(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



The EmpaTeach intervention for reducing physical violence from teachers to students in Nyarugusu Refugee Camp: A cluster-randomised controlled trial

['Camilla Fabbri', 'London School Of Hygiene', 'Tropical Medicine', 'London', 'United Kingdom', 'Katherine Rodrigues', 'International Rescue Committee', 'New York', 'United States Of America', 'Baptiste Leurent']

Date: 2021-10

There was no evidence that the EmpaTeach intervention effectively reduced physical violence from teachers towards primary or secondary school students in Nyarugusu Refugee Camp. Further research is needed to develop and test interventions to prevent teacher violence in humanitarian settings.

We conducted a 2-arm cluster-randomised controlled trial with parallel assignment. A complete sample of all 27 primary and secondary schools in Nyarugusu Refugee Camp were approached and agreed to participate in the study. Eligible students and teachers participated in cross-sectional baseline, midline, and endline surveys in November/December 2018, May/June 2019, and January/February 2020, respectively. Fourteen schools were randomly assigned to receive a violence prevention intervention targeted at teachers implemented in January–March 2019; 13 formed a wait-list control group. The EmpaTeach intervention used empathy-building exercises and group work to equip teachers with self-regulation, alternative discipline techniques, and classroom management strategies. Allocation was not concealed due to the nature of the intervention. The primary outcome was students’ self-reported experience of physical violence from teachers, assessed at midline using a modified version of the ISPCAN Child Abuse Screening Tool–Child Institutional. Secondary outcomes included student reports of emotional violence, depressive symptoms, and school attendance. Analyses were by intention to treat, using generalised estimating equations adjusted for stratification factors. No schools left the study. In total, 1,493 of the 1,866 (80%) randomly sampled students approached for participation took part in the baseline survey; at baseline 54.1% of students reported past-week physical violence from school staff. In total, 1,619 of 1,978 students (81.9%) took part in the midline survey, and 1,617 of 2,032 students (79.6%) participated at endline. Prevalence of past-week violence at midline was not statistically different in intervention (408 of 839 students, 48.6%) and control schools (412 of 777 students, 53.0%; risk ratio = 0.91, 95% CI 0.80 to 1.02, p = 0.106). No effect was detected on secondary outcomes. A camp-wide educational policy change during intervention implementation resulted in 14.7% of teachers in the intervention arm receiving a compressed version of the intervention, but exploratory analyses showed no difference in our primary outcome by school-level adherence to the intervention. Main study limitations included the small number of schools in the camp, which limited statistical power to detect small differences between intervention and control groups. We also did not assess the test–retest reliability of our outcome measures, and interviewers were unmasked to intervention allocation.

School-based violence prevention interventions offer enormous potential to reduce children’s experience of violence perpetrated by teachers, but few have been rigorously evaluated globally and, to the best of our knowledge, none in humanitarian settings. We tested whether the EmpaTeach intervention could reduce physical violence from teachers to students in Nyarugusu Refugee Camp, Tanzania.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: the EmpaTeach intervention was developed by KR at the International Rescue Committee and MK, GTE, and ADF at the Behavioral Insights Team. They have provided input into study design but final decisions on study design and procedures were taken by KD.

Funding: The research was supported by funding from an anonymous donor to the International Rescue Committee and by a research grant by MRC/DfID/NIHR MR/S023860/1 to KD at the London School of Hygiene and Tropical Medicine. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: Fully anonymised data are available on request from the LSHTM Data Repository ( https://doi.org/10.17037/DATA.00002474 ) for researchers who meet the criteria for data access and whose intended analyses fall under the scope of the PVAC study. The LSHTM Research Data Manager, based in the Library & Archives Service, is responsible for managing data in the repository and can be contacted at [email protected] .

