Optimizing machine learning models for predicting iron supplementation uptake among pregnant women in Somaliland: insights from the 2020 Somaliland demographic and health survey data

Iron supplementation during pregnancy is crucial, fulfilling increased demands for placental and fetal development. Despite WHO recommendations and efforts to promote iron intake, uptake remains suboptimal in many regions, including Somaliland, where maternal and child health indicators are poor due to limited healthcare access and nutritional deficiencies. This study aims to identify determinants of iron supplementation to inform targeted interventions. This cross-sectional study utilized data from the 2020 Somaliland Health and Demographic Survey (SLHDS) for a sample of 2,983 pregnant women. Explanatory variables included maternal age, education, employment status, ANC visits, residence type, region, media exposure, and wealth. Supervised machine learning models including, Logistic Regression, Random Forest, XGBoost, LightGBM, Support Vector Machine, and K-Nearest Neighbors were employed to predict iron supplementation uptake. Performance was evaluated using accuracy, precision, recall, F1-score, and AUROC. Overall, 28.83% of pregnant women reported taking iron supplements. Bivariate analysis revealed significant associations (p < 0.05) between iron supplementation and maternal age (χ2 = 15.00, p = 0.020), educational level (χ2 = 117.3, p < 0.001), employment status (χ2 = 5.5, p = 0.019), ANC visits (χ2 = 259.5, p < 0.001), region (χ2 = 103.5, p < 0.001), media exposure (χ2 = 22.3, p < 0.001), and wealth quintile (χ2 = 261.1, p < 0.001). The Random Forest model demonstrated the best performance, achieving an accuracy of 0.785 and an AUROC of 0.81. Iron supplementation uptake in Somaliland remains suboptimal, with only 28.83% reporting adherence, underscoring a critical need for enhanced interventions. The Random Forest model highlighted key predictors of iron supplementation uptake: wealth status, region, antenatal care visits, and maternal age. These findings emphasize the importance of socioeconomic factors, geographical location, and access to healthcare services in influencing iron supplementation behaviors.