Predicting awareness of where facility-based abortion services can be accessed among women in Ethiopia using explainable machine learning
Year: 2026
Type: article
Abstract: Access to safe abortion care remains a significant public health concern in low- and middle-income countries, where unsafe abortion contributes substantially to maternal morbidity and mortality. In Ethiopia, although abortion is legally permitted under specific circumstances and facility-based abortion services are available, awareness of where these services can be accessed remains limited. Evidence on predictors of such awareness remains scarce. Therefore, this study aimed to predict women’s awareness of where facility-based abortion services can be accessed in the community and identify its key predictors using nationally representative data. Data from 8,894 women participating in the 2019 Performance Monitoring for Action (PMA) Ethiopia survey were analyzed. The data were stratified into development (80%) and independent test (20%) sets. Within each training fold, imputation, categorical encoding, Boruta feature selection, and SMOTENC were performed before model fitting to minimize information leakage. Eight supervised machine learning algorithms were developed and compared using repeated stratified 10-fold cross-validation with five repeats. Grid-search hyperparameter tuning was performed for the selected model using the development data. The final model was refitted on the full development set and evaluated on the independent test set using ROC-AUC, PR-AUC, balanced accuracy, F1-score, precision, Brier score, and calibration measures. SHapley Additive Explanations (SHAP) were used to identify features contributing to the final model’s predictions. XGBoost was the top-performing algorithm and, on the independent test set, achieved a ROC-AUC of 0.856 (95% CI 0.835–0.878), balanced accuracy of 0.784 (95% CI 0.761–0.804), and Brier score of 0.178 (95% CI 0.167–0.188), with a calibration intercept of − 0.003 and calibration slope of 0.959. SHAP analysis identified knowledge of legally permitted abortion circumstances (rape, risk to the mother’s life, and incurable fetal disease), media exposure to family planning information, family planning method use, region (Tigray), age, and parity as influential features in model predictions. The XGBoost model showed good predictive discrimination for women’s awareness of where facility-based abortion services can be accessed, with consistent performance between model development and independent test evaluation. SHAP analysis identified knowledge of legally permitted abortion circumstances, media exposure to family planning information, family planning method use, region, age, and parity as influential features in model predictions. These findings highlight potentially relevant domains for further research and for the development and evaluation strategies to improve awareness of where facility-based abortion services can be accessed in the community and related reproductive health information. However, these features represent predictive contributions within the fitted model and should not be interpreted as causal effects or independent determinants of awareness.
Source: Scientific Reports
Institutions: University of Gondar, Debre Tabor University
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7228050458
