Predicting pregnancy loss and its determinants among reproductive-aged women using supervised machine learning algorithms in Sub-Saharan Africa.

Background: Pregnancy loss is a significant public health issue globally, particularly in Sub-Saharan Africa (SSA), where maternal health outcomes continue to be a major concern. Despite notable progress in improving maternal health, pregnancy-related complications, including s due to miscarriages, stillbirths, and induced abortions, continue to impact women's health, social wellbeing, and economic stability in the region. This study aims to identify the key predictors of pregnancy loss and develop effective predictive models for pregnancy loss among reproductive-aged women in SSA. Methods: We derived the data for this cross-sectional study from the most recent Demographic and Health Survey of Sub-Saharan African countries. Python software was used to process the data, and machine learning techniques such as Random Forest, Decision Tree, Logistic Regression, Extreme Gradient Boosting, and Gaussian were applied. The performance of the predictive models was evaluated using several standard metrics, including the ROC curve, accuracy score, precision, recall, and F-measure. Result: The final experimental results indicated that the Random Forest model performed the best in predicting pregnancy loss, achieving an accuracy of 98%, precision of 98%, F-measure of 83%, ROC curve of 94%, and recall of 77%. The Gaussian model had the lowest classification accuracy, with an accuracy of 92.64% compared to the others. Based on SHPY values, unmarried women may be more likely to experience pregnancy loss, particularly in contexts where premarital pregnancies are stigmatized. The use of antenatal care and family planning services can significantly impact the risk of pregnancy loss. Women from lower-income backgrounds may face challenges in accessing prenatal care or safe reproductive health services, leading to higher risks of loss. Additionally, higher levels of education are often correlated with increased awareness of family planning methods and better access to healthcare, which can reduce the likelihood of unintended pregnancy loss. Conclusion: The Random Forest machine learning model demonstrates greater predictive power in estimating pregnancy loss risk factors. Machine learning can help facilitate early prediction and intervention for women at high risk of pregnancy loss. Based on these findings, we recommend policy measures aimed at reducing pregnancy loss Sub-Saharan African countries.