Determinants of adoption of household water treatment in Haiti using two analysis methods: logistic regression and machine learning.

Household water treatment (HWT) is recommended when safe drinking water is limited. To understand determinants of HWT adoption, we conducted a cross-sectional survey with 650 households across different regions in Haiti. Data were collected on 71 demographic and psychosocial factors and 2 outcomes (self-reported and confirmed HWT use). Data were transformed into 169 possible determinants of adoption across nine categories. We assessed determinants using logistic regression and, as machine learning methods are increasingly used, random forest analyses. Overall, 376 (58%) respondents self-reported treating or purchasing water, and 123 (19%) respondents had residual chlorine in stored household water. Both logistic regression and machine learning analyses had high accuracy (area under the receiver operating characteristic curve (AUC): 0.77-0.82), and the strongest determinants in models were in the demographics and socioeconomics, risk belief, and WASH practice categories. Determinants that can be influenced inform HWT promotion in Haiti. It is recommended to increase access to HWT products, provide cash and education on water treatment to emergency-impacted populations, and focus future surveys on known determinants of adoption. We found both regression and machine learning methods need informed, thoughtful, and trained analysts to ensure meaningful results and discuss the benefits/drawbacks of analysis methods herein.

Hydrogeochemical and sediment parameters improve predication accuracy of arsenic-prone groundwater in random forest machine-learning models.

The relative importance of groundwater geochemicals and sediment characteristics in predicting groundwater arsenic distributions was rarely documented. To figure this out, we established a random forest machine-learning model to predict groundwater arsenic distributions in the Hetao Basin, China, by using 22 variables of climate, topographic features, soil properties, sediment characteristics, groundwater geochemicals, and hydraulic gradients of 492 groundwater samples. The established model precisely captured the patchy distributions of groundwater arsenic concentrations in the basin with an AUC value of 0.84. Results suggest that Fe(II) was the most prominent variable in predicting groundwater arsenic concentrations, which supported that the enrichment of arsenic in groundwater was caused by the reductive dissolution of Fe(III) oxides. The high relative importance of SO42- indicated that sulfate reduction was also conducive to groundwater arsenic enrichment in inland basins. Nevertheless, parameters of climate variables, sediment characteristics, and soil properties showed secondly important roles in predicting groundwater arsenic concentrations. The other two models, which excluded parameters of groundwater geochemicals and/or sediment characteristics, showed much worse predictions than the model considering all variables. This highlights the importance of variables of groundwater geochemicals and sediment characteristics in improving the precision and accuracy of predicting results. Future studies should probe a method constructing the random forest predicting model with high precision based on the limited number of groundwater samples and sediment samples.