Predicting Basic Sanitation Service Achievement Among Rural Households in Ethiopia: An Integrated Logistic Regression and Explainable Machine Learning Approach

Access to basic sanitation remains a major public health challenge in rural sub-Saharan Africa, where open defecation and unimproved facilities sustain preventable morbidity and mortality. Despite government-led initiatives, rural sanitation coverage in Ethiopia remains critically below the Sustainable Development Goal 6 target. This community-based cross-sectional study among 842 rural households across seven Ethiopian regions aimed to identify determinants of basic sanitation service achievement and to develop predictive models using multivariable logistic regression and explainable machine learning algorithms. Only 20.9% of households achieved at least basic sanitation service, while 55.8% used unimproved facilities and 17.1% practiced open defecation. Logistic regression identified privacy motivation, perception of sanitation as a constitutional right, and access to professional guidance as the strongest positive predictors, while households motivated primarily by peer observation had significantly lower odds of service achievement. Among machine learning models, Categorical Boosting outperformed Extreme Gradient Boosting in accuracy (72.6% versus 60.1%) and balanced accuracy (74.7% versus 68.8%). Shapley Additive Explanations analysis confirmed that rights-based perception and professional consultation were the leading protective factors, while peer-driven motivation and absence of household toilet prioritization were the most influential predictors of non-achievement. Basic sanitation failure in rural Ethiopia reflects perceptual and behavioral barriers beyond infrastructure gaps. Sanitation programs should strengthen rights awareness, extension-worker engagement, and household prioritization alongside infrastructure investment.