Predicting determinants of modern contraceptive use among reproductive-age women in Ethiopia using machine learning...

Globally, around 40% of women report unintended pregnancies, with approximately 214 million women in developing countries wanting to avoid pregnancy but not using any contraception. Modern contraceptives (MCs) are effective tools for preventing unintended pregnancies, controlling rapid population growth, and reducing fertility and maternal mortality rates, particularly in developing countries. This study aimed to identify the determinants of modern contraceptive use among Ethiopian women of reproductive age using machine learning (ML) algorithms.

The study utilized secondary data from the 2019 Performance Monitoring and Accountability (PMA) Ethiopia survey, analyzing 8,837 samples. Preprocessing steps included data cleaning, feature engineering, dimensionality reduction, and splitting the data, with 80% used for training and 20% for testing the algorithms. Six supervised ML algorithms were employed and assessed using confusion matrices, with information gain applied to identify critical attributes for predicting MC use.

Only 24% of participants used modern contraceptives [95% CI (23.1%, 24.9%)]. Extreme gradient boosting (XGB) demonstrated the highest predictive accuracy (81.97%, 95% CI {79.06%, 82.7%}) and area under the ROC curve (76.63%), followed by logistic regression (80.52%) and support vector machines (80.41%). Key determinants of MC use included starting family planning at age 20 or older, being single, having partner approval, being the wife of the household head, being between 36 and 49 years old, advice from healthcare providers, concerns about side effects, and having a household size of five or more.

The use of modern contraceptives among Ethiopian women remains low. Extreme gradient boosting proved most effective in predicting determinants of MC use. Based on the results of predictive associations, improved counseling during antenatal and postnatal care visits, promoting partner discussions on family planning, and addressing concerns about family size and contraceptive use are recommended strategies to enhance MC uptake.