Copyright: © 2021 Fabbri et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

In this study, we aimed to determine whether a short peer-led intervention focused on reducing impulsive violence—known as EmpaTeach—reduced physical violence from school staff to students in Nyarugusu Refugee Camp, Tanzania, using a cluster-randomised controlled trial. EmpaTeach takes a psychological approach, aiming to improve teachers’ self-efficacy, self-regulation, and empathy for students, and to reduce teachers’ stress levels. It provides teachers with information about alternative disciplinary methods and positive classroom management strategies, and creates social support for skill development. The intervention is implemented by trained peer teachers over a 10-week period, and is thus feasible to deliver in highly resource-constrained contexts, including emergency settings.

While corporal punishment often occurs as a normalised behaviour, it can also occur impulsively rather than instrumentally [ 23 , 24 ] and can be exacerbated by perceived stress [ 25 ]. Stress and other life events that generate negative affect and emotional distress can be triggers of violent behaviours and aggression [ 26 , 27 ]. In contexts such as humanitarian crises, characterised by high levels of stress, overcrowding, and severe resource constraints, it is plausible that teachers’ psychological well-being may be compromised and their emotion regulation weakened, resulting in increased risk of impulsive violent behaviours. To date, no specific interventions focusing on reducing teachers’ impulsive use of violence have been tested, but correlational studies suggest that integrating content on self-regulation and coping, along with positive discipline methodologies, into teacher training could be an effective mechanism to reduce violent discipline [ 25 ].

Schools can be a main site of exposure to violence, from peers and from school staff [ 11 – 14 ], but also offer tremendous opportunities for primary prevention of violence [ 15 ]. Particularly for displaced children, schools may offer some sense of normality in very challenging conditions and provide an opportunity to build aspirations and recover from trauma [ 16 ]. WHO recommends school-based approaches to improve a range of child and adolescent health outcomes [ 17 , 18 ], but only a handful of evaluations of programmes to prevent violence from teachers to students have been conducted globally [ 19 , 20 ] and, to the best of our knowledge, none in humanitarian settings. Two interventions have been rigorously trialled, and both were proven effective in reducing violence from school staff to students [ 21 , 22 ]; however, both programmes were implemented in low-resource but stable contexts in Uganda and Jamaica. One is a whole-school approach focused on creating a conducive operational culture in schools and encouraging behaviour change for teachers, students, and school administration over 18 months [ 22 ]. The other is focused on teacher skill development and behaviour change over 8 months, with a supported process for teachers to learn positive reinforcement, classroom management strategies, and proactive approaches to prevent child misbehaviours [ 21 ].

More than 1 billion children experience physical, sexual, or emotional violence each year [ 1 ]. For the 33 million children estimated to be forcibly displaced by war and conflict globally as of 2019 [ 2 ], levels of violence are likely to be even higher, as several features of humanitarian crises are known risk factors [ 3 ]. These include widespread economic and social insecurity, weakened social ties and alterations to traditional household composition, threats to basic livelihoods and individual freedoms, and the breakdown of protective systems and service provision [ 4 ]. In Nyarugusu Refugee Camp, where this study takes place, violence in schools and on the way to and from schools is perceived as prevalent [ 5 ]. Norms that promote the use of violence as a way of disciplining students and supporting student learning are widespread in the camp [ 6 ], and although the majority of students report feeling safe in school, the ‘stick’ and other forms of corporal punishment are frequently used to manage large classrooms [ 5 ]. Violence in childhood and adolescence is associated with future depression, suicide attempts, violence victimisation and perpetration, and poor educational outcomes [ 7 – 10 ], and prevention of violence is a focus of Sustainable Development Goals 5, 8, and 16.

Methods

The Preventing Violence Against Children in Schools (PVACS) study consists of a cluster-randomised controlled trial, a parallel qualitative study, an economic evaluation, and a process evaluation. The study was approved by the London School of Hygiene & Tropical Medicine Ethics Committee (ref. 16000) and the Tanzania National Institute for Medical Research (ref. NIMR/HQ/R.8a/Vol.IX/2920). Our protocol is registered at ClinicalTrials.gov (NCT03745573) and is published elsewhere [28]. This paper describes our main trial results and is reported as per the Consolidated Standards of Reporting Trials (CONSORT) guideline (S1 CONSORT Checklist).

