Unsafe drinking water is a leading cause of death and disability globally, responsible for millions of deaths and disability-adjusted life years (DALYs) annually from enteric infections. While Rwanda has significantly improved access to improved water sources, a substantial portion of households still rely on water contaminated with fecal bacteria, resulting in enteric infections being a significant cause of child mortality. Household water treatment and safe storage (HWTS) interventions, such as filtration, boiling, or chemical disinfection, offer interim solutions for obtaining safe drinking water. However, despite the disproportionate burden on the poorest households in low- and middle-income countries (LMICs), HWTS interventions often struggle to reach these populations at scale. This implementation gap stems from a lack of replicable, evidence-based models that ensure sustained coverage and use. The effectiveness of HWTS interventions depends on factors like technology acceptability, the local pathogen environment, and the intervention's delivery and promotion strategy. Rwanda's CBEHPP, implemented since 2009, aims to combat childhood diarrheal disease through hygiene promotion, but a previous study demonstrated its ineffectiveness in improving water quality or reducing diarrhea. The CBEHPP primarily promotes boiling and safe storage but household practice rates remain low. The 2014 Tubeho Neza campaign, in contrast, successfully utilized a large-scale distribution of household water filters coupled with intensive promotion, leading to improvements in water quality and child health outcomes. However, this model's sustainability was limited due to funding constraints. This study aimed to investigate whether incorporating a similar filter delivery model into the existing CBEHPP infrastructure could improve drinking water quality and child health, by leveraging CBEHPP's established national reach to distribute and promote filters in a less intensive manner than Tubeho Neza.

Existing literature highlights the effectiveness of various HWTS interventions in reducing diarrheal disease in settings with unsafe drinking water. Studies have demonstrated the positive impact of interventions such as filtration, boiling, and chemical disinfection on reducing the incidence of diarrheal illnesses. However, a significant implementation gap exists, particularly in reaching the poorest populations in LMICs. The literature points to several challenges, including the lack of replicable and evidence-based models for sustained HWTS coverage, the importance of local context in influencing acceptability and technology use, and the crucial role of delivery and promotion strategies. Studies in Rwanda specifically have shown the limitations of behavior-change communication alone in improving water quality and reducing diarrhea, suggesting a need for integrated interventions that combine behavior change with provision of effective water treatment technologies. The successful Tubeho Neza campaign provided a model demonstrating the effectiveness of large-scale filter distribution combined with intensive community engagement and household support. However, the limitations of this model, particularly in terms of long-term sustainability due to funding constraints, highlight the need for exploring alternative models which incorporate proven interventions into existing public health programs.

This study employed a cluster-randomized controlled trial (CRCT) design in Rwamagana district, Rwanda. Sixty villages were randomly selected and stratified by sector, with 30 allocated to the intervention group (CBEHPP + filter) and 30 to the control group (CBEHPP alone). Probability proportional-to-size sampling (PPS) was used for village selection based on Community Health Club (CHC) size. Within each village, 25 households were randomly selected for enrollment, prioritizing households with children under 5 or pregnant women. Baseline data were collected before intervention delivery, with midline and endline surveys conducted approximately 6 months and 13-16 months after intervention, respectively. The LifeStraw Family 2.0 filter, meeting WHO guidelines for comprehensive protection, was the intervention's core component. The intervention involved a less intensive promotion approach than the Tubeho Neza campaign, mainly relying on CHC facilitators for filter distribution, training, and limited follow-up visits. Data collection included household surveys, drinking water samples for *E. coli* detection, and caregiver-reported child health information. The primary outcome was the presence of *E. coli* in drinking water samples, categorized by WHO risk levels. Secondary outcomes were caregiver-reported diarrhea and healthcare visits for diarrhea among children under 5 and under 2. Data analysis used log-binomial generalized estimating equations (GEE) to account for village-level clustering, adjusting for socioeconomic status. Toothache prevalence served as a negative control to assess reporting bias.

The study enrolled 1,199 households, with approximately 600 in each arm. The intervention significantly reduced the proportion of households with detectable *E. coli* in drinking water samples by 20% (PR 0.80, 95% CI 0.74–0.87, *p* < 0.001). Moderate and higher fecal contamination (≥10 CFU/100 mL) was reduced by 35% (PR 0.65, 95% CI 0.57–0.74, *p* < 0.001). The intervention also resulted in a 49% reduction in the proportion of children under 5 experiencing diarrhea in the last week (aPR 0.51, 95% CI 0.35–0.73, *p* < 0.001). Similar reductions were observed in children under 2. Reported healthcare visits for diarrhea treatment were also significantly lower in the intervention group (PR 0.46, 95% CI 0.22–0.96, *p* value = 0.039). Filter coverage in the intervention group was almost universal (99%), with high reported use (95%). However, there was a decline in reported filter use and observed filter condition between midline and endline surveys, possibly attributed to seasonal variations in water access and limited support for maintenance. The negative control (toothache) showed no significant difference between groups.

The findings demonstrate that integrating household water filters into the existing CBEHPP framework effectively improves drinking water quality and reduces child diarrhea. The significant reduction in *E. coli* contamination and diarrhea prevalence strongly supports the intervention's effectiveness. The results are comparable to the more intensive Tubeho Neza campaign, showcasing the feasibility of a less resource-intensive approach. This contrasts with previous findings highlighting the ineffectiveness of the CBEHPP alone, emphasizing the importance of providing effective HWTS technology alongside behavior change communication. The observed decline in filter use and condition highlights the need for sustained support and maintenance programs. Seasonal variations in water availability may have also influenced filter use, suggesting that future interventions should account for these factors. The study's strengths include its rigorous CRCT design, large sample size, and long follow-up period. The consistency of results in both unadjusted and adjusted models strengthens the evidence of the intervention's effectiveness.

This study demonstrates the significant impact of integrating household water filters into Rwanda's CBEHPP program. The intervention resulted in substantial improvements in water quality and a significant reduction in child diarrhea. The findings provide strong evidence to support the integration of proven HWTS technologies into existing community-based health programs. Future research should focus on strategies to sustain filter use and address challenges related to seasonal variation in water access and implementation sustainability. Research is also needed to compare different implementation strategies to optimize the scalability of safe water solutions in diverse contexts.

The study's limitations include the unblinded nature of the intervention, which could have introduced bias related to reporting and household reactivity. Reliance on caregiver-reported diarrhea may not fully capture all cases of diarrhea, and cross-sectional measurement of water quality is an imperfect proxy for exposure. The observed decline in filter use over time warrants further investigation into the long-term sustainability of the intervention. Further research to better understand the reasons for filter underutilization and deterioration is needed to improve program effectiveness. Despite these limitations, the consistency of findings and the use of a negative control contribute to the strength of the study.