Machine Learning and Survival Analysis Models for Postpartum Depression: A Comprehensive Risk Factor Analysis

Postpartum depression (PPD) is a widespread issue affecting countless women globally, characterized by profound feelings of sadness, anxiety, and hopelessness. While extensive research has concentrated on assessing the overall risk of PPD, there is a crucial gap exists in understanding individual onset, duration, and risk factors. This study addresses this gap by employing sophisticated Machine Learning (ML) and Survival analysis models, focusing on Edinburgh Postnatal Depression Scale (EPDS) scores 12 months post-childbirth and relevant social factors. The ML models, including Random Forest, Decision Tree, and Extreme Gradient Boosting, identified significant risk factors such as lower education, anti-depression medication usage during pregnancy, pregnancy number, and EPDS scores at specific months, with Random Forest leading with 98% accuracy. The Survival analysis, utilizing the Cox proportional hazards model and Kaplan-Meier curve, highlighted additional risk factors like gestational age, antidepressant usage, past medication, current medication, and EPDS scores at specific three months. This comprehensive approach enhances our understanding of PPD, providing crucial insights for global maternal mental health interventions.