Using Spatial Technologies to Assess Risk Factors for Diarrheal Disease Under Environmental Variability in Bangladesh: A Machine Learning Study.

backgroundDiarrheal disease (DD) remains a major public health challenge and is the leading cause of malnutrition and the second leading cause of death among children under five globally. Although DD can be caused by a wide range of pathogens, its primary drivers are often linked to unimproved sanitation, limited access to clean drinking water, and poor hygiene practices. Low- and middle-income countries, particularly those in South Asia, experience the highest burden. These regions are also increasingly vulnerable to climate change and land use/cover changes, which may further exacerbate DD risk. However, the relative influence of environmental and social drivers at localized scales is not well understood. This gap presents a critical opportunity to identify scalable, data-informed interventions that address environmental determinants of health in the context of a changing climate.

methodsTo investigate these dynamics, we analyzed 21,779 records from the Demographic and Health Surveys (DHS) for Bangladesh, integrating them with remotely sensed data on forest cover change, temperature, and rainfall. Using Random Forest machine learning models, we assessed the relative importance of both environmental and socio-demographic variables at household and regional (village) levels.

resultsThe results show that DD risk varies across scales: household-level outcomes are primarily associated with socio-demographic characteristics, while regional-level outcomes are more strongly influenced by environmental and geographic features, including precipitation, elevation, and proximity to water bodies.

conclusionsThese findings underscore the importance of scale-sensitive approaches when assessing environmental health risks and developing climate-adaptive public health strategies.