Preventing violence against children in schools (PVACS): protocol for a cluster randomised controlled trial of the EmpaTeach behavioural intervention in Nyarugusu refugee camp

The primary aim of this project is to test the effectiveness of the Empateach intervention to prevent physical violence from teachers to students in Nyarugusu Refugee Camp, Tanzania. Our secondary objectives are to assess the impact of the EmpaTeach intervention on students' depressive symptoms, experience of emotional violence and educational test scores. We also plan to conduct a parallel qualitative study, process and economic evaluations.

We will conduct a two-arm cluster RCT with parallel assignment.

Our collaboration currently involves five partners, convened by the International Rescue Committee (IRC), an international non-governmental organization (NGO) delivering services in conflict-affected settings. Since 2016, IRC has been developing and piloting the EmpaTeach intervention in conjunction with the Behavioural Insights Team (BIT), a team of UK and USA based behavioural scientists. IRC implements the EmpaTeach intervention in this study. Innovations for Poverty Action (IPA) is the quantitative data collection partner for the study and collaborates on a range of projects with IRC. IPA is an international organisation which facilitates data collection for large scale quantitative research projects via its local country offices. IRC has been working with IPA to collect data in relation to its projects in Africa for over 5 years.

In July 2018, the London School of Hygiene and Tropical Medicine (LSHTM) was invited to join the collaboration to conduct an independent evaluation of the EmpaTeach intervention. LSHTM leads the trial and quantitative aspects of the process and economic evaluations; the Tanzania National Institute for Medical Research leads the qualitative aspects of the research.

Nyarugusu refugee camp in Kigoma, Tanzania, was established in 1996 to host refugees fleeing conflict in Democratic Republic of Congo and is one of three refugee camps in Kigoma region. According to the latest UNHCR data, the camp hosts around 150,000 refugees, of whom 80,000 are Congolese refugees who have been in the camp since its opening and around 70,000 Burundian nationals who have found home in the camp after the conflict in Burundi broke out in April 2015. Many Burundians were displaced for longer periods previously in Tanzania due to recurring inter-ethnic violence.

The Camp is made up of 146 villages, spread across 14 zones. Zones are largely divided by population origin: zones 1–7 and 14 host Congolese refugees and zones 8–13 host Burundian refugees. The population is very young, with 58% of residents under 18 and only 3% of adults older than 60 years of age. The two most common ethnic groups in 2017 were Hutu (47.8%) and Bembe (42.8%) and most refugees in the camp speak Kiswahili or Kirundi. Refugees are not allowed to engage in formal employment while in Tanzania according to the Refugees Act, and the government enforces restrictions on the movement of refugees once they are hosted in a designated area. Therefore, the Camp population largely relies on food and commodities distribution by NGOs and United Nations agencies active in the camp.

IRC is responsible for all education activities in the camp and for protection and gender-based violence support services. Schools serve Congolese or Burundian students; children in the camp learn the curriculum of their country of origin. The language of instruction is French for the Congolese refugees, while Kirundi is used as the language of instruction until Grade 5 for the Burundian refugees, and after Grade 5, French becomes the medium of instruction. Kiswahili and English, which are the languages in the host country and in the region, are delivered as subjects. According to a recent joint education needs assessment conducted across the three refugee camps in Kigoma region, 21% of Burundian refugee children attend pre-school, 78% are in primary school and only 3% attend secondary school. Among Congolese refugee children, 45% attend pre-school, 98% attend primary school and 60% attend secondary school. The children most likely to be out of school within the refugee community are children from very poor families, orphans, unaccompanied minors, children with disabilities and adolescents and youth who have reached secondary and post-secondary levels [12]. More than 70% of child refugees report feeling safe at school, though a household survey revealed a number of risks associated with commuting to school or being in school including petty thieves and robberies, natural hazards, sexual violence, and violence in schools [12]. Around 30% of households in two of Kigoma’s refugee camps reported incidences of sexual violence and harassment experienced in camps. School safety is undermined by unsanitary bathroom conditions, and dilapidated and poorly maintained buildings. Corporal punishment appeared to be common in classrooms, though not formally reported on [12].

