Effects of adding household water filters to Rwanda's Community-Based Environmental Health Promotion Programme: a cluster-randomized controlled trial in Rwamagana district

Unsafe drinking water remains a leading risk factor for global mortality and morbidity. In Rwanda, despite high coverage of improved water sources, most households’ drinking water is contaminated with fecal bacteria and enteric infections remain a major cause of under‑5 mortality. Household water treatment and safe storage (HWTS) technologies (e.g., filtration, boiling, chemical and solar disinfection, safe storage) can serve as interim solutions, but sustained coverage and correct use at scale are challenging. Rwanda’s Community‑Based Environmental Health Promotion Programme (CBEHPP) uses Community Health/Hygiene Clubs (CHCs) to promote hygienic behaviors including boiling and safe storage; however, a prior 12‑month trial found CBEHPP did not improve drinking water quality or reduce diarrhea, likely due to reliance on behavior change without provision of effective hardware. In contrast, the Tubeho Neza (“Live well”) campaign delivered LifeStraw Family 2.0 filters and improved cookstoves to low‑income households with intensive promotion and support, resulting in reduced fecal contamination in drinking water and lower reported child diarrhea and Cryptosporidium IgG responses. Policymakers are considering integrating household filters into CBEHPP using a lighter‑touch model than Tubeho Neza. This study asks whether adding a household water filter with safe storage to the CBEHPP can improve household drinking water quality and reduce child diarrhea when delivered through existing CHC structures.

Previous evidence shows HWTS interventions can reduce diarrheal disease in settings with unsafe drinking water, but scalability and sustained use in low‑ and middle‑income countries have been limited. In Rwanda, CBEHPP’s behavior‑change approach without hardware did not improve water quality or child health outcomes in a cluster‑randomized trial. Conversely, the Tubeho Neza campaign in Western Province, which provided LifeStraw Family 2.0 filters with intensive follow‑up and support, reduced detectable fecal contamination in household drinking water by 38% and caretaker‑reported child diarrhea by 29%, and lowered serologic responses to Cryptosporidium in young children. These mixed findings suggest that provision of effective HWTS hardware and implementation intensity are critical. The current study builds on this literature by testing whether a lighter‑touch integration of proven filters into an existing national CHC platform (CBEHPP) can achieve similar benefits.

Design: Cluster‑randomized controlled trial in Rwamagana district, Rwanda, with 60 villages (clusters) randomly selected by probability proportional to size and stratified by 13 sectors; 30 villages allocated to intervention (CBEHPP + LifeStraw Family 2.0 filter) and 30 to control (CBEHPP alone). Households were eligible if CHC members with at least one child under 5 or a pregnant person and an adult available to consent. From village lists (10–72 eligible households per village), up to 25 households were randomly selected per village. Enrollment totaled 1,199 households (608 intervention; 591 control) with 759 and 724 under‑5 children, respectively. Timeline: Baseline survey December 2018–March 2019; intervention delivery March–June 2019; midline 5–7 months post‑delivery (October–December 2019); endline 13–16 months post‑delivery (July–September 2020). Blinding: Allocation concealed at baseline; post‑implementation blinding not possible for enumerators/households; PI remained blinded; data analyst was not. Intervention: Delivery and promotion of LifeStraw Family 2.0 tabletop ultrafiltration systems (80 µm pre‑filter, 20 nm hollow‑fiber membrane, backwash lever, covered 5.5 L storage). The system meets WHO “comprehensive protection” criteria and has capacity up to ~18,000 L. Filters were distributed at sector health centers to all eligible households; CHC facilitators conducted initial household visits for training and a second visit ~6 months later for maintenance and reinforcement. Messaging was integrated into CHC meetings. Implementers included CRS and SNV with AEE, supported by Amazi Yego. Outcomes: Primary—detectable E. coli in 100 mL point‑of‑use drinking water sample at follow‑up visits. Water quality also categorized using WHO risk categories: ≥10 CFU/100 mL (moderate+), and ≥100 CFU/100 mL (very high). Secondary—caregiver‑reported 7‑day diarrhea and healthcare visits for diarrhea in children <5 and <2 years. Filter coverage, use, and acceptability were assessed via observation and respondent report. Water sampling and analysis: At each follow‑up, enumerators collected 100 mL drinking water samples in sodium‑thiosulfate Whirl‑Pak bags; processed within 8 hours using membrane filtration onto CompactDry media, incubated at 30°C for 24 hours. Dilutions (50, 20, 10, 5 mL) used to avoid TNTC; TNTC assigned 300 CFU. Duplicates and blanks included daily; counts normalized to CFU/100 mL. Statistics: Intention‑to‑treat. Binomial regression with log link and generalized estimating equations (GEE) with robust standard errors to account for clustering at village level. Child health models adjusted for age (months) and sex; final models additionally adjusted for government‑defined socioeconomic status (SES) due to baseline imbalance and association with diarrhea. Sensitivity analyses adjusting for handwashing access and clustering at household level showed comparable results. Power: Designed to detect 25% reduction in detectable E. coli, assuming 50% control prevalence, ICCs (village 0.14, household 0.21), two follow‑ups, 25 households/village; 60 villages targeted to ensure power with attrition. Ethics and registration: PACTR201812547047839; Emory IRB and Rwanda National Ethics Committee approvals; informed consent obtained.

