Effect of hygiene & menstrual hygiene interventions on learning and psychosocial wellbeing in Madagascar

This project builds on the work of our partner NGO, CARE, in rural Madagascar, which was evaluated in our previous RCT (Kilonga I project). The activities included in this evaluation were evaluated in the previous RCT as a package, and include the following:
• Infrastructure. Construction of latrines and handwashing basins in schools that lack this infrastructure.
• Young Girl Leaders. CARE will recruit Jeune Fille Leaders (JFLs, “Young Girl Leaders”): schoolgirls who are willing to speak out against harmful social norms and act as advocates for menstrual hygiene in their school. Using data from teacher and professional staff assessments, these girls will be selected based on their willingness to actively and openly speak about menstruation, despite the taboo nature of the topic, and on their broader leadership skills. YGLs are trained on a curriculum focusing on key hygiene behaviors, the reduction of menstrual stigma, and leadership skills, and are asked to promote hygiene and reduce stigma among their peers and classmates at school. The YGLs are thus intended to act as a prominent example of someone engaging in behaviors that undermine the norm of stigmatization from within the social network.
• Teacher sensitization: Training sessions for teachers on water, sanitation, and hygiene (WASH) practices, including menstrual hygiene. After the training, teachers are encouraged to dedicate five minutes each week to discussing these topics with their students.
• Distribution of free sanitary pads: Each girl receives a free voucher that can be exchanged for recyclable menstrual pads produced by local seamstresses.
• WASH committee: Each school forms a committee consisting of students, teachers, and parents to motivate and organize hygiene-related activities.
• WASH competition: A school competition organized by the NGO to promote cleanliness and motivate schools to maintain tidy premises.

2024-11-01

2027-07-01

• Learning. Standardized test scores in mathematics and literacy (direct assessment of 17 girls per school), and official school grades (79 girls per school).
• Psychosocial wellbeing and physiological stress: objective measures of stress, including heart rate, blood pressure, and salivary cortisol reactivity. The prespecified primary cortisol endpoint is the area under the salivary cortisol concentration curve with respect to increase (AUCi), constructed from three saliva samples collected during the endline biomarker visit.
• Menstrual stigma (self-reported)

Salivary cortisol will be measured at endline, up to approximately 19 months after baseline, in a nested biomarker subsample covering approximately 80 of the trial's 140 schools. The sample will include up to approximately 17 girls per school, with a planned maximum of approximately 1,360 girls.

Three saliva samples will be collected: (1) when the girl arrives at the clinic, before clinical procedures; (2) approximately 5-8 minutes before venipuncture; and (3) approximately 20 minutes after venipuncture. Venipuncture provides the acute stressor. Let C₁, C₂, and C₃ denote cortisol concentrations measured at elapsed times t₁, t₂, and t₃. We will calculate the area under the curve with respect to ground using the trapezoidal rule: AUCg = [(C₁ + C₂)/2] × (t₂ − t₁) + [(C₂ + C₃)/2] × (t₃ − t₂). The primary cortisol endpoint will be AUCi = AUCg − C₁ × (t₃ − t₁) and will be reported in μg/dL × min.

Cortisol AUCi is the single primary cortisol outcome. Individual cortisol concentrations, AUCg, changes between individual samples, percentage changes, and classifications based on cortisol response or recovery will be treated as secondary or exploratory outcomes.

• Hygiene and menstrual hygiene knowledge and behaviour (self-reported measures, distinguishing between menstrual health practices and other health & hygiene practices).
• Student and teacher motivation (self-reported measures)
• Network integration and bullying (self-reported)
• Self-reported psychosocial wellbeing
• Information diffusion measures (self-reported exposure to discussions on hygiene/menstruation)
• School attendance (official school attendance and self-reported measures)
• School educational outcomes (grades progression, repetition, and previous school history).
• Results on standardized cognitive tests (direct assesment)
• Measures regarding early sexual activity and pregnancy (self-reported)
• Health – health measures including self-reported infectious diseases, middle upper-arm circumference measures, urinary tract infections, and inflammation (blood-based C-reactive protein (CRP), measured during the endline biomarker visit in the same nested biomarker subsample)
• Additional outcomes for YGLs: self-confidence/self-perception, openness to discussion

CRP will be measured once from blood collected during the endline biomarker visit and reported in milligrams per liter (mg/L). CRP is a secondary outcome and is separate from the primary cortisol endpoint.

The study will be an RCT in 140 secondary schools in rural areas of the Amoron’I Mania region in Madagascar. The schools will be randomly assigned to one of 4 treatment arms:
1. Control (35 schools)
2. Infrastructure only (35 schools) – only construction of latrines and handwashing basins
3. Sensitization interventions + free sanitary pads (35 schools) – only Young Girl Leaders, teacher sensitization, school-based hygiene competition, and free distribution of menstrual pads. 2-6 girls are nominated in each school to be Young Girl Leaders.
4. The full package (35 schools) – combine all interventions from (2) and (3).

