Machine learning-driven prediction of risk factors for postoperative re-fractures in elderly OVCF patients with...

Area of Science:

• Orthopedics
• Geriatrics
• Biomedical Engineering
• Data Science

Background:

• Elderly osteoporotic vertebral compression fracture (OVCF) patients with comorbidities face high re-fracture rates (up to 52%).
• Existing risk models fail to capture complex disease interactions in this population.
• A novel machine learning (ML) framework is developed to address these limitations.

Purpose of the Study:

• To pioneer a machine learning (ML) framework for predicting postoperative re-fractures in elderly osteoporotic vertebral compression fracture (OVCF) patients with comorbidities.
• To identify novel predictors of re-fracture by integrating multidimensional patient factors.
• To improve risk stratification and postoperative care for this high-risk group.

Main Methods:

• Analysis of 560 OVCF patients with comorbidities who underwent percutaneous vertebroplasty (PVP).
• Feature engineering selected 14 variables, including scoliosis, chronic kidney disease (CKD), mental disorders, and cardiovascular comorbidities.
• Six ML models were trained and validated, with performance assessed using AUC-ROC, precision-recall curves, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) were used for interpretability.

Main Results:

• The Random Forest (RF) model demonstrated superior predictive performance (test AUC = 0.88, sensitivity = 0.77, specificity = 0.87).
• Scoliosis (SHAP = 0.14), mental disorders (0.12), and CKD (0.10) were identified as the top three risk factors.
• Decision curve analysis (DCA) confirmed high clinical utility, with RF offering the greatest net benefit.

Conclusions:

• Machine learning offers a transformative approach for re-fracture prediction in OVCF patients with comorbidities.
• Underappreciated risk factors like scoliosis, mental disorders, and CKD require integrated management.
• This ML framework enables personalized risk stratification and enhances postoperative care strategies.