Impact Evaluation of Accelerating Climate-Resilient Water, Sanitation and Hygiene Services for Schools and Healthcare Facilities in Karamoja

BACKGROUND AND RATIONALE Access to safe water and sanitation in schools is essential for children's health, dignity and learning. However, evidence on the effects of school WASH interventions remains mixed. A systematic review of 38 studies found reductions in school absence ranging from 21 to 61 per cent. Several trials reported larger effects for girls. However, some cluster-randomised trials found no effects. These include a trial in 60 Ethiopian schools that found no improvement in attendance. The largest effects are generally found in settings with severe water scarcity. A Kenyan trial found a 58 per cent reduction in the odds of absence among girls in schools receiving water treatment. However, it found no overall effect across all schools. The intervention also reduced diarrhoea prevalence by 66 per cent in water-scarce schools. It had no effect in schools that already had access to water. A recent meta-analysis of 50 trials found a similar pattern. WASH interventions reduced diarrhoea risk by 33 per cent during the dry season, compared with 18 per cent during the rainy season. Evidence on menstrual health interventions is also mixed. In Uganda, girls are nearly six times more likely to miss school during menstruation than on other days. Multi-component menstrual health programmes consistently improve knowledge, attitudes and self-efficacy. However, the most rigorous Ugandan trial to date found no effects on school attendance or educational performance after one year. This was despite improvements across almost every measured dimension of menstrual health. This trial takes place in a setting with severe baseline deprivation. Existing evidence suggests that WASH interventions may have larger effects under such conditions. The trial therefore examines whether the intervention produces detectable health and education effects in Karamoja. SETTING AND CLUSTER SELECTION The trial covers all nine Karamoja districts. These are Abim, Amudat, Kaabong, Karenga, Kotido, Moroto, Nabilatuk, Nakapiripirit and Napak. A rapid vulnerability assessment was conducted between October and November 2025. It assessed all 182 operational primary schools using 12 water-security indicators. These indicators were adapted from the Household Water Insecurity Experiences Scale. We classified each school as water secure or as experiencing mild, moderate, or severe water insecurity. The 60 schools with the highest composite priority scores form the trial sample. Eligibility was also conditional on hydrological viability, as confirmed through groundwater surveys. The nine districts were grouped into three geographic zones. The groupings reflect Uganda's 1974 provincial zoning and Karamoja's agroecological gradient. The North zone includes Kaabong and Karenga and has 17 study schools. The Central zone includes Kotido, Abim and Nabilatuk and has 25 schools. The South zone includes Moroto, Napak, Nakapiripirit and Amudat and has 18 schools. RANDOMISATION Schools were allocated proportionally within each zone. This ensured that every zone contributed to all three study arms in proportion to its share of the sample. In the North, eight schools were assigned to Phase 1, six to Phase 2 and three to the pure control group. The corresponding allocation was 13, eight and four schools in the Central zone. In the South, we assigned nine schools to Phase 1, six to Phase 2, and three to the pure control group. The study used covariate-constrained randomisation. We generated 10,000 candidate allocations. Each candidate simultaneously assigned all 60 schools while maintaining the zone-level quotas. For each allocation, we calculated the Mahalanobis distance for every pair of study arms. The calculation considered all balancing covariates and accounted for correlations between them. We used the fifth percentile of the maximum Mahalanobis-distance distribution as the acceptance threshold. This produced about 500 well-balanced allocations. We then randomly selected one allocation from this set. Random selection from the acceptable set preserves the reference distribution required for valid permutation-based inference. We did not use the conventional chi-squared threshold. Preliminary analysis showed that it accepted all 10,000 candidate allocations and was therefore equivalent to simple randomisation. We used three covariates to assess balance. These were the HWISE water-insecurity score, the infrastructure vulnerability deficit score and the school enrolment score. We normalised each measure to a scale from zero to one. The standardised mean differences between Phase 1 and the pooled comparison group were -0.10 for water insecurity, -0.07 for infrastructure vulnerability and -0.24 for enrolment. We also examined several variables that were not used to constrain randomisation. These included proximity to a health facility, girls' share of enrolment, the pupil-teacher ratio and Primary Leaving Examination scores. These variables were closely balanced across the study arms. We use 0.25 as the reporting benchmark for individual standardised mean differences. This threshold is appropriate for a small trial with prespecified ANCOVA adjustment. Achieving differences below 0.10 across all variables was not considered feasible with only 10 pure control schools. Randomisation was completed before baseline data collection. This allowed implementing partners to scope the infrastructure requirements of Phase 1 schools and begin procurement. INTERVENTION Schools in Phase 1 and Phase 2 receive the same intervention package. The two arms differ only in implementation timing. Phase 1 schools begin receiving the intervention in 2026. Phase 2 schools begin after the midline survey in the first quarter of 2028. Pure control schools receive no intervention during the evaluation period. Infrastructure The intervention provides climate-resilient water systems, including motorised boreholes. Fifteen boreholes will have data loggers for remote monitoring of water levels and functionality. The intervention also constructs or rehabilitates sex-disaggregated latrines and handwashing stations. These facilities will meet national pupil-to-latrine standards. Dedicated menstrual health facilities will also be constructed. They will include enclosed changing areas, a water supply and hygienic waste-disposal systems. Community water access points will be installed next to participating schools. The package also includes groundwater monitoring, water-quality testing and surveillance. Capacity-building and social and behaviour change District Water Officers will receive training on school WASH monitoring and water-quality testing. School Management