Effectiveness of a girls’ empowerment programme on early childbearing, marriage and school dropout among adolescent girls in rural Zambia: Study protocol for a cluster randomized trial

Background: Adolescent pregnancies pose a risk to the young mothers and their babies. In Zambia, 35% of young girls in rural areas have given birth by the age of 18 years. Pregnancy rates are particularly high among out-of-school girls. Poverty, low enrolment in secondary school, myths and community norms all contribute to early childbearing. This protocol describes a trial aiming to measure the effect on early childbearing rates in a rural Zambian context of (1) economic support to girls and their families, and (2) combining economic support with a community intervention to enhance knowledge about sexual and reproductive health and supportive community norms. Methods/design: This cluster randomized controlled trial (CRCT) will have three arms. The clusters are rural schools with surrounding communities. Approximately 4900 girls in grade 7 in 2016 will be recruited from 157 schools in 12 districts. In one intervention arm, participating girls and their guardians will be offered cash transfers and payment of school fees. In the second intervention arm, there will be both economic support and a community intervention. The interventions will be implemented for approximately 2 years. The final survey will be 4.5 years after recruitment. The primary outcomes will beincidence of births within 8 months of the end of the intervention perio,incidence of births before girls’ 18th birthda andproportion of girls who sit for the grade 9 exa. Final survey interviewers will be unaware of the intervention status of respondents. Analysis will be by intention-to-treat and adjusted for cluster design and confounders. Qualitative process evaluation will be conducted. Discussion: This is the first CRCT to measure the effect of combining economic support with a community intervention to prevent adolescent childbearing in a low- or middle-income country. We have designed a programme that will be sustainable and feasible to scale up. The findings will be relevant for programmes for adolescent reproductive health in Zambia and similar contexts. Trial registration: ISRCTN registry: ISRCTN12727868 , (4 March 2016). The intervention packages will be examined in a cluster randomized controlled trial with two intervention arms and one control arm (Fig. 1). The randomization units will be basic schools and their surrounding communities. (Basic schools offer grades 1–9. Primary school in Zambia comprises grade 1–7 and grades 8 and 9 are referred to as junior secondary school.) The selected schools are at least 8 km apart from each other. Flow chart of trial The participants will be girls enrolled in grade 7 in 2016 in rural basic schools in 12 districts in Zambia: Kalomo, Choma, Pemba, Monze, Mazabuka, Chikankata, Chisamba, Chibombo, Kabwe, Kapiri Mposhi, Mkushi, and Luano. These districts were selected as they have medium school dropout rates, and adolescent marriage and childbearing are common. All girls enrolled in grade 7, including anyone who is already married or has children, will be eligible to participate. Girls who drop out of school after they have been recruited will still be followed up and can continue to receive the interventions. To achieve community acceptance, chiefs, headmen, religious – and informal – leaders, headmasters and Parent Teacher Association (PTA) committee members were oriented and asked to support the trial activities before the recruitment was initiated. Local radio has also been used for community sensitization. Communities and schools were informed that any individual school in which > 15% of the girls did not assent would not be included in the trial. The interventions will be launched in September 2016 and will last for 27 months, until November 2018 (the end of the academic year when the girls who attend school are expected to complete grade 9). The intervention may be extended one more year, given supplementary funding. In all the study arms, girls will be offered writing materials (exercise books, pencils and pens) as an incentive to participate. Apart from this, only standard school and health services will be offered in the control arm. In the intervention arms, girls and their parents/guardians will be offered economic support, consisting of a monthly cash transfer for the girl (ZMW 30), an annual cash grant to her parents/guardians (ZMW 350/year) and payment of school fees for girls who enrol in grade 8 and 9 (up to ZMW 500 per term). The support package targets the key actors in the decisions leading to early pregnancy and marriage. Cash transfers target the poverty dimension, by making it somewhat less urgent for the guardians that the girl gets married and for the girl to receive gifts from a boyfriend. The payment of school fees will be made directly to the school bank account for girls who get a place in grade 8 and 9. The money for the girls and the guardians will be disbursed by a cash transfer committee consisting of a teacher and two parents from the PTA committee. At least two of the cash transfer committee members will be present during disbursement to witness that the right persons receive the cash. Girls will also be asked in every follow-up contact how much money they have received, and all participants will be encouraged to contact