Machine Learning Prediction of Iron Supplement Utilization Among Pregnant Women in Somaliland: Evidence from the Somaliland Demographic and Health Survey 2020

Abstract Background Iron supplementation during pregnancy is a key public health intervention for preventing maternal anemia and improving maternal and neonatal outcomes. This study aimed to identify the determinants and predict iron supplement utilization among pregnant women in Somaliland using machine learning techniques. Methods This study used data from the 2020 Somaliland Demographic and Health Survey (SLDHS), a nationally representative cross-sectional survey. The outcome variable was iron supplement utilization during pregnancy. Descriptive statistics and bivariate analyses were conducted to examine associations between independent variables and iron supplement use. Several supervised machine learning algorithms, including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes, and Extreme Gradient Boosting (XGBoost), were applied to predict iron supplement utilization. The dataset was split into training (80%) and testing (20%) subsets, and model performance was evaluated using accuracy, precision, recall, specificity, and area under the receiver operating characteristic curve (AUC-ROC). Results The prevalence of iron supplement utilization among pregnant women was 27%, indicating low coverage in the study population. Bivariate analysis revealed that region, educational level, wealth index, distance to a health facility, husband’s employment status, antenatal care (ANC) visits, and media exposure were significantly associated with iron supplement use (p < 0.05). Among the machine learning models, Support Vector Machine achieved the highest accuracy (82.6%), followed by Logistic Regression (81.7%) and Random Forest (80.6%). Logistic Regression (AUC = 0.853), Random Forest (AUC = 0.852), and SVM (AUC = 0.850) demonstrated the strongest discriminatory performance. Feature importance analysis indicated that ANC utilization, husband’s employment status, media exposure, and distance to health facilities were the most influential predictors of iron supplement utilization. Conclusion Iron supplementation during pregnancy remains substantially low in Somaliland. Maternal healthcare utilization, socioeconomic status, and access to health information play important roles in determining iron supplement use. Machine learning approaches demonstrated strong predictive performance in identifying key determinants of iron supplementation. Strengthening antenatal care services, improving maternal health education through mass media, and addressing geographic barriers to healthcare access may significantly improve iron supplementation coverage and maternal nutrition in Somaliland.