence, communication, empathy, coping with negative emotions, resilience, cooperation, leadership, critical thinking and conflict resolution in order to reduce intolerance, gender discrimination, and peer violence. The programme has been translated and adapted for different settings. In Pakistan, where the NGO has worked since 2008, the material has been translated into the Urdu and Sindhi languages. Although tested and refined over nearly 20 years across 18 countries, the programme had not been previously evaluated for long-term impact on children.
We report the results of an evaluation which assessed the impact of Right To Play's structured play-based intervention in single-sex, public schools in Pakistan. The objective was to determine whether this play-based life skills Programme could reduce peer violence perpetration and victimization, and depression among children in 6 th to 8 th grades, who were mostly aged 12-14 years at enrollment.

Study setting
We evaluated the intervention in 40 public middle schools in the Hyderabad district of Sindh province, in classes from grades 6-8 between December 2015 and January 2018. A full account of the methods [19] and analyses from baseline data [9,20 -22] are published elsewhere.

Design
We conducted a two-arm cluster-randomized controlled trial with parallel group assignment. One arm received the intervention for 2 years and the control arm, received the intervention in 2018 after data collection concluded. We encouraged participation in the control arm schools by offering schoolwide incentives such as providing drinking water tank or repairing school benches.
Only single-sex government schools educating students in grades 6-8 were eligible. Schools were required to have sufficiently large spaces (either indoor or outdoor) for safe play. A further requirement for inclusion was that the schools had 35 or more students in grade 6 for whom parental consent could be sought. Using the Pakistan Government's official 2015 schools list as a sampling frame, we found 56 potentially eligible secondary schools in Hyderabad. We excluded 16 as they had fewer than 40 enrolled students in grade 6 or were campus schools with a single administration responsible for multiple schools in the same area. We planned to include only schools that were more than 1 km away from the nearest eligible boys or girls-only school, to avoid contamination between the study arms. Forty schools met this requirement -20 boysonly schools and 20 girls-only schools. After stratifying by gender, each school had an equal chance of being assigned to either the intervention, or control arm. We issued consent forms to all children enrolled in grade 6. We decided to recruit grade 6 students to maximize our chances of two-year follow-up.

Sampling
Cluster randomized controlled trial software [23] was used to determine minimum sample and cluster sizes required to show statistically significant differences. Power was set at 0.80, alpha at 0.05, the inter-cluster correlation coefficient (ICC) was 0.1 and the effect size was 0.20 [24]. In the absence of local data, the effect size was guided by the findings of an evaluation of a violence prevention intervention among school students in South Africa [25]. We determined that at least 25 students per school and 20 schools per arm would be required, giving 1000 students. We were advised that in Pakistan 30% of students drop out of school between grades 6 to 8. For caution, we assumed up to 40% attrition, and inflated the minimum cluster size to 35 students per school. This yielded a recruitment goal of 1,400 students from 40 schools.

Ethics and data collection
Permission to undertake the study was granted by the Director of School Education-Hyderabad and Principals of the potential study schools. Parental consent forms were issued to 2,486 children of which 75% (1,858) were returned with signed parental consent. Of these, 1,767 children agreed to participate (assented), 71.1% of those initially given consent forms. If parents did not provide written consent, their children did not participate in the study. However, if they were in the intervention school they were still exposed to the intervention. A clinical psychologist was available to accept referrals for any children who had difficulties during the course of the study.
We conducted baseline interviews between November and December 2015, after which the intervention commenced. The 12-month assessment was undertaken between December 2016 and January 2017 and the 24-month assessment was undertaken between November 2017 and February 2018. We declared a participant unavailable for interview after three repeat visits at midline, and five repeat visits at end line. The primary reason for attrition was students dropping out or moving schools. We lacked resources to follow up the missing children.

Intervention
The intervention was delivered to the children by 10 male and 10 female adult coaches employed by the Right To Play NGA. They were later supported by 120 junior leaders. Further information on training and selection is published elsewhere [26] The intervention in Pakistan was based on 103 play-based learning activities each with a specific goal as specified in the manual. Coaches selected an activity for a session from the manual. After the game, they led a three-step discussion following the formula Reflect-Connect-Apply, which involved reflection on the activity and how it made participants feel or what had been learned from it, discussion connecting this to daily life, and application more broadly to other circumstances. Over the 2 years of this study, 120 sessions were conducted in each class, with, on average, two sessions of 35 minutes per week, resulting in 60 sessions in a year. Some of the activities were conducted more than once with the class. The fidelity of the intervention was monitored by Right To Play's staff and the main research partner, Aga Khan University. The research team verified completion of the intended number of sessions and observed a session in each school each month to ensure compliance with the manual.
In addition, the NGO organized summer camps, and also invited parents to sports tournaments and thematic play days. Right To Play also held quarterly awareness sessions with parents on child rights, gender equality and positive discipline and provided some training sessions for teachers (about three per school) on positive child and youth development, positive disciplining, gender equity, and child protection.