At the school level, all 27 primary and secondary schools in Nyarugusu refugee camp in Tanzania will be invited to participate. The intervention is delivered to individual teachers, and all teachers working in intervention schools will be eligible to receive the intervention. The hypothesized intervention effect will be in all students being taught by participating teachers; however, we will measure effects of the intervention in students who are aged 9 years and over as they are better able to respond to survey questions. Data are being collected from both students and teachers.

Headteachers will be approached by the research team for permission to conduct research in their schools and to participate in the intervention. Headteachers will be invited to provide written consent to participation on behalf of individual students aged less than 18 years. Individual students under aged 18 years will be asked to provide assent for participation; students aged 18 years and over will be asked to provide written informed consent. Individual teachers will be asked to provide written informed consent to participate in teacher surveys. During consent procedures it will be communicated to all individual respondents that participation in study activities is completely voluntary and that individuals will be free to withdraw from the survey at any point in time without any penalty. Teachers will be informed that if an interviewer feels that they are putting children at risk of serious violence, a child protection officer may need to be notified. Students under 18 years will be informed that if an interviewer feels their safety is at risk, a child protection officer may need to be notified. Students over 18 years will also be offered referral for further support. This is described further in the child protection section below.

We will conduct surveys at three time points: a baseline prior to the intervention, a midline soon after the 10-week intervention; and an endline at least 6 months after the end of the intervention (Fig. 1).

Data will be collected at each time point from students and teachers during individual interviews by Kiswahili or Kirundi-speaking interviewers. Before each round of data collection interviewers will receive 2 weeks of in-depth training on how to ask sensitive questions in a non-judgemental manner, survey procedures, and child protection procedures. Individual interviews will be done using a questionnaire programmed into tablet computers. Interviews with students will be conducted at schools, in places where interviewer-participant pairs are out of earshot but within sight of others, to protect confidentiality and ensure child safety. Female interviewers will interview both male and female participants; male interviewers will interview only male participants.

Survey data collected by tablet will be stored on a password protected database that is online-accessible only to senior study personnel, and backed-up daily in encrypted folders. The electronic master database will be held on IPA’s secure server. Devices will be password protected at all times to prevent accidental discovery of personal information by third parties during data collection. Data will contain identifying information as we intend to link data from individuals over time, and to link individual survey data with administrative data held by schools.

The aim of the EmpaTeach intervention is to improve student and teacher well-being, self-regulation, teacher classroom management, and teachers’ use of positive discipline techniques. Teachers in the intervention condition will receive a 10-week group intervention. Groups will meet 14 times for 1–1.5 h length sessions, which will be led by peer teachers. The intervention uses cognitive behavioural therapy techniques to change negative thought and behaviour patterns related to corporal punishment. The teachers will receive information on alternatives to corporal punishment, planning exercises and reinforcement SMS, and because the intervention is in a group setting, social support to change their behaviours. They will discuss their experiences and challenges in group sessions. Table 1 contains a detailed description of the EmpaTeach intervention.

These schools will receive no specific interventions related to violence prevention during the study, but will receive the intervention after the study is over if it is shown to be effective and funding permits.

The primary outcome will be past-week prevalence of physical violence from teachers to students, measured by students’ self-reports using a modified version of the International Society for the Prevention of Child Abuse and Neglect Screening Tool-Child Institutional (ICAST) [10, 13]. This will be assessed at midline. Secondary outcomes will include: students' self-reports of physical violence from school staff assessed at the endline; students' self-reports of emotional violence from school staff at midline and endline; students' depressive symptoms as measured by the Mood and Feelings Questionnaire [14] at midline and endline, and student attendance and educational test scores on standardised tests, using administrative data from the camp schools, assessed at the end of the school year. The questionnaire based measures have been adapted, cognitively tested and piloted by IRC and BIT in a neighbouring refugee camp with a similar population of teachers and children aged 7 and above, and by IPA in Nyarugusu refugee camp. Based on this process children aged 9 and above were deemed reliably able to answer questions on the above measures.

