Objective To compare the effect on prevalence of diarrhoea and mortality of household based treatment of drinking water with flocculant-disinfectant, sodium hypochlorite, and standard practices in areas with turbid water source in Africa.
Design Cluster randomised controlled trial over 20 weeks.
Setting Family compounds, each containing several houses, in rural western Kenya.
Participants 6650 people in 605 family compounds.
Intervention Water treatment: flocculant-disinfectant, sodium hypochlorite, and usual practice (control).
Main outcome measures Prevalence of diarrhoea and all cause mortality. Escherichia coli concentration, free residual chlorine concentration, and turbidity in household drinking water as surrogates for effectiveness of water treatment.
Results In children < 2 years old, compared with those in the control compounds, the absolute difference in prevalence of diarrhoea was –25% in the flocculant-disinfectant arm (95% confidence interval –40 to –5) and –17% in the sodium hypochlorite arm (–34 to 4). In all age groups compared with control, the absolute difference in prevalence was –19% in the flocculant-disinfectant arm (–34 to –2) and –26% in the sodium hypochlorite arm (–39 to –9). There were significantly fewer deaths in the intervention compounds than in the control compounds (relative risk of death 0.58, P = 0.036). Fourteen per cent of water samples from control compounds had E coli concentrations < 1 CFU/100 ml compared with 82% in flocculant-disinfectant and 78% in sodium hypochlorite compounds. The mean turbidity of drinking water was 8 nephelometric turbidity units (NTU) in flocculant-disinfectant households, compared with 55 NTU in the two other compounds (P < 0.001).
Conclusions In areas of turbid water, flocculant-disinfectant was associated with a significant reduction in diarrhoea among children < 2 years. This health benefit, combined with a significant reduction in turbidity, suggests that the flocculant-disinfectant is well suited to areas with highly contaminated and turbid water.