Predicting iron folic acid utilization among pregnant women in Ethiopia leveraging machine learning

Iron-Folic Acid (IFA) is a crucial process indicator, that helps to halt the burden of iron deficiency anemia worldwide, especially in low and middle income countries like Ethiopia. Iron-folic acid supplementation is a crucial method in Ethiopia to reduce maternal deaths from anemia and poor birth outcomes. Despite these initiatives, there is still extremely little adherence to IFA consumption for the 90-day or more WHO-recommended timeframe. Therefore, the aim of this study was to use the DHS 2016 information to create a predictive model for IFA consumption and figure out the key factors influencing it among Ethiopian pregnant women. A secondary data analysis of Ethiopian DHS from 2016 was performed. A total weight of sample of 3109 women (15–49) was included in this study. Data were extracted using STATA version 17 software, and imported to a Jupyter notebook for further analysis. Machin learning algorithm using different classification models were implemented. All analysis and calculation were performed using Python 3 programming language in Jupyter Notebook using imblearn, sklearn, xgboost, and shap packages. Random Forest demonstrated the best performance with accuracy of 92.93%, precision of 91.32%, recall of 94.77%, F1-score of 93.99% and AUC 94%. Region, distance to health facility, age at first marriage, household head, number of ANC visit, media exposure, community literacy, husband occupation, gravidity, family size were top predicting factors of IFA utilization among women in Ethiopia. Random Forest were best predictive models with improved performance. Highly predicting factors of IFA utilization were identified using machine learning algorithms particularly random forest. These findings can guide policymakers and program managers to target interventions more effectively, promote ANC attendance, enhance community literacy and address household-level determinants to improve IFA supplementation rates. However, as the analysis is based on cross-sectional survey data, causal inferences cannot be made.