BACKGROUND: This paper estimates the impact on childhood drowning rates of community-based introduction of crèches or playpens or both in rural Bangladesh for children aged 0-47 months. METHODS: A baseline census of the whole population of 270,387 households in 51 unions, 451 villages from 7 rural sub-districts in Bangladesh was conducted in 2013. The baseline census determined retrospective, age-specific, and cumulative drowning incidence rates (IR) experienced in the target households in the 12 months prior to the intervention. Beginning in late 2013, creches for drowning prevention were established across the study area. Acceptance into creches was provided and written assent to attend a creche was obtained for all children aged 9-47 months in all participating unions. Playpens were provided to 45,460 of these children, of which 5981 children received only the playpens. All children were followed-up until their 48-month birthday or administrative censoring (fixed timepoint to stop observing the drowning deaths), after a two-year implementation period (2014-2016). Drowning IR were estimated for children and compared to corresponding baseline rates from 2012. Age-specific drowning IR under different "as treated" categories (playpen-only, creche-only, and playpen-plus-creche) were compared to the baseline rates experienced by the categorized households prior to intervention. RESULTS: A total of 3205 creches (average of 7 creches per village) were established, and 116,054 children aged 9-47 months were exposed to the intervention packages. Aggregated drowning IRs between age 0 and 47 were estimated per 100,000 population per year at 86.73 (95% CI: 69.67-107.97) and 43.03 (95% CI: 35.55-52.10) in the baseline and post implementation period, respectively. Risk ratios were 0.40 (95% CI: 0.28-0.57) overall, and 0.34 (95% CI: 0.13-0.90), 0.09 (95% CI: 0.02-0.36), and 0.04 (95% CI: 0.002-0.60) in children under the creche-only, aged, 1, 2, and 3 years old respectively. Inexplicably, drowning rates were statistically significantly higher post-intervention in children 0-11 months. There was no mortality reduction with playpen use (alone or in combination), and this group may actually have had a higher risk of drowning. CONCLUSIONS: Creches are effective for preventing childhood drowning in rural Bangladesh for children above age 1-year, and should be considered for further scale-up.