Abstract
Deep venous thrombosis (DVT) is a leading cause of cardiovascular-related mortality, with an increasing incidence in elderly patients. However, existing risk assessment tools remain limited for this population. This study aimed to develop and validate machine learning (ML)-based models for predicting DVT risk in elderly patients. We retrospectively analyzed data from 1226 elderly patients discharged from the cardiovascular surgery department between January 2022 and December 2023. Risk factors were identified using the least absolute shrinkage and selection operator (LASSO), and seven ML models were subsequently trained on the selected features. Optimal hyperparameters for each model were selected through grid search with ten-fold cross-validation. Logistic regression (LR) and random forest (RF) demonstrated the best performance, with areas under the receiver operating characteristic curve (AUCs) of 0.835 and 0.819, respectively. SHapley Additive exPlanations (SHAP) revealed swelling, pain, albumin (ALB), and D-dimer as key predictors. These models may facilitate accurate risk stratification in elderly patients and provide clinical decision support through an interactive web-based tool.