Identification of features associated with dysmenorrhea in women aged 20-50 years using logistic regression and synthetic minority oversampling technique.

This study examined which clinical, biochemical, and lifestyle features are associated with self-reported dysmenorrhea in 1,038 Taiwanese women aged 20-50, and whether SMOTE improves detection in this imbalanced dataset. Logistic regression with SMOTE substantially increased sensitivity (83.2% vs. 51.9%) at the cost of precision and specificity, while both models achieved similar AUC values (~0.92 ROC). The authors identified six key predictors (childbirth history, age, GPT, GOT, menstrual flow volume, and diastolic blood pressure), suggesting potential use in primary care screening pending external validation.