the study team if they do not receive the right amount. There will be no age limit for the economic support for girls who are in school, but for girls who drop out of school, it will stop after the 18th birthday. The economic support will also be discontinued for girls who do not participate in the follow-up contacts. The second intervention arm will combine the economic support with a community-oriented intervention consisting of (1) community and parent meetings employing a community dialogue approach in promoting supportive community norms around education for girls and postponement of early marriage and early childbearing; and (2) establishment of youth clubs in order to provide comprehensive sexual and reproductive health education among in- and out-of-school adolescent girls and boys. Girls participating in the trial and boys who attend grade 7 in 2016 in the randomly selected schools will be invited to participate in a youth club every fortnight during the school terms, and they will all be welcome to continue in the youth club even if they quit school. The meetings will include interactive discussions on education, early marriage, the risks of early pregnancy, gender roles, and sexual and reproductive health, including myths around modern contraceptives. We will test a model where a teacher is linked with a community health assistant (CHAs) or a community health worker (CHW) to run the youth club together. Meetings will be held to inform parents about the content of the youth club sessions. Before the intervention is launched, the selected teachers and CHAs/CHWs will be given a 5-day training which will focus on the SRH curriculum, facilitation techniques and approaches to community mobilization. Refresher training will be held halfway during the intervention period. In addition, orientation meetings will be held to inform other healthcare workers in the catchment area of the schools about the project and the importance of providing youth-friendly health services. For each combined intervention cluster, two young (<20 years) unmarried women from the local community will be selected as youth peer educators. The role of the peer educators will be to mobilize girls and boys to come to youth club meetings. Meetings with the community at large will be organized bimonthly. These meetings will be conducted using a dialogue approach [56] and will discuss topics such as the value of education, and the risks and benefits of early childbearing. Films or role plays will be used to start discussions. The outcomes, measurement tools, measurement times and validation tools are listed in Table 1. Outcomes, measurement tools, measurement times and validation tools We used PASS 14 (NCSS Statistical Software, Kaysville, UT, USA) to calculate sample size required for a cluster randomized trial with the primary outcomes “incidence of births within 8 months of the end of the intervention period”, “incidence of births before girls’ 18th birthday” and “proportion of girls who sit for the grade 9 exam”. We used the 2010 census estimates of the percentages reporting ever giving birth in the study districts to estimate the incidence of childbearing in the control arm: 2% at the average age of 14.5 years, 4% at 15.5 years, 9.5% at 16.5 years, 22% at 17.5 years, and 35% at 18.5 years. The interventions will not affect childbearing until approximately 9 months after the start of the intervention, when the average age will be 15. We assume that 3% of the girls will have given birth before any effects of the interventions can be seen. The other assumptions for the sample sizes required to detect differences for the primary outcomes “incidence of births before girls’ 18th birthday” and “incidence of births within 8 months of the end of the intervention period” are presented in Tables 2 and ​and3.3. We find (using two-sided tests) that with 63 clusters in each of the intervention arms and 31 in the control arm we will have 70% power to detect the assumed difference between the economic and the combined interventions, 80% power for the comparison of the economic intervention versus the control arm, and > 95% power to detect the assumed difference between the combined intervention versus the control arm for the outcome “incidence of births before girls’ 18th birthday”. For the outcome “incidence of births within 8 months of the end of the intervention period” we will have > 90% power for the comparison of the combined intervention versus the control group. Assumptions for sample size required to measure the primary outcome “incidence of births before girls’ 18th birthday” Assumptions for sample size required to measure the primary outcome “incidence of births within 8 months of the end of the intervention period” The percentage of girls enrolled in grade 7 who sat for the grade 9 exam was approximately 50% in 2012. In 2015 the Ministry of General Education (MoGE) removed the previous entry requirements for grade 8 in Central province and all pupils who sit for the grade 7 exam in Chibombo, Chisamba, Kapiri Mposhi, Kabwe, Mkushi and Luano will be admitted to grade 8. To be enrolled they will have to pay school fees. The percentage who enrols in grade 8 is very likely to increase due to this policy change. This policy has not been introduced in Southern province (where the districts Kalomo, Choma, Pemba, Monze, Mazabuka and Chikankata are located). We have assumed that the percentage will be 70% overall in the control arm. See Table 4 for the other assumptions. With 30 clusters in each arm, we will have 95% power or more for each of the three comparisons. Assumptions for sample size required to measure the primary outcome “proportion of girls who sit for the grade 9 exam” Taking the largest of these sample sizes, we need 63 clusters in each of the intervention arms. Since we expect larger differences between each of the intervention arms and the control arm than between the two intervention arms themselves we can reduce the total number of clusters by allowing for a lower sample size in the control arm. Thus we will include at least 63 economic intervention clusters, 63 combined intervention clusters and 31 control clusters, i.e. 157 clusters with a total of 31 × 157 = 4867 or approximately 4900 girls. The recruitment started in March and will be completed in July 2016 (Fig. 2: SPIRIT figure). The parents/guardians of girls in grade 7 are invited to an information meeting. If their daughter is < 18 years, they are asked to give consent to her participation in the trial (Additional file 1). After the consent is obtained, girls are informed and asked to assent by a research assistant (Additional file 2). Any girl aged ≥ 18 years is informed directly and asked to consent (Additional file 3). Those who consent/assent and participate in the baseline survey interview are enrolled. SPIRIT figure. Schedule of enrolment, interventions, and assessments Randomization will take place in July 2016, after recruitment is completed. We will organize six randomization ceremonies, each for two districts, where the schools will be stratified by district and randomly allocated to the three arms. Before each ceremony, 1000 allocations will be computer-generated by an independent scientist from the Centre for Interventions Science in Maternal and Child health (CISMAC), and each allocation will be numbered. Officials from the study districts, chiefs, head teachers and PTA members of the trial schools will be invited to be present. Community members will be welcome too. Tickets with numbers corresponding to a specific allocation will be drawn from a box. There will be no blinding of participants, but the team doing the final survey will be independent from the intervention delivery and they will be unaware of the intervention status of respondents. Biannual follow-up contacts with the participants and the final surveys will be conducted by research assistants who are independent from the intervention implementation team. All the tools have been translated to Tonga, Nyanja, Bemba and Lenje, the major local languages in the study districts, and then back-translated to ensure that the content is maintained. Interviews are conducted in English or one of these four local languages, depending on the preference of the participant. The main data collection tools are described in Table 5. Data collection elements FGD focus group discussion, IDI in-depth interview, SRH sexual and reproductive health We will collect information about the location and size of schools that are excluded due to suboptimal participation. All the quantitative tools have been piloted to ensure that the questions and the translations are relevant and comprehensible. Data from interviews will be captured electronically using tablets and CommCare (https://www.commcarehq.org/home/) as our data management software. The forms have inbuilt check-and-skip rules to minimize data entry errors. Names and telephone numbers and addresses will not be recorded in the same forms as sensitive data. Each participant will be given a unique identifier, and this number will be used when storing forms. Only the Data Manager, the Principal Investigator and the Co-Principal Investigator will have access to the personal identifiers. All data will be saved on password-protected computers and tablets. The database is hosted by Dimagi (http://www.dimagi.com/), and weekly downloads of data are done to a secure server owned by the University of Bergen. When the trial is completed, all personal identifiers will be deleted. In the qualitative process evaluation, we will seek permission to audio record the individual interviews and focus group discussions using digital recorders. The recordings will be transcribed and translated verbatim. A senior researcher will review the transcripts to ensure that they contain word by word transcriptions and translations that retain original meanings, with a particular caution to retain culturally embedded content. The data will be analysed with Stata 14 (StataCorp, College Station, TX, USA) software. The outcomes will be captured in time-to-event, binary and continuous variables. Descriptive statistics will be used to describe continuous and categorical variables. We will compare outcomes between the three arms: the economic arm versus the control arm, the combined arm versus the control arm, and the economic arm versus the combined intervention arm. Analyses will be by intention-to-treat (ITT). The childbearing outcomes will be measured based on the participants’ responses to the questions “Have you ever given birth to a baby who was born alive?”, “Have you ever given birth to a baby who was born dead?”, “If yes, on which date and which year did you give birth?” and “How many months pregnant were you when you gave birth?”. We will count stillbirths after 6 months or 28 weeks pregnancy as births. Pregnancies conceived before the intervention starts will be excluded from the analysis of effects on childbearing and pregnancy. For the incidence measures of childbearing, survival analysis using Cox regression will be employed. Time-on-study will be used as the timescale when measuring outcomes occurring within 8 months or 2 years of the end of the intervention period, and age will be the timescale when measuring outcomes before the 16th and 18th birthdays. Proportions will be compared using generalized estimating equations (GEE) with a log link. The ICC will be reported for all primary and secondary outcomes. Any remaining imbalances of predictors of the outcomes after randomization will be adjusted for in the regression models. In all models, we will also adjust for potential confounders such as age. When there are missing values, we will explore whether data is missing completely at random (MCAR) or whether data is missing at random (MAR). If MCAR and MAR are satisfied, we will do complete case analysis and explore multiple imputation using Stata software. Equity effects will be examined by comparing distributions of primary and secondary outcomes between wealth tertiles and educational tertiles within each trial arm. Inequality will also be analysed using the concentration index. Further details will be provided in a separate statistical analysis plan. The content of the qualitative data from IDIs and FGDs will be explored on the same day as interviews and FGDs are done, and the researchers will adjust the interview guides to enhance their relevance. The transcribed and translated texts will be entered into QSR NVIVO 10. The analysis of the interviews, FGDs, open-ended questions from log forms and observations will follow a classical approach employing Malterud’s ‘Systematic text condensation’ [57], a descriptive and explorative method for thematic cross-case analysis drawing upon Giorgi’s psychological phenomenological analysis. Systematic text condensation consists of the following four steps: (1) total impression; (2) identifying and sorting meaning units; (3) condensation; (4) synthesizing. A Data Monitoring Committee (DMC) with three members has been established. The committee will be independent of the project management team. The committee will advise on study modification or termination based on its reviews of data. A charter for the DMC has been developed and can be obtained from the authors. The DMC will review the follow-up rates from the 6-monthly surveys. Participants in the intervention arms will benefit from what the intervention packages offer. If the intervention packages are found to have a positive impact and the government consequently introduces a similar programme, adolescents and their families in many communities may benefit. Unanticipated problems or adverse events such as deaths and hospitalizations and whether these are likely to be related to the trial, will be recorded and reported to the ethics committees (UNZABREC and REK-West) and the Data Monitoring Committee. All community and parent meetings will be captured photographically if those present consent, and a report form will be filled in that automatically captures GPS coordinates to verify the place and time of the meetings. Youth club and community meetings will be regularly monitored by the research team to observe the quality of the delivery of the intervention. An external expert assigned by CISMAC will conduct annual monitoring visits during study implementation. If challenges are detected, an action plan to address these will be prepared. The intervention will be adaptive, i.e. elements in the intervention package delivered to a cluster will be modified if obstacles to the implementation of the intervention are encountered, the participation is suboptimal or if important changes occur. When the Project Management Team believes it is important to make substantial adaptions of the trial design, CISMAC will be asked to critically consider whether the proposed adaptions are necessary and sufficient. Important protocol modifications will be reported to the DMC, UNZABREC and REK-West, and registered in ISRCTN and clinicaltrials.gov. The data will be owned by the University of Bergen and the University of Zambia. Project management team members will have access to the data. Other requests for access will be considered after 3 years of trial completion. This protocol was written following the Standard Protocol Items: Recommendations for Interventional trials (SPIRIT) checklist (see Additional file 10). Dissemination of the research findings will be done through scientific articles in peer-reviewed journals, reports and presentations at national and international academic- and policy-related conferences. We will report findings according to the Consolidated Standards of Reporting Trials (CONSORT) guidelines. Protocol modifications will be reported when disseminating findings. Authorship of scientific articles emerging from the study will be decided upon following guidelines from the International Committee of Medical Journal Editors. We have established an advisory group with representatives from the MoGE, Ministry of Health (MoH), Ministry of Gender, and the Ministry of Traditional Affairs. We will share and discuss research findings with the group and other key stakeholders in Zambia and the study communities.