Prediction of unsuccessful endometrial ablation: random forest vs logistic regression

Abstract Background Five percent of pre-menopausal women experience abnormal uterine bleeding. Endometrial ablation (EA) is one the treatment options for this common problem. However, technique shows a decrease in patient satisfaction and efficacy on long term. Study objective To develop prediction model to predict surgical re-intervention (for example re-ablation or hysterectomy) within 2 years after endometrial by using machine learning (ML). The performance developed was compared with previously published multivariate logistic regression (LR). Design This retrospective cohort study, minimal follow-up time years, included 446 (18+) that underwent an EA complaints heavy menstrual ML LR area under receiving operating characteristic (ROC) curve. Results We found out (AUC 0.65 (95% CI 0.56–0.74)) not superior 0.71 0.64–0.78)) predicting outcome EA. Based model, dysmenorrhea duration menstruation have highest impact rate. Conclusion Although techniques are gaining popularity development clinical tools, study necessarily traditional statistical techniques. Both should be considered when developing model. models can identify predictors contribute shared decision-making process practice.