A Web-Based Machine Learning Calculator for Predicting Preoperative Deep Vein Thrombosis in Elderly Hip Fractures...

Area of Science:

• Geriatric Medicine
• Data Science in Healthcare
• Vascular Surgery

Background:

• Hip fractures are a significant health concern in the elderly population.
• Deep vein thrