Abstract
CONTEXT: Recent randomised controlled trials in Bangladesh and Kenya concluded that household water treatment, alone or in combination with upgraded sanitation and handwashing, did not reduce linear growth faltering or improve other child growth outcomes. Whether these results are applicable in areas with distinct constellations of water, sanitation and hygiene (WaSH) risks is unknown. Analysis of observational data offers an efficient means to assess the external validity of trial findings. We studied whether a water quality intervention could improve child growth in a rural Indian setting with higher levels of circulating pathogens than the original trial sites. METHODS: We analysed a cross-sectional dataset including a microbiological measure of household water quality. All households accessed water from an improved source. We applied propensity score methods to emulate a randomised trial investigating the hypothesis that receipt of drinking water meeting Sustainable Development Goal (SDG) 6.1 quality standards for absence of faecal contamination leads to improved growth. Growth outcomes (stunting, underweight, wasting, and their corresponding Z-scores) were assessed in children 12-23 months of age. For each outcome, we estimated the mean and 95% confidence interval of the absolute risk difference between treatment groups. FINDINGS: Of 1088 households, 442 (40.62%) received drinking water meeting SDG 6.1 standards. The adjusted risk of child underweight was 7.4% (1.3% to 13.4%) lower among those drinking water satisfying SDG 6.1 norms than among controls. Evidence concerning the relationship of drinking water meeting SDG 6.1 norms to length-for-age and weight-for-age was inconclusive, and there was no apparent relationship with stunting or wasting. CONCLUSIONS: In contexts characterised by high pathogen transmission, water quality improvements have the potential to reduce the proportion of underweight children, but are unlikely to impact stunting or wasting. Further research is required to assess how these modelled benefits can best be achieved in real world settings.