Participants: 2,226 household observations and 2,455 child observations analyzed across midline and endline; slight differential attrition without evidence of bias by key characteristics. Coverage/use/acceptability (intervention arm): Filter observed in 98.5% of visits overall; in good condition in 93.0%. Reported current use 95.1%; filled in last 7 days 94.2%. Storage container had water at observation 78.2%. Drinking water sample reportedly treated by filter 87.4% overall (declined from 94.6% midline to 80.7% endline). At least one under‑5 child drank filtered water yesterday in 81.3% of households. Acceptability of appearance, smell, and taste ≥99%; acceptability of time to filter 89.8% (declined over time). Water quality (primary outcome and WHO risk categories): • Detectable E. coli (≥2 CFU/100 mL): 69.9% (649/929) intervention vs 87.0% (730/839) control; PR 0.80 (95% CI 0.74–0.87), p<0.001 (unadjusted and adjusted similar). • Moderate+ contamination (≥10 CFU/100 mL): 49.2–49.3% intervention vs 74.7–75.0% control; PR 0.65–0.66 (95% CI 0.57–0.75), p<0.001. • Very high contamination (≥100 CFU/100 mL): 22.4% intervention vs 39.8–39.9% control; PR 0.56 (95% CI 0.46–0.68), p<0.001. Mean CFU/100 mL: Intervention arithmetic mean 91.8 (95% CI 80.6–103.1), Williams mean 14.1 (95% CI 12.3–16.1); Control arithmetic mean 175.3 (95% CI 158.3–192.2), Williams mean 44.4 (95% CI 38.8–50.8). Child health: • Under‑5 diarrhea, last 7 days: 4.8–4.9% intervention vs 9.3% control; PR 0.51–0.54 (95% CI 0.35–0.78), p≤0.001; adjusted for SES aPR 0.51 (95% CI 0.35–0.73), p<0.001. • Under‑5 CHW/clinic visit for diarrhea: 1.5% vs 3.0%; PR 0.52–0.53 (95% CI ~0.27–0.98), p≈0.044–0.045. Clinic visits alone: PR 0.46–0.48, p≈0.039–0.043. • Under‑2 diarrhea, last 7 days: 8.7–8.8% vs 15.6%; PR 0.55 (95% CI 0.37–0.83), p=0.005. Negative control (toothache) showed no significant effect after SES adjustment. Temporal trends: Declines from midline to endline in observed water in storage container, reported recent filter use, and acceptability of time to filter; dry‑season shifts to unimproved sources noted.

Adding LifeStraw Family 2.0 filters to Rwanda’s CBEHPP improved point‑of‑use drinking water quality and reduced recent caregiver‑reported diarrhea and related healthcare visits among children under 5, addressing the study’s primary research question. These effects were achieved using a lighter‑touch implementation than the earlier Tubeho Neza campaign, suggesting that integrating proven HWTS hardware into existing CHC structures can yield meaningful benefits without intensive external support. The findings contrast with prior evaluations of CBEHPP’s behavior‑only approach, reinforcing the importance of providing effective WASH hardware alongside behavior change. Declines in some indicators of use and acceptability over time likely reflect seasonality (e.g., increased use of unimproved, turbid sources in the dry season), possible program slippage, and reduced implementation intensity—factors compounded by COVID‑19 restrictions affecting CHC activities. Despite these challenges, improvements in water quality and child diarrhea remained consistent across both follow‑ups, and the water quality reductions corroborate reported health effects while a negative‑control outcome showed no effect, reducing concern about pure reporting bias. The results support using CHCs as delivery platforms for HWTS at scale while highlighting the need for ongoing support to sustain use and functionality and for parallel investments in water supply infrastructure.

Integrating microbiologically effective household water filters with safe storage into Rwanda’s CBEHPP significantly improved drinking water quality and reduced recent child diarrhea in a cluster‑randomized trial, demonstrating a practical, scalable enhancement to the national CHC platform. The study contributes evidence that providing HWTS hardware through existing community structures can improve health outcomes where behavior‑only approaches have fallen short. Future work should evaluate long‑term sustainability and adherence beyond 16 months, assess strategies for routine monitoring of use and water quality, optimize implementation intensity and maintenance/repair systems, and examine scalability and cost‑effectiveness alongside continued investments in reliable water supply.

The trial was unblinded post‑implementation, and key outcomes (diarrhea, use) were caregiver‑reported, introducing potential courtesy/social desirability and recall bias; however, reductions in E. coli contamination and null effects on a negative‑control outcome mitigate these concerns. Cross‑sectional water quality measurements at visit time may not perfectly capture exposure over the prior week. Unannounced visits can still induce reactivity. Some intermediate outcomes (use/acceptability) declined over time, possibly due to seasonality, program slippage, or filter clogging with turbid water. COVID‑19 restrictions likely reduced CHC meeting frequency and implementation intensity. There were missing water samples (often due to no water available at visit), though missingness was not evidently associated with key characteristics and household size was balanced between arms. The follow‑up period was limited to 13–16 months.