Speakers
Description
Background: Access to safely managed sanitation services is credited with improving the health and well-being of all. However, millions of people, many in low- and middle-income countries, still have no basic sanitation facilities. This study aimed to estimate the proportion of households having safely managed sanitation service and identify its important predictors.
Methods: A cross-sectional survey was conducted in 12,400 households in Gondar City and Gorgora town Health and Demographic Surveillance System sites in northwest Ethiopia. Sociodemographic, housing, and household characteristics were collected. Data quality was assured by pretesting, training, and supervision. A multivariable binary logistic regression was fitted. Nine supervised classification machine-learning algorithms (Support Vector Machine, Random Forest, Gradient Boosting, AdaBoost, Bagging, XGBoost, LightGBM, K-Nearest Neighbors, and MLPClassifier) were trained to build a suitable model. Accuracy, precision, recall, F1 score, and the ROC curve were used to evaluate model performance.
Results: The proportion of safely managed sanitation services in households of the urban sites was 29.9% (95% CI: 29.1-30.7). Male, young, married, illiterate heads and those engaged in agricultural investment; small family households; non-poorest households; households not a member of health insurance; and houses that have good structures were more likely to have a safely managed sanitation service. The LightGBM algorithm was the best classifier (accuracy = 81.6%, precision = 81.1%, recall = 81.6%, F1 score = 81.3%) in predicting household sanitation services. Wall-building, flooring materials, wealth, occupation, education, and age were the top six important predictors of household sanitation service identified by the random forest model.
Conclusion and recommendations: The coverage of safely managed household sanitation service in urban areas of northwest Ethiopia was higher than the national average. Poor housing structure and certain household and sociodemographic characteristics were factors associated with safely managed sanitation service. Criterion-based sanitation technologies should be promoted for poor HHs, large families, and CBHI member HHs. Households that have poor housing should be consulted to construct house-friendly and low-cost basic sanitation facilities. Results from machine learning could have a significant impact on the delivery of interventions by prioritizing problems in accessing safely managed sanitation services.