Hypotheses being tested:
• Effect of infrastructure. Understanding the impacts of the infrastructure by comparing arms with and without infrastructure is crucial for evaluating the overall cost-effectiveness of the program, since it is expected to contribute to about 60% of the overall cost. Measuring the psychosocial wellbeing effects of physical infrastructure is also crucial. Our qualitative work has indicated that in this setting, with very high rates of poverty and little state presence, having salient interventions in schools (visits from officials or construction of visible infrastructure) could have important effects on motivation and other psychosocial factors driving education, even if they do not target these outcomes directly.
• Effect of sensitization + sanitary pad interventions. Conversely, since the rest of the bundle of interventions is substantially cheaper to implement, finding that it is sufficient to generate learning effects could imply that it is very cost effective, competitive with the most cost-effective interventions described by a Global Advisory panel on education interventions (GEEAP 2023).
• Complementarity between infrastructure and other interventions. There may be important complementarities between the two types of interventions: for example, girls may only be able to change their hygiene behaviour and begin handwashing if they have access to handwashing basins at school. The 4th arm thus enables us to test this complementarity.

Data collection:
• Baseline (pre-intervention, October 2024), midline (after 1 year of intervention), and endline (after 2 years of intervention)
• Girls’ survey (17 girls per school) using surveys at the household of the 2,380 teenage girls and their mothers; teacher and director survey (maximum 3 per school); school-level survey.
In midline and endline, we also add approximately 4-5 girls per school to the sample who were potential nominees to be Young Girl Leaders (both in treatment and control schools, regardless of whether they were eventually trained), in order to measure causal effects on YGLs themselves.

BIOMARKER OUTCOMES:
Biomarker outcomes will be collected at endline in a nested subsample of approximately 80 schools spanning all four randomized treatment arms. The sample will include up to approximately 17 girls per school, with a planned maximum of approximately 1,360 girls. The biomarker sample will be selected before biomarker assay results and treatment-effect analysis. School selection reflects the operational accessibility and cold-chain requirements of biomarker collection while preserving coverage and balance across treatment arms. Within selected schools, the target is up to 17 girls from the original study sample. Up to three Young Girl Leaders will be prioritized where available.

We will estimate intent-to-treat effects according to each school's original randomized assignment. The main specification will regress each biomarker outcome on indicators for the infrastructure-only, sensitization-plus-free-pads, and combined-treatment arms, with control schools as the omitted category. We will also report the p-values pooled factorial comparisons of infrastructure versus no infrastructure and sensitization versus no sensitization, together with tests comparing the treatment arms.

Regressions will include randomization-stratum fixed effects and baseline-grade fixed effects. We will use post-double-selection LASSO to select additional controls from prespecified pre-treatment covariates (Belloni, Chernozhukov, and Hansen, 2014). The selection procedure will take the union of variables that predict the outcome and randomized treatment assignment, while always including randomization strata and baseline grade into the model. An exact baseline measure of the outcome will also be included when one exists (although no baseline salivary cortisol or CRP measure was collected). Standard errors will be clustered at the school level.

The set of prespecified pre-treatment covariates for the post-double-selection LASSO was fixed prior to endline data collection and outcome unblinding. As a robustness check, we will additionally report an unadjusted specification, without LASSO-selected controls, to demonstrate that results are not sensitive to the covariate-selection procedure.

Reference
Belloni, A., Chernozhukov, V., & Hansen, C. (2014). Inference on treatment effects after selection among high-dimensional controls. The Review of Economic Studies, 81(2), 608–650. https://doi.org/10.1093/restud/rdt044

Not available

Randomization done in office by a computer

Cluster randomized at the school level

Yes

140 schools

The main evaluation sample remains approximately 17 girls in each of 140 schools, for a planned total of 2,380 girls. Salivary cortisol and CRP will be collected at endline in a nested biomarker subsample covering approximately 80 schools, with up to approximately 17 girls per school and a planned maximum of approximately 1,360 girls.

35 schools in each of the treatment arms.

The design yields an MDE of 0.15, 0.12 and 0.11SD for learning, stress (as measured by heartrate), and school grades, respectively for the infrastructure versus no-infrastructure comparisons (70 vs 70 schools). The design also gives an MDE of 0.18, 0.15 and 0.12 SD for comparing the infrastructure only with the interventions including all other components (35 versus 70 schools). These MDEs align with the impact sizes observed in our first RCT, ensuring adequate power to detect meaningful effects. Our main statistical analysis will use ANCOVA regressions to compare the treatment arms. The detailed baseline modules for the questionnaire are attached.