Method: We analysed 107,000 end-user drinking water samples for an association between detection of Escherichia coli and drinking water system features, sample year and season using NSW Health Drinking Water Monitoring Program data, 2001–2007. We used negative binomial generalized estimating equations with adjustment for autocorrelation and clustering.
Results: We detected E. coli in over 2% of samples from 40% (129/323) of systems. E. coli detection was significantly more common in earlier years and during summer (p<0.001). On multivariate analysis E. coli detection was significantly associated with smaller systems; watercourse sources; no disinfection or disinfection with ultraviolet only; and higher post-treatment mean turbidity (all p0.01). Detection was most strongly associated with lack of disinfection (incidence rate ratio 12.6, p<0.001) and smaller supply systems (1% reduction in E. coli detection for each 1,000 person increase in supply population, p=0.004). Ultraviolet disinfection alone was the least effective disinfection method (p<0.001).
Conclusion: Even in developed countries, drinking water systems without disinfection or serving small populations appear vulnerable to the effects of faecal contamination, which presents a risk of waterborne disease outbreaks.