Procedures
We conducted a public randomization where a Government School Superintendent drew the names of the intervention schools. Schools and children were not blinded to the study arm.

Instruments
All measures are described in Table 1. The primary outcomes were peer victimization and peer perpetration during the previous four weeks, and symptoms of depression in the previous two weeks with the latter assessed by using Child Depression Inventory II (CDI-2) [20,27]. The intervention was intended to impact upon multiple aspects of children's lives. We hypothesized that a reduction in peer violence would be accompanied by improved child mental health. Neither the children nor the coaches were aware that the main focus was in measuring peer violence.
Secondary outcomes were gender attitudes, roles and norms [22], school corporal punishment [21], experience of physical punishment of at home, witnessing their father or another relative hitting their mother, witnessing their father fighting with other men, self-rated school performance, number of school days missed, and preparations for marriage (Table 1). We assessed the social and demographic characteristics of the children (see Table 1). Selfreported absences were validated against school records.
Instruments were translated into Urdu and Sindhi and self-completed, however a fieldworker read out the questions to a group of four children to facilitate comprehension. We ensured children did not copy answers and reassured them repeatedly that there were no right or wrong answers. Further information on the instruments is given elsewhere [19].

Statistical analysis
We defined study loss to follow-up as missing at both the 12 and 24 months interviews. The final analysis excluded those lost to follow-up. Because the percentage of missing data for items was low, ranging from 0.1% to 3.6%, and mostly with just one missing item, we did not impute missing data. We examined items for randomness in missing data using Little's covariate-dependent missingness (CDM) test. The test showed no association between missing data and baseline covariates.
For all outcomes, except for 'days of school missed', 'depression' and 'preparation for marriage', we derived a score by adding the items. For depression, we derived raw scores for the CDI-2 scale and then converted these to T-scores as described in the CD1-2 technical manual [27]. A student was deemed to have 'preparation for marriage' if they responded yes to either of the two questions. We assessed the percentage change in mean scores for the continuous outcomes, or in the proportion of learners 'promised/ prepared for marriage' from baseline to 24 months' post-baseline.
Cohen's standardized difference was used to compare study outcomes and socio-demographic characteristics at baseline between the two study arms. Logistic regression tested for association between loss to follow up and study arm, socio-demographic characteristics, and baseline levels of all outcomes for both boys and girls. Generalized linear mixed effects Coded as 'never' = no to both items, 'sometimes' = sometimes to one or both, 'often' any response of 'every week' or 'most days' modelling (multi-level model for change) with the Gaussian link function was used to compare mean scores at end-line for all outcomes except for the binary outcome of 'preparation for marriage'. Full information maximum likelihood estimation was used to deal with missing data in study outcomes due to missed interviews at 12 or 24 months postbaseline. For the binary outcome, a generalized linear mixed effects model with a logit link function was used to compare the effect of the intervention between the two study arms. The fixed effects terms included study arm, gender and data collection wave, an interaction term for study arm and data collection wave, and an interaction term for study arm and gender. School was added as the random effects term. The Kenward-Roger method was used to calculate denominator degrees of freedom due to the small number of clusters in each study arm for boys and girls [28]. The effect of the intervention was assessed at 24 months using linear combinations of the fixed effect terms in the model (study arm, gender and data collection wave). We compared boys from the control and intervention arm, and girls from control and intervention arm. Residual plots were computed for each continuous outcome to check for distribution assumptions. For each outcome, we assessed standardized residuals from the model without the random component as well as from the model with random components.
As a sensitivity analysis, we used generalized linear mixed effects modelling to perform a cluster-level analysis where all outcomes were aggregated at school level (mean scores for continuous outcomes and proportions for binary outcomes). The cluster-level model was adjusted for baseline outcomes aggregated at cluster level, and we used the Kenward-Roger method. Analyses were conducted in Stata 14 with statistical significance set at 5%.