Setting Nyarugusu Refugee Camp was formed in 1996. It hosts about 80,000 Congolese refugees, some of whom have been there since the camp’s establishment, and about 60,000 Burundian refugees, who have been arriving since 2015. The camp is operated by the United Nations High Commission for Refugees and the Tanzania Ministry of Home Affairs, and the International Rescue Committee (IRC) provides all education and gender-based violence response services and, at the time of the trial, provided all child protection services. Children attend 1 of 27 schools in the camp, which teach either a Burundian or Congolese curriculum. A 2018 survey across Nyarugusu and 2 smaller camps in the Kigoma region showed only 56% of school-aged children were enrolled. However, 78% of Congolese children in Nyarugusu attended school, owing to the protracted displacement of the Congolese population [5]. Schools in the camp face numerous challenges, including poor teacher attendance and shortages of qualified teachers. In Nyarugusu, in 2017–2018, there was a teacher to pupil ratio of up to 1:200 in primary schools; teachers received low pay and often needed to engage in other income-generating activities [5].

Study design and participants We conducted a 2-arm superiority cluster-randomised controlled trial with parallel assignment between November 2018 and February 2021. A cluster design was appropriate because the intervention was delivered to all teachers within a given school. A complete sample of all 27 primary and secondary schools in the camp was invited to participate All schools agreed to participate and were randomised to the intervention or to act as controls. Allowing for a loss to follow-up of 2 schools per arm, assuming 50 interviews with students per school, a prevalence of past-week physical violence of 50%, and an intra-cluster correlation coefficient of 0.10 [22], we had 80% power to detect a 19% difference in the prevalence of our primary outcome between the intervention and control arms with a 2-sided significance level of 5%. Three rounds of cross-sectional surveys were conducted in schools: baseline (from 19 November 2018 to 8 December 2018), midline (from 15 May 2019 to 7 June 2019, 2 months after the end of the intervention), and endline (from 23 January 2020 to 21 February 2020, 10 months after the end of the intervention). A cross-sectional design was selected rather than a cohort design to avoid issues related to attrition of individual students over time, as there was some likelihood of repatriation of Burundian refugees over the course of the study. Headteachers provided consent for data collection and intervention implementation in schools, and consent to approach individual students. Students provided informed assent, and individual teachers provided informed consent. Parental consent for children’s participation in this study was not required by relevant ethics committees. In initial consultations with camp stakeholders, it was felt that parents in this setting would prefer headteachers to consent for children in school as they had full responsibility for students during school hours, and that seeking active parental consent would preclude large numbers of children from participating, given high numbers of unaccompanied youth in the camp. Up-to-date lists of all teachers and students in grade 2 and above were obtained from IRC’s education team. These lists formed the sampling frame for the study, and a stratified random sample of 60 students per school was selected to ensure representation from different age groups. All students who could speak Kiswahili or Kirundi and who were deemed by interviewers to understand the consent procedures were eligible. We collected survey data from students in grade 2 or above and aged 9 years and over, as they were able to respond to questions in survey format in our pre-testing. A simple random sample of school staff was invited to provide informed consent for their participation in survey data collection. At least 2 repeat visits were made to find students and staff who were sampled but absent on the day of the survey. The sample is thus intended to be representative of teachers and of students in grade 2 or above and age 9 years and over attending school in the camp. Survey data were collected by interviewers recruited from the camp population, who received 2 weeks of intensive training before each data collection round. Interviewers read questions and response options to participants from tablet computers. Algorithms were in place to prevent accidental skipping of questions, and referrals for those who disclosed abuse (see below) were automatically triggered based on survey responses. Backchecks and data quality audits were conducted on 10% of the sample at each round of data collection by experienced interviewers who acted independently from the core field team and who reported directly to the local field coordinator. Backchecks were used to assess enumerator performance, identify errors in survey programming, and record discrepancies to address during data cleaning.