We will include a number of measures which may mediate or moderate the effects of the intervention, including: teachers’ reports of the use of physical and emotional violence; teachers’ mental health and history of trauma and violence; students’ self-reports of other forms of violence from school staff and other perpetrators; students’ perceptions of school connectedness, students’ perceptions of safety and student mental health. Some measures for students will be administered only to those aged 11 years and over, as some content has been deemed by our study team to be only relevant and appropriate for slightly older children, including all questions on sexual violence and PTSD. We have used all measures with children of the relevant age groups previously in other similar studies (although not previously with refugee children). Table 2 contains a summary of the tools used to measure study outcomes.

Assuming 25 of 27 schools in the camp agree to participate; an intra-class correlation coefficient (ICC) of 0.10 [10]; a significance level of 0.05; 80% power; and a 50% prevalence of past week violence, [10] we would be able to detect a 19% difference in prevalence of physical violence from school staff to students if we surveyed 50 students per school. We thus plan to invite 60 students per school to participate to allow for non-response (n = at least 1500 in total). We also plan to survey about 20 teachers per school (n = 500 teachers in total) to explore the intermediate outcomes of the intervention. We planned to stratify randomisation by predictive factors to ensure that the ICC is at the lower end of the range, including whether school are mainly Congolese or Burundian, and whether they are primary or secondary.

This is a behavioural intervention which is considered low risk, so no interim data analyses are planned during intervention implementation.

A stratified allocation list will be produced by LSHTM. Allocation will take place at a public meeting of all headteachers, where each headteacher or school representative will place their school name in an opaque bag according to their stratum. Names will then be drawn in turn from each stratum specific bag by one person nominated by the group, and allocated to either the intervention or wait-list control condition in the sequence on the allocation lists.

Due to the nature of the intervention, participants, implementers and research partners should be considered unblinded. The statistician performing the main trial analysis will be blind to allocation.

We have chosen to randomise at the school level to minimise possible contamination. As the intervention is mainly delivered to groups of teachers in a targeted way, we do not expect much natural diffusion due to the nature of the intervention itself. We expect that teacher transfer will be our main possible source of contamination. Based on 2017 data, 10% of teachers transferred to a different school during the school year (although we are informed that these levels are higher than normally expected due to a reorganisation of schools in 2017). We recognise that this contamination potentially reduces the observable effect, and we intend to collect data to assess whether contamination may have taken place.

All primary analyses will be carried out according to the principle of intention-to-treat (ITT) and using Generalised Estimating Equations (GEE) to take into account clustering at the school level. Our main analysis of the primary outcomes, and other secondary outcomes will be based on cross-sectional analyses comparing the outcome at midline for two main reasons. One, the intervention will be delivered to all teachers in a given school and is thus expected to impact on all pupils, not just on those pupils who were present at baseline. Two, the literature suggests that in cluster randomised trials, when migration into or out of the clusters is high over time, the baseline cohort may not remain representative of the cluster and therefore repeated cross-sectional analysis is preferred to minimise bias. However, where possible we will link student data across time points and will use analyses that include all students at all time-points, which essentially provides a repeat cross-sectional analysis with a nested longitudinal cohort.

Data will be analysed using multivariate regression models. We will carry out unadjusted analyses and analyses adjusted for any stratification factors and other pre-hypothesised potential confounders as covariates. Formal testing will be restricted to a pre-specified number of the secondary outcomes. A further small number of exploratory subgroup analyses will be specified in advance. These will include subgroups by sex, ethnicity, primary versus secondary school, baseline level of violence and student’s disability.

Other secondary analyses will include teacher outcomes and will be carried out according to the principle of ITT using a similar approach to modelling as described for the student outcomes.

The main aim of the qualitative study will be to explore the experience of teachers, students and school administration with the intervention, including what they perceive to be violence; to understand what participants felt ‘worked’ and ‘didn’t work’, and to gain insight into the potential mechanisms of action for teachers and for students. We will conduct semi-structured individual interviews with approximately 10–30 teachers and 10–30 students at three time points at the start of the intervention, during the intervention or immediately after the intervention, and at around 6 months after the intervention. We will attempt to follow the same participants over time. Additionally, we will conduct semi-structured key informant interviews with a sample of approximately 20–40 headteachers, discipline teachers, and group coordinators to understand barriers and facilitators to implementation and contextual information about schools and groups that may have influenced uptake of the intervention. We will conduct approximately 10–12 focus groups with students and teachers. Questions asked will differ based on whether teachers and students are in intervention or control schools. The sample will be weighted so that more participants (approx. 80%) are in intervention schools. Participants will be purposively selected based on analysis of baseline data and recommendations from stakeholders.

