A Machine-Learning-Based Clinical Decision Model for Predicting Amputation Risk in Patients with Diabetic Foot Ulcers
Summary: Retrospective study (n=149 hospitalized DFU patients) developed a support vector machine (SVM) model to predict lower-limb amputation risk using CRP and Wagner grade as key predictors. Model achieved AUC 0.89, accuracy 82.4%, sensitivity 79.6%, specificity 86.5%; outperformed LDA/KNN. Decision curve analysis showed net benefit at 30% threshold. Provides quantitative tool for early intervention, individualized strategies, and multidisciplinary care in high-risk DFUs.
Key Highlights:
- Predictors: CRP and Wagner grade (independent via LASSO, p<0.05).
- Performance: AUC 0.89; moderate calibration (p=0.012); net benefit 0.351.
- Validation: 5-fold cross-validation; 82.4% correct classification.
- Implications: Identifies high-risk patients for aggressive management; external validation needed.
- Authors: Lei Gao, Zixuan Liu, Siyang Han et al.
Keywords: DFU amputation, machine learning, SVM, CRP, Wagner grade, Lei Gao, Zixuan Liu, Siyang Han