Randomisation and masking To ensure balance across arms, schools were stratified according to whether they served a Congolese or Burundian population, and were primary or secondary schools. An allocation list was generated by EA with a computer random number generator and an algorithm in Stata (version 15 [29]). Allocation took place at a public meeting where a representative of each school within each stratum was invited to place the name of their school in an opaque bag. A nominated person from each stratum then withdrew names from the bag, and schools were allocated either to receive the intervention or to the control condition in the sequence on the allocation list, recorded by MZ. The intervention is behavioural in nature, and it was not possible to mask participants to allocation. Similarly, allocation was not discussed with interviewers tasked with collecting survey data, but it is reasonable to assume that they could determine allocation. The statistician (BL) was masked to allocation when performing the main trial analyses. Some secondary analyses (e.g., adherence) were performed after unmasking. EA, MZ, and BL are authors and members of the PVACS research team.

Intervention Content. The EmpaTeach intervention is a behaviourally informed, self-guided teacher training intervention designed to reduce and prevent teachers’ use of corporal punishment in the classroom [30]. The content of the intervention is focused on empathy-building exercises and on group work to learn and practice self-regulation techniques, strategies to promote well-being, positive disciplinary methods, and classroom management strategies. To create intervention materials and exercises around empathy building, alternative discipline, and other ingredients, the intervention designers drew on existing programmes and also created new content [31–33]. Box 1 describes implementation in the trial [34]. The intervention theory of change is illustrated in S1 Fig. Box 1. Description of the EmpaTeach intervention Overview The project targets teachers’ attitudes, beliefs, and behaviours. Teachers go through self-guided group sessions inspired by cognitive behavioural therapy—an approach that has been effectively applied to many problems, from reducing destructive behaviours to improving well-being—to help teachers challenge thinking and patterns of behaviour related to using violence as a form of punishment. Content and materials The intervention was based on a booklet developed specifically to self-guide teachers through each of the 12 sessions in the programme (developed by the Behavioural Insights Team [BIT] and IRC in English, and translated into Kiswahili and Kirundi). The booklet contained learning materials for all sessions and space to commit to practicing new strategies and record reflections on their homework assignments, which primarily consisted of home and classroom practice of the intervention techniques. For 6 of the sessions, there were accompanying videos that were produced locally as part of the intervention. Two of the sessions involved playing an interactive game with playing cards to help teachers apply learned concepts. Specifically, throughout the intervention, teachers engaged in a series of value-affirmation and empathy-building exercises based on self-reflection, and they received information about alternative disciplinary techniques (including de-escalation strategies and techniques to reward positive behaviours) and emotional regulation tools inspired by cognitive behavioural therapy to identify triggers for impulsive reactions. The intervention supported teachers in creating action plans for change for responding to students’ positive and negative behaviours and reflecting on future actions when they would encounter problems. Finally, the intervention generated social support through the group setting so that teachers could count of peers for support and advice throughout the change process. The intervention material was designed to be suitable for application with children of different age groups. The description of each classroom strategy illustrated in the booklet included adaptations for older children, and gender considerations. A shared tablet computer was required for each group to view the videos during the sessions. Session 5, which involved learning how to co-create classroom rules with students, required a large piece of posterboard paper, a marker, and tape or glue to affix the paper to the wall. All teachers were served lunch during the introductory meeting and the programme ending party; however, neither teachers nor group coordinators received any form of payment for their participation in the intervention. Guides were provided for the coordinators of the sessions, which contained the agenda and content for each meeting. Procedures A BIT programme developer, IRC education technical unit staff, and local refugee incentive workers provided a 3-day training to 85 teachers who were nominated as group coordinators by their peers (as well as a separate 1-day training with school headteachers and discipline teachers). Day 1 was focused on facilitation skill building, and beginning to review the facilitator roles and responsibilities and the programmatic themes. On day 2, facilitators finished reviewing the programme themes and roles and responsibilities. On day 3, facilitators were split into smaller groups and practiced facilitation of sessions. The group facilitators were trained in 3 separate groups: 2 groups of Congolese facilitators and 1 group of Burundian facilitators. During the training itself, teachers participated in smaller group work activities. There were 2 trainers (1 English speaker and 1 native Kiswahili or Kirundi speaker) at each of the group facilitator trainings. As such, the trainer:teacher ratio was 2:27 to 2:30, depending on the group. These training sessions provided participants with an overview of the intervention and their roles and responsibilities, built group facilitation skills, provided a space for practicing facilitation skills and learning from other participants, prepared facilitators for potential challenges that could arise during the intervention, and built their buy-in. The coordinator training contained a review of the main lessons from each programme module, facilitator roles and responsibilities, and facilitation and organisation skills, including practice sessions facilitating mock sessions. Group coordinators did not receive additional facilitation support during the intervention implementation period, as the curriculum is intended to be self-guided. The selection of teacher facilitators followed a staged approach: First, teachers were asked to nominate peers they admired or learned from. Second, nominated teachers completed a survey about their attitudes towards the use of corporal punishment. Teachers who expressed supportive views towards the use of