Bayesian modeling of Escherichia coli contamination in household drinking water in Bangladesh: evidence from the Multiple Indicator Cluster Survey 2019.

backgroundFrom a public health standpoint, there is merit in determining the levels of Escherichia coli in drinking water, but surveillance datasets often report censored values that may hinder traditional statistical analysis. This study aims to identify sociodemographic factors associated with the presence of E. coli in household drinking water in Bangladesh using Bayesian models for censored data, utilizing data from 6069 households in the Multiple Indicator Cluster Survey 2019.

methodsIn terms of censoring, we considered two different Bayesian regression strategies: Bayesian Tobit Poisson regression and Bayesian Censored Generalized Poisson regression.

resultsThe Bayesian Censored Generalized Poisson regression model was identified as the optimal model for analyzing household fecal contamination. Regression analysis revealed significant associations between household E. coli levels and various factors including division, livestock ownership, location of water sources, treatment of drinking water, household head education, wealth index, source of drinking water, place of handwashing and toilet facility. Households using tube wells had lower E. coli levels than those using other sources. Furthermore, households using pit latrines had 1.03 times higher contamination levels than those using flush latrines.

conclusionsLevels of fecal contamination in household water in Bangladesh were alarming. Our findings underscore the need for targeted policy interventions in specific population segments to address household fecal contamination, highlighting the link between sociodemographic and environmental factors with E. coli levels in drinking water.