Abstract / Summary
Background. Safely managed drinking-water monitoring under Sustainable Development Goal target 6.1 uses source-water quality, yet Escherichia coli is common in household drinking water in Bangladesh despite near-universal improved sources. We compared source and household point-of-use (POU) E. coli detection within households and estimated how the safely managed estimate changes if POU replaces source E. coli. Methods. Cross-sectional secondary analysis of the Bangladesh MICS 2025 water-quality subsample (February-June 2025; detection >=1 colony-forming unit/100 mL). We classified 5,944 households with readable paired results into four joint patterns and, exploratorily, examined POU detection among negative-source households and two-week diarrhoea in children under five with survey-weighted modified Poisson models. Results. E. coli was detected in 47.4% of source and 84.8% of POU samples; only 12.7% of households were negative at both. Among negative-source households, 75.8% had E. coli in household water, versus 94.7% with a detected source, persisting (33.3%) even direct from the source. The JMP-aligned safely managed estimate was 42.5% (95% CI 40.8-44.1); substituting POU for source E. coli gave 12.6% (11.4-13.8), not an SDG indicator estimate. The diarrhoea prevalence ratio for POU detection (1.04, 0.59-1.86) was too imprecise to interpret. In exploratory spatial analysis, POU detection clustered geographically (Moran's I 0.16 cluster level, 0.31 district level, both p<0.001) and was independently associated with lower elevation (odds ratio 0.62, 0.56-0.70) and longer travel time to healthcare (1.65, 1.29-2.12); five districts had complete detection among 84-94 households tested. Conclusions. Source and POU classifications differed substantially; the survey cannot show where or when contamination occurred. Contamination was spatially clustered, arguing against pure noise and pointing to districts for follow-up. The findings support evaluating standardised POU sampling as a complement to source-based monitoring, with targeted follow-up and cross-season, cross-country replication.