harsh discipline were removed from the potential facilitator cohort. Third, among the remaining teachers, those who were most proficient in facilitation skills were selected to be facilitators, and the remaining teachers became ‘back-up facilitators’. Each group coordinator facilitated the programme in a face-to-face fashion with a group of 3–15 teachers. The first 4 sessions were condensed into two 4-hour sessions delivered over 2 weekend days. (This resulted in there being 12 sessions in total, instead of the 14 sessions specified in the PVACS study protocol.) The remainder of the sessions were held weekly until the end of the programme, and lasted about 1–1.5 hours each—with the exception of weeks 5 and 11, when groups met a second time during the week to further engage with the techniques they had learned by playing an interactive learning game. The teachers did homework each week, taking about 30 minutes. They also received 2 SMS texts per week from their group facilitators to reinforce aspects of the group sessions or homework. Each group self-determined when and where it met each week. Each session started with a review of the previous week’s session, reflection on key concepts, and sharing of homework, including any challenges encountered. This was followed by an introduction of a short slogan capturing the main learning of the session. Then participants engaged in a series of stories that illustrated a hypothetical but common classroom situation and reflection activities, and presentation and discussion of simple classroom management and self-regulation activities they could use, followed by homework that allowed real-world practice of new techniques in teachers’ own classrooms. Modifications During programme implementation, some group coordinators expressed being unclear about how to facilitate the card games with teachers. In light of this, IRC staff offered a 1-day refresher training on the teacher card games specifically. A number of teachers were made redundant (i.e., laid off) during the implementation of the intervention and were subsequently re-hired within weeks. This disrupted implementation of EmpaTeach across a number of teacher groups. As a result, IRC’s programme team had to re-assemble some of the intervention groups and make plans to allow for the implementation of the intervention within a shorter period of time. Nine out of 77 groups received a compressed version of EmpaTeach over a 6-week period (instead of 10 weeks). Intervention development and pilot trial. The intervention design was informed by qualitative interviews and focus group discussions with teachers, school administrative staff, students, parents, and humanitarian frontline workers in the camp, and developed via iterative co-creation sessions with teachers. The intervention was intentionally designed to be short, with limited ongoing support for implementation, and focused primarily on teachers so that it could be implemented in humanitarian contexts with highly mobile populations and significant resource constraints. Teachers were selected mainly because they were seen as credible, relatable messengers with relevant experience, helping to dispel possible perceptions among teachers that the intervention was being externally imposed. High levels of stress were reported by teachers in qualitative interviews conducted during the formative research stage, resulting from overcrowded classrooms and poor resources, as well as the teachers’ own experiences as forcibly displaced persons. The designers thus included techniques inspired by cognitive behavioural therapy focused on enhancing teacher well-being, empathy, and emotional self-regulation, which the designers hypothesized would lead to a reduction of stress-induced or impulsive uses of corporal punishment. A rapid pilot trial aimed at identifying effective behavioural strategies to shift attitudes was conducted in 2018; results showed that an empathy-building approach was the most promising in changing teachers’ attitudes about violence, compared to rights-based and information-based approaches [35]. The designers also hypothesized that teachers’ lack of training in alternative disciplinary techniques would be a barrier to reducing use of violent discipline. The initial third of the EmpaTeach curriculum was therefore dedicated to teaching and practicing positive discipline strategies and classroom management skills. This content was co-created and validated with refugee teachers through a series of iterative sessions where teachers provided feedback on the clarity of exercises, examples of local positive discipline strategies, and scenarios and stories relevant to the camp cultural context. During co-creation sessions, teachers became defensive of, and expressed support for, use of corporal punishment and violent discipline when the topic was broached. The designers thus determined that directly addressing norms on the acceptability of violence would have required highly trained facilitators to avoid inadvertent reinforcement of norms. Such facilitators are not available in the camp setting. Therefore, the intervention did not directly discourage the use of physical violence for classroom management, but instead encouraged teachers to adopt new and more effective behavioural strategies. In 2018, a controlled before-and-after pilot study was conducted, where the whole intervention was implemented in 4 schools in a nearby refugee camp. The pilot study suggested that the intervention was acceptable to participants and feasible to deliver, and preliminary estimates indicated that the direction of effect would be towards a reduction in teachers’ use of physical violence [30]. Delivery. All teachers in intervention schools were invited to participate in EmpaTeach, and were assigned to a group composed of other teachers from their school. While participation was not mandatory, it was strongly encouraged and communicated as expected by school administrators. All students in each school thus had the potential to be exposed to the intervention. Groups met 12 times, for two 4-hour sessions and ten 1- to 1.5-hour sessions, which were led by peer teachers (group facilitators) guided by a tailored booklet featuring stories, lessons, and exercises. Teachers received information on positive discipline, as well as planning exercises and reinforcement SMS texts. Given that the intervention was implemented in a group setting, teachers also received social support from peers to change their behaviours. Teachers had the opportunity to discuss their experiences and challenges in the group sessions. The groups met over a 10-week period as originally intended (‘original version’) or over a 6-week period (‘compressed version’). Reasons for this are described in the Results. Schools in the control arm did not receive any specific intervention related to violence prevention during the study.