Trial profile
The trial profile is shown in Figure 1. A total of 446 boys and 483 girls from the intervention arm and 376 boys and 447 girls from the control arm were enrolled in the trial. The 12 months follow-up rate was 86.5% (1515/1752) and at 24 months, it was 84.5% (1480/1752). Of the 1752 learners enrolled into the study, 1397(79.7%) were interviewed at all the three data collection points, 118 (6.7%) were available at baseline and 12 months only, 83 (4.7%) were available at baseline and the 24 months only. No schools (clusters) were lost to follow up. Loss to follow up of participants is shown in Figure 1. Reasons for loss were children having transferred to other schools, long absence from schools, and for boys, having left schools for employment, and for girls, for household chores. Two girls had married. There were no reports of serious adverse events. Table 2 shows the participants' baseline characteristics. Most participants were aged 12-14 years and many of the participants were not food secure, sometimes or often going without food. Between 20% and 25% had repeated an academic grade. Many were exposed to violence at home and in the community including witnessing their father fighting with another man and their mother being beaten by their father or another relative in the past month.

Background characteristics
There were no significant differences in sociodemographic characteristics between intervention and control arms for both boys and girls and no significant differences at baseline in any of the study outcomes among boys. However, girls in the intervention arm had slightly higher mean scores for peer violence victimization, depression or corporal punishment than control arm girls (effect size≥0.2). Table 3 shows the results of analysis of the effect of the intervention on the primary outcomes peer violence victimization, peer violence perpetration and depression for boys and girls at 24 months. After adjusting for the baseline level of each variable, there were statistically significant differences in mean scores at end-line between control and intervention arms. For boys, the percentage reduction in the mean peer-victimization score was 27.8% in the control versus 33.3% in the intervention arm. For girls, the percentage reduction in this measure was 21.3% in the control versus 58.5% in the intervention arm. For boys, the percentage reduction in the mean peer perpetration score was 11.1% in the control versus 25.3% in the intervention arm. For girls, the percentage reduction in this was 27.6% in the control versus 55.6% in the intervention arm. The mean depression score in boys dropped by 4.8% in the control versus 7.2% in the intervention arm and in girls it dropped by 5.6% in the control versus 9.5% in the intervention arm. Table 4 shows the changes observed in the secondary outcomes. Gender attitudes changed significantly for boys and girls, becoming less patriarchal in the intervention arm than the control arm, although the reduction in scores was greater for girls than boys. Corporal punishment was reported significantly less often during the last four weeks, by both boys and girls in the intervention arm compared to the control arm, with the reduction greater for girls. Physical punishment at home during the last four weeks was reported significantly less often in the intervention arm, with a greater reduction for girls than boys. The difference in the proportion promised in marriage for boys and girls showed possible evidence in change in the desired direction, but was not statistically significant.

Secondary outcomes
Two of the secondary outcomes did not show evidence of desired change: the self-assessed school performance score and the number of days of school missed in the previous four weeks. The latter was higher at 24 months for both boys and girls than at baseline. For boys, it was 77% higher in the control arm versus 55% higher in the intervention arm, and for girls, it was 36.7% in the control arm and 43.8% higher for girls in the intervention arm. These differences were not statistically significant. Table 5 shows an analysis of three exploratory outcomes: experiences of hunger, witnessing of fighting between the child's father and another man, and witnessing the child's mother being beaten at home. For both boys and girls, these outcomes were reported significantly less often reported among children in the intervention arm compared to the control arm at end line. The results of the sensitivity analysis performed at cluster level were consistent with the individual-level analysis above.

