Evaluation of School-based Health Promotion Programmes in Bihar, India

Background and objective:

India is home to 358 million young people in the age group of 10 to 24; of these 243 million are between 10 and 19 years age. This represents a huge opportunity that can transform the social and economic fortunes of the country. Key public health challenges for adolescents in India include unwanted pregnancies, sexually transmitted and reproductive tract infections, injuries, growing misuse of alcohol, tobacco and other substances, and mental health problems such as depression, anxiety disorders and suicide. Schools provide an ideal platform for addressing these health issues and the WHO's Health Promoting Schools framework seeks to address some of the structural determinants of these health concerns. The recent Cochrane review assessed school-based interventions and found that 'whole school' or 'school environment' interventions are effective in addressing a range of health outcomes among adolescents including bullying, aggression, and tobacco use. However, there is no comparable evidence on effectiveness and cost-effectiveness on school-based health promotion programmes in India.

SEHER-Strengthening Evidence base on school-based intErventions for pRomoting adolescent health, is a jointly funded programme (by The MacArthur Foundation and United Nation's Population Development Fund, India) that seeks to develop and evaluate a comprehensive adolescent health promotion intervention delivered by two different delivery agents viz. teacher as SEHER Mitra (TSM) and a lay health worker called as SEHER Mitra (SM) in government run secondary schools in Bihar, India.

Hypotheses:

1. For primary outcome measure: It is hypothesized that the intervention strategies in addition to usual intervention (Tarang-AEP) will have a graded effect on overall school climate.
2. For secondary outcome measures: The interventions will build positive attitude towards gender equity, build knowledge of and attitude towards reproductive and sexual health, reduce self-reported bullying, violence and depression.
3. For exploratory outcome measures : The interventions will decrease tobacco, alcohol and other substance use, reduce suicide behaviour (suicide thoughts and attempts) and increase safe sexual behaviour.

An addition of the SEHER intervention with the already existing Tarang-AEP will enhance the effect on the above mentioned outcomes; low resource intensity intervention (Teacher-as-SEHER Mitra: Arm-I) having minimal to moderate effect, and the high resource intensity intervention (School Mitra: Arm-II) having the greatest and significant effect.

Randomisation:

The school as the unit of randomisation will be allocated to one of the intervention arms or comparison using minimisation during the pilot study (June 2014-March 2015). Although, usually a pilot study would be conducted in different schools from those in the main trial, it has been possible for SEHER to conduct the pilot study in the main trial schools because a new cohort of standard IX students will be entering school every year.

Before randomisation, a list of eligible schools for randomisation (n=112) from the total number of secondary and higher secondary schools in Nalanda district of Bihar (n=136) was generated. The following criteria were used to generate the list: implementation of Tarang-AEP programme, total number of students in the school, and total number of employed teachers in school.

Of the 112 schools, 75 were randomly selected for the random allocation. To have a representative pool of 75 schools, and to ensure an equal number of schools of each type in each of the three trial arms, 68 % of co-educational (63 out of 93), 69% of only girls' (9 out of 13) and 50% only boys' schools (3 out of 6) were selected. All the 75 schools are allocated to comparison or one of the intervention arms by using minimization. To carry out the allocation by minimisation, the arms were balanced on the following variables (each of the variable will be classified into categories):

1. Type of school (secondary= 1; and secondary and higher secondary school= 2);
2. School size (small=100-300; medium=301-600, and large=601 and above students ), and
3. Nature of school (Co-education=1; only boys'=2; and only girls' =3). The random allocation by using minimization was carried out by an independent statistician (Gian Luca DiTanna, LSHTM) using the R software package .

Sample size:

The sample size estimations are based on various scenarios for the changes in the primary and two secondary outcome measures, viz. overall school climate, self-reported bullying experience, and prevalence of moderate to severe depression, respectively. The baseline average score and prevalence rate/proportion of these outcome indicators and respective intra-class coefficients (ICCs)are based on the data collected during the pilot testing of the SEHER outcome assessment questionnaire in 15 schools (5 schools in each arm; N=1722 participants).