Harms The intervention is behavioural in nature, and we anticipated that this would result in minimal risk for participants. We did speculate that for teachers who had experienced recent trauma, some of the empathy and reflection exercises may have resulted in increased emotional distress; however, we did not record increased levels of distress among teachers in the intervention group. All teachers were given information about where they could seek psychological support. At endline, IRC’s programme team installed a letterbox in a central location within the camp to allow for anonymous complaints and feedback on the study procedures. All teachers were informed about the existence of the letterbox during consent procedures, but no complaints were reported. We did not expect any adverse effects of the intervention among children in intervention schools, but anticipated that during survey data collection, children would disclose experiences of abuse that would necessitate referral to child protective services. During intervention implementation, IRC’s child protection team made regular visits to study schools as part of their routine activities and offered routine support and safeguarding services to students. During research activities, children were informed during the consent process that we would need to pass their details on to child protection officers if they disclosed information that made the research team think that ‘their safety or well-being might have been at risk or that they had been hurt’. The student survey contained a number of algorithms based on students’ age and responses to survey questions that automatically generated different interview finishes that prompted enumerators to follow recommended referral pathways. Specific disclosures about forms and time frames of abuse that would necessitate referrals were based on predefined criteria agreed on with IRC’s child protection team. Teachers were told that information they disclosed during the surveys might be reported to child protection officers if the research team thought that ‘a child’s safety or welfare might be at risk’. All participants were offered counselling regardless of what was disclosed during the survey. Adults were offered referral to IRC services.