Interviews and focus groups will be conducted in Kiswahili, Kirundi or French by a trained research coordinator. These will take place in private locations convenient and safe for both the interviewee/group participant and interviewer/facilitator. With permission, interviews will be recorded, transcribed and translated. Analysis will be thematic, drawing on techniques from Grounded Theory, including constant comparison and searching for deviant cases. This approach will allow us to answer our main questions about how the intervention might work and contextual factors that might affect this, and also for other salient themes to emerge from the data that will provide further information about context more generally.

An economic evaluation will be conducted to estimate the costs and cost-effectiveness of the EmpaTeach intervention. A combination of top-down and ingredients-based costing approaches will be used to generate total cost estimates for each school site. All costing will be estimated from the provider’s perspective and financial and economic costs will be calculated for all inputs. The results of cost analysis will assess the costs of setting up and running the intervention, describe the distribution of costs across different forms of inputs, the unit cost per student reached, the cost of delivering all activities in the intervention site and the cost per unit of measure for selected intermediate intervention outcomes. Outcome measures for the cost-effectiveness analysis will be the change in violence outcomes (cost per case of violence averted). We will estimate the incremental cost effectiveness of the intervention relative to the status quo (represented by the control sites). The cost-effectiveness measure proposed here will be compared to similar violence-prevention programmes in the region and it will inform programme replication, scalability and financial sustainability.

We have chosen a repeat cross-sectional design for our trial, which should minimize risk related to possible attrition stemming from both ongoing resettlement of Burundian refugees and school re-organisations. Burundian refugees in the Camp have the option to voluntarily return home; but as of November 2018 only 6000 of 70,000 living in Nyarugusu Camp had signed up to return. We therefore expect that the number of Burundian students and teachers who are resettled is likely to be small during the period of the study. Obtaining a full sampling frame (in this case, a list of all students and teachers for each Camp school) has also proved challenging, with several re-organisations of schools involving large scale transfers of students which occurred in 2017 and 2018 meaning that school rosters are not up to date. We have also stratified allocation to intervention or control conditions by whether schools are Burundian or Congolese, and thus any outflow of Burundian teachers and students, or transfers between schools, should thus be balanced across arms.

There is also the possibility of another influx of refugees if there is another outbreak of conflict in a surrounding area. If the camp receives another influx of refugees, children would be admitted into existing schools (and/or new schools would be created, depending on the number of new students). Whether additional teachers could feasibly be trained would depend on the timing of the influx, and whether this aligned with the programme implementation period. In this event, we would attempt to document and explore the impact of this via our qualitative and quantitative implementation research.

The intervention is designed for teachers and students, and thus is not designed to directly affect violence outside of schools. In Nyarugusu Camp, about 50% of children are registered with primary or secondary schools, meaning that the intervention will not directly reach about half of the children in the Camp, and our research results should not be interpreted as generalizable to those children who do not attend school in the Camp. It is unknown at present if there are any systematic differences between those children who are and who are not registered in Camp schools.

If the EmpaTeach intervention is successful in reducing violence from teachers to students in Nyarugusu Camp, this will represent the first proven intervention model to reduce this form of violence in a conflict setting. IRC currently works in 40 countries, and provides a platform via which to scale EmpaTeach quickly where it already implements education programming.

Empateach and similar interventions are likely to be of interest in non-emergency contexts as well. For example, reports by NGOs indicate that secondary school girls and boys in Tanzania consistently describe high levels of violence and harassment in their schools [27]. Teachers discuss beating students to cause ‘fear’ and to get them into ‘shape’ [27]. If Empateach is successful, further adaptation of non-emergency contexts should be explored.

In Tanzania, IRC is also a leader within the Education Quality Improvement Programme (EQUIP), a consortium which develops and implements best practices to strengthen the quality of education in seven regions across Tanzania, and to ready such interventions for national scale-up. IRC also provides technical support to the Ministry of Education and Vocational Training on teacher professional development, and NIMR advises the Ministry of Health, Community Development, Gender, Elderly and Children. These connections will be leveraged to advocate for uptake.