Discussion
We have shown that Right to Play's Positive Child and Youth Development Programme in Pakistan, delivered under our study conditions, enabled a significant reduction in peer violence victimization, and perpetration and depression among boys and girls in the intervention schools compared to the control schools. Girls reported a greater decline in violence outcomes than boys. The programme showed positive impact on gender attitudes, corporal punishment, and child physical punishment at home, again with greater changes reported by girls. The reduction in depression adds confidence to the interpretation of these findings, as depression is a wellestablished consequence of peer violence and the complexity of the CDI-II measure enhances response validity.
This research has advanced the small body of evidence on the prevention of peer violence victimization and perpetration in low-and middle-income countries in general, and Pakistan in particular. It is the first evaluation of an intervention here to show positive impact. Globally, this is the first evaluation of the impact of a play-based life-skills intervention on peer violence among school children. The findings show that the intervention is effective in preventing violence victimization and perpetration amongst girls as well as boys in and out of school settings, as well as some other forms of out of school violence. This intervention is particularly important because evidence shows that school violence has far-reaching effects and that different forms of violence are interconnected [29]. For example, peer violence  perpetrators are more likely to engage in dating and partner violence [30]. Using schools as an entry point for reducing violence overall is particularly important in a country like Pakistan which has relatively high levels of culturally sanctioned violence and where oppression of women and girls is widespread.
We found greater violence reduction reported by girls than boys, and this may be because teenage girls are more mature than boys and so may benefit more from an intervention that requires discussion and processing of information. It is also possible that boys' violence is harder to prevent. Peer violence in Pakistan and many other countries is more socially acceptable behaviour for boys than girls, and consequently, there is a higher prevalence among boys, and this may explain why it is harder to change their behaviour [31].
We did not show an impact on self-reported school performance measures nor on child-reported days missed from school. We recognize that the main drivers of days missed (child labour and help with domestic chores) were not changed through this intervention and the measures were self-reported. We were pleased to see that the children did not give socially desirable answers to these questions, and this further enhances our trust in the other reported outcomes. The changes in peer violence and depression were recorded in both intervention and control arms, as is common in randomized-controlled trials. This may be associated with age-related changes in peer violence and associated depression, or be an artefact of the repeated measurements in the trial [32].

Strengths and limitations
The study strengths included its large sample, drawn from poorer schools in a major city in Pakistan, and the fact no schools dropped out. Our results may be generalizable to similar settings in Pakistan. The likelihood of replicability of the results in another setting is enhanced by the long-established approach to work to Right To Play. The study was well powered, even after taking into account changes in the control arm, retention was high and there was little missing data. The evaluation was conducted by an independent research team and the involvement of Right To Play staff was limited to delivering the intervention. The coaches were not informed of the primary outcome measures.
A key limitation was that the outcome measures were self-reported and we had to assume that the 'last four weeks' was typical, but reporting biases should have been similar between the two study arms. Our confidence in our results was enhanced by the concurrent findings of reduction in depression, and the fact that we did not see positive changes in all of our measures.
The evaluation had features of both an effectiveness and efficacy trial. We did not consider individual-level dose in the analysis and did not influence intervention attendance. In this evaluation the coaches were trained for a little longer than was usual due to a research-related delays in starting the intervention, and the research team's oversight of the sessions delivered may have strengthened adherence to the manual. Right To Play has now adopted these changes into the normal practice.

Conclusion
We found the play-based life-skills intervention of Right To Play, when delivered over 2 years to children in grades 6-8 in Hyderabad, Pakistan, reduced peer violence in school and family violence in the home. Play-based life-skills interventions thus constitute a new dimension in the repertoire of effective approaches to prevent peer violence and should be evaluated in other settings. Further research is needed to determine whether the impact is sustained beyond the 2 years of a study and whether positive impact persists if the intervention is taken to scale. This study increases understanding of the types of interventions that are effective in reducing peer violence in low-and middle-income countries. This study has broadened our understanding of the types of interventions that are effective in reducing peer violence in low-and middle-income countries.

Disclosure statement
No potential conflict of interest was reported by the authors.

Ethics and consent
The study complied with all ethical standards required to conduct randomized controlled trial. The study was approved by Aga Khan University's Ethics Review Committee (Protocol Number 3705) and South African Medical Research Council (Protocol Number EC025-8-2015). Informed consent was obtained from parents and assent from all children included in the study.

Funding information
This is an output from the What Works to Prevent Violence: A Global Programme, funded by the UK Aid from the UK Department for International Development (DFID) for the benefit of developing countries, managed by the South African Medical Research Council;Department for International Development, UK Government.

Paper context
Systematic reviews have observed that there has been little research evaluating interventions to prevent peer violence in low-and middle-income settings. This paper describes an evaluation of an intervention in Pakistan which was found to be effective in preventing peer violence perpetration and victimization. This is the first time a rigorous evaluation of a structured play-based life-skills programme has been conducted and shown impact on peer violence prevention. It is the first time any peer violence prevention intervention in a low-and middle-income country setting has been evaluated and shown to be impactful. Pakistan is a country in which violence is highly prevalent and so it is important that interventions are shown to work in this context. The study represents an important advance in peer violence prevention globally. Research is encouraged in other regions.