This manuscript extends our previously published work (based on data from one clinic) on the association between three drinking water-treatment modalities (boiling, filtering, and bottling) and diarrhoeal disease in HIV-positive persons by incorporating data from two additional clinics collected in the following year. We conducted a cross-sectional survey of drinking water patterns, medication usage, and episodes of diarrhoea among HIV-positive persons attending clinics associated with the San Francisco Community Consortium. We present combined results from our previously published work in one clinic (n = 226) with data from these two additional clinics (n = 458). In this combined analysis we employed logistic regression and marginal structural modelling of the data. The relative risk of diarrhoea for ‘always’ vs. ‘never’ drinking boiled water was 0.68 (95% CI 0.45–1.04) and for ‘always’ vs. ‘never’ drinking bottled water was 1.22 (95% CI 0.82–1.82). Drinking filtered water was unrelated to diarrhoea [1.03 (95% CI 0.78, 1.35) for ‘always’ vs. ‘never’ drinking filtered water]. Adjustment for confounding did not have any notable effect on the point estimates (0.61, 1.35 and 0.98 for boiled, bottled, and filtered water respectively, as defined above). The risk of diarrhoea was lower among those consuming boiled water but this finding was not statistically significant. Because of these findings, the importance of diarrhoea in immunocompromised individuals, and the limitations of cross-sectional data further prospective investigations of water consumption and diarrhoea among HIV-positive individuals are needed.