Development and Validation of a Machine Learning-Based Risk Prediction Model for Limb Amputation in Patients with Diabetic Foot Complications: A Retrospective Cohort Study

This retrospective cohort study analyzed 364 consecutive patients with diabetic foot ulcers treated at three Ibb University-affiliated tertiary hospitals between March 1, 2018 and December 3, 2024. We developed a Random Forest classifier incorporating 37 clinical variables spanning demographics, comorbidities, wound characteristics, laboratory parameters, and treatment factors. Model performance was evaluated using temporal validation (2024 cohort) with discrimination assessed via AUC-ROC and calibration via Brier scores.