Committees and Parent-Teacher Associations will receive training on facility operation and maintenance. This will cover basic repairs, record-keeping and maintenance budgeting. Teachers and school prefects will receive training on menstrual hygiene management. The programme will establish and support School Health Clubs. It will also conduct community engagement and social and behaviour change activities. Schools and communities will receive information, education and communication materials. Selected participants will also be trained to produce menstrual health products locally. DESIGN AND MEASUREMENT The trial uses a repeated cross-sectional design. A new sample will be drawn from current enrolment records during each survey round. The study will not attempt to recontact pupils interviewed in earlier rounds. This design avoids attrition caused by grade progression and graduation. It also limits attrition associated with the high residential mobility of Karamoja's pastoral communities. The design estimates population-level effects within each school. It does not estimate changes within individual pupils over time. Each school-level sample will include 20 post-menarche girls aged approximately 11-17 years. It will also include 20 general pupils drawn from Primary 2 to Primary 7. Approximately two-thirds of the general pupil sample will be boys. Both groups will be selected systematically from school enrolment registers. Sampling will be proportional to enrolment across grades. Each sampled pupil will be matched to a parent or legal guardian. Caregivers will be interviewed separately from the pupils. We will randomly select six classroom teachers from each school's staff register. The head teacher will also be interviewed. The study will purposively select four community leaders from each school catchment. One structured WASH facility observation and one classroom roll call will be conducted in each school during every round. The study will also digitise school enrolment, attendance and examination registers. This will cover the full academic year preceding each survey. All quantitative data will be collected on tablets using SurveyCTO. The electronic instruments will include range and skip-logic checks. The instruments will be reviewed for contextual relevance before each survey round. They will then be translated into Ngakarimojong and back-translated. Trained female enumerators will administer all modules for post-menarche girls. Quality assurance will include daily high-frequency checks against prespecified range and logic rules. Supervisors will back-check approximately 10 per cent of completed interviews. Unannounced spot checks will also be conducted across all field teams. School records will be digitised through a two-part procedure during a single school visit. First, whole-school enrolment, attendance and examination registers will be photographed. The pupil name and identification columns will be physically covered before each photograph is taken. The photographs will therefore show only grade, age, sex, attendance marks and examination scores. These records will be de-identified and will not require individual parental consent. Second, we will transcribe individual register information for the approximately 40 pupils sampled in each school. This will apply only to pupils whose pupil-caregiver dyads completed the consent process and participated in the survey. We will link their register information to their study identification numbers. STATISTICAL POWER The power calculations assume 80 per cent power and a 5 per cent significance level. They consider intra-cluster correlations of 0.10 and 0.15. They also assume an R-squared of 25 per cent at the individual and school levels. At midline, Phase 1 will include 30 schools, and the pooled comparison group will include 30 schools. This comparison can detect standardised effects of approximately 0.20 to 0.23 in the full pupil sample. It can detect effects of approximately 0.22 to 0.25 among post-menarche girls. At endline, the comparison between all treated schools and pure control schools can detect effects of approximately 0.27 to 0.31 standard deviations. The larger minimum detectable effect reflects that most schools are allocated to treatment by that stage. The comparison between Phase 1 and Phase 2 can detect differential-exposure effects of approximately 0.22 to 0.26 standard deviations. Comparable WASH trials in the region have reported effects ranging from 0.17 to 0.52 standard deviations. The largest effects have generally been observed in settings with severe baseline water scarcity. ANALYSIS A detailed pre-analysis plan will specify all regression models, outcome definitions and subgroup analyses. We will register it before the first follow-up survey. The study will also conduct a prospective cost-effectiveness analysis. Programme expenditure will be coded by output and geographic scope at the transaction level. The analysis will report the cost per beneficiary reached and per unit of outcome achieved. ETHICS The study was approved by the Mildmay Uganda Research Ethics Committee on 3 September 2026 (REC-REF 2308-2026). The Uganda National Council for Science and Technology approved the study. Parents or legal guardians will provide written informed consent before any pupil interview. Pupils will then provide their own assent. Adult respondents will provide informed consent before participating. School administrators will provide permission for facility observations and access to school records. Interviews with minors will be conducted in semi-private locations away from the main school area. A designated member of school staff will remain within sight during each interview. For menstrual health modules, a female staff member will perform this role. She will remain far enough away to ensure that responses cannot be overheard. Study data will be recorded using study identification numbers. Direct personal identifiers will be stored separately in encrypted files. A child-protection referral protocol will guide responses to any disclosures requiring further action. Participants will receive in-kind compensation worth approximately UGX 3,000. This recognises their time without creating undue inducement. Policy equipoise exists regarding the intervention's educational effects. The theory of change predicts that reduced illness, better menstrual hygiene management and less time spent collecting water will improve attendance and educational achievement. However, evidence linking school WASH interventions to measurable educational gains remains limited and contested. The phase-in design follows the programme's existing implementation plan. The team determined this plan independently of the evaluation. No school has been delayed because of the evaluation design.