Predicting long-term clinical mortality of elderly patients with vertebral compression fractures

Vertebral compression fractures (VCFs) are prevalent among the elderly, often leading to significant complications and mortality. We aimed to develop and validate a predictive model for long-term mortality in patients with VCFs, utilizing a comprehensive dataset from a single-center retrospective study. A total of 440 patients aged 65 years and older, diagnosed with VCFs between 2017 and 2020, were included. The participants were divided into a training set (n = 296) and a validation set (n = 144). We employed five survival analysis models: Cox Proportional Hazards, LASSO, Random Survival Forests, Gradient Boosting Machine, and Extreme Gradient Boosting (XGB). The XGB model demonstrated superior predictive performance, achieving a C-index of 0.753 with the top five predictive variables, outperforming other models in the validation set. SHAP analysis revealed age, sex, previous fracture, history of cancer, and co-morbidity as significant predictors of mortality. The model's robustness was confirmed through Kaplan-Meier survival analysis, which showed significant stratification of high- and low-risk groups (p < 0.001). Calibration curves and decision curve analysis further validated the model's clinical utility. The XGB model's interpretability was enhanced using SHAP values, providing insights into the influence of individual features on mortality risk. In conclusion, we developed a novel risk-stratification model for predicting mortality for VCFs. Our model can also aid in stratifying patients, with high discriminative ability. The use of an explainable machine learning model can aid physicians in making individualized treatment decision in VCFs patients.