Process data collected from intervention schools During the intervention period, teachers’ attendance at each EmpaTeach session and whether they completed intervention homework was recorded by their group facilitator. Tracking sheets were delivered to headteachers and collected on a weekly basis by IRC’s programme team. IRC conducted a series of teacher classroom observations and random spot checks that, together with our survey and qualitative data, will inform a separate process evaluation that will explore the intervention’s mechanisms of action. Movement of individual teachers across schools was recorded centrally by IRC’s education team; the team was also alert to movement of groups of students, but individual student movement across schools was not tracked during the study.

Outcomes The primary outcome was students’ self-reports of experience of past-week physical violence from school staff at midline, measured using an adapted version of the ISPCAN Child Abuse Screening Tool–Child Institutional (ICAST-CI) [36]. Secondary outcomes were students’ self-reported experience of past-week physical violence from school staff at endline, students’ self-reported experience of past-week emotional violence from school staff at midline and endline (measured using the ICAST-CI), students’ self-reported depressive symptoms (Mood and Feelings Questionnaire [MFQ] [37] score of 12 or above) at midline and endline, and students’ self-reported past-week school attendance at midline and endline. Students were categorised as having experienced physical violence from school staff in the past week if they had experienced at least 1 out of 23 different violent behaviours in the past week. The depressive symptom scale had a good internal consistency both at midline (Cronbach’s alpha = 0.86) and endline (Cronbach’ alpha = 0.88). The exact wording of the questions used for the creation of the individual-level outcomes is provided in S1 Table. We constructed 7 intermediate outcome variables to assess teachers’ adoption of the self-regulation and positive discipline strategies that were described in the EmpaTeach booklet, and teachers’ self-control, attitudes toward corporal punishment, and job satisfaction. Detail on the construction of these variables is included in S1 Table. All instruments have been widely used and validated in international settings. Items were translated to Kiswahili and Kirundi, cognitively tested with a sample of 5 students and 5 teachers from the same camp population, and adapted where necessary, prior to the baseline survey. Cognitive testing participants were then excluded from the trial sampling frame.

Statistical analysis We performed an intention-to-treat analysis with data from our cross-sectional surveys. All analyses were done in Stata/IC 16 [29]. Analyses were done with individual student data, accounting for clustering of students within schools using generalised estimating equations (GEEs) with an exchangeable correlation structure, and robust standard errors. For binary outcomes, GEEs with a log link were fitted to estimate risk ratios (RRs), and for continuous outcomes, a Gaussian link was used. Adjusted estimates controlled for the stratification variables and also incorporated the cluster-level mean of the outcome at baseline for continuous outcomes. We also explored heterogeneity of the treatment effect by testing for an interaction with pre-specified subgroups (sex, country of origin, school level, functional difficulty, and baseline level of past-week physical violence), and conducted sensitivity analyses for missing data. Finally, in the schools that received the intervention, we explored the association between the primary outcome and school-level adherence to the intervention, defined as the proportion of EmpaTeach groups where average attendance to sessions was at least 80%. Results were considered statistically significant at the 2-sided 5% level. No formal adjustment for multiple comparisons was performed, but a limited number of pre-specified tests were conducted. Further details of the study and analysis methods are reported in the statistical analysis plan [38]. Following suggestions from reviewers, we conducted additional post hoc analyses including exploratory analyses of intermediate teacher outcomes that we hypothesized would be on the pathway to impact for the main trial outcome, analyses of the intervention’s dose and quality using monitoring data from IRC, and analyses of emotional violence that adjusted for the school-level mean of the outcome at baseline.

[END]

[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003808

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/