• The arithmetic mean score on Beyond Blue School Climate scale is 20.7 (SD 6.7) with an ICC of 0.018. The minimum and maximum score on this 28 item scale range from 0 to 28, with higher scores indicating a more favourable experience of School Climate.
• 18% students in secondary school in Nalanda, Bihar reported of experiencing bullying in last 30 days (In the last 30 days, I were made fun of with sexual jokes, comments, or gestures in the school) with an ICC of 0.03
• The arithmetic mean score on the Primary Health Questionnaire -9 (PHQ-9) is 6.12 (SD 5.04) with an ICC of 0.0001
• The cluster number is 25 per arm and average cluster size is 200 (average classroom size in standard IX in schools in Nalanda district is 70 students with on average 3 divisions in each school)
• Assumed 15% loss to follow-up

Based on these assumptions, the proposed trial will have 98% power to detect an effect size of 0.2 (difference in means/SD) average school climate score between the comparison and intervention arms, with 95% confidence and an ICC of 0.02 (88% power among boys and 93% among girls respectively).

The trial will have 83% power to detect a 6% absolute difference in the proportion of students who reported of experiencing bullying between the comparison and intervention arms, with 95% confidence and an ICC of 0.03 (82% power to detect a 7% difference in boys and 83% power for a 6% difference in girls respectively). Likewise, the trial will have at least 93% power to detect an effect size of 0.1 in depression score overall, with 95% confidence and an ICC of 0.0001 (88% power among boys with an ICC of 0.03 and 92% among girls with an ICC of 0.005 respectively).

Process Evaluation: Process evaluation will be an integral part of the research design and will examine the quality of programme implementation (i.e. its integrity or fidelity), the completeness of its delivery, and the extent to which stakeholders engage with it. In doing so, it can help explain the programme's outcomes and identify ways to improve and/or replicate it. Two types of indicators will be collated to evaluate the fidelity of the delivery of the SEHER interventions, viz. their quantity and quality.

Data management: Two types of quantitative data will be collected: intervention process, and outcome assessment. All outcome assessment data will be collected in paper form, as will process data from TSMs/SMs. All the data will be manually entered into Access database. Range and consistency checks will be performed at monthly intervals for all process data. Queries identified will be resolved promptly by the trial management team, and the database updated, maintaining the audit trail. All data will be kept in separate databases and only merged into a master database after data collection is completed and each individual database will be locked. All process data will be backed-up on external hard disks on monthly basis. Access to pre-locked data will be password protected at multiple levels and no member of the trial team apart from the data manager and independent statistician will have access to these passwords. After the dataset is locked, the dataset will remain password-protected and trial investigators will have access to the datasets. For all data, a separate file linking names and trial identification codes will be kept and password protected.

Analysis:Quantitative analyses will be carried out using STATA (version 13). Below is a summary of our approach to the analysis. A detailed analysis plan will be agreed with the Data Safety and Monitoring Board towards the end of the trial and/or before any analysis is undertaken.

Descriptive analyses: Initial analyses will compare characteristics of enrolled participants within and across arms (for example: age and socioeconomic background).An initial stage of statistical analysis will also involve examining the balance of characteristics within and across three arms of the outcome measures used in the SEHER study.

The outcome measures will be summarised at baseline and follow-up by each cluster, arm-wise and overall. These will be summarised by means (SD), medians (Inter-Quartile Range) or numbers and proportions as appropriate by key relevant subgroups (such as age, gender, and baseline outcome score). For continuous outcomes, histograms will also be plotted within each arm to assess normality, and whether any transformation is required.

Outcome analyses: The primary analyses, based on repeated cross sectional surveys, will be intention-to-treat at the end of the trial and using multi-level modelling to take into account clustering at the school level. The sub-group analysis will involve gender-wise as well as analyses with students who have at least 60% attendance in the academic year. The trial outcomes will be evaluated for the primary and secondary outcomes based on the change in the intervention arm compared to the change in the comparison relative to "baseline". The primary trial impact comparison will be analysed between the baseline assessment and the 6-7 month follow-up assessment.

Mixed effects regression will be performed to analyse changes in primary and secondary outcomes at school-level. The multilevel analyses will contain two levels: the school level and the individual student level. Effect sizes will be presented as risk ratios for binary outcomes (for e.g. experience of bullying), and as mean differences for continuous outcomes (for e.g. overall school climate score); 95% confidence intervals (CI) will be derived for both.

The reporting and presentation of this trial will be in accordance with the CONSORT guidelines for cluster randomised trials, with the primary comparative analysis being conducted on an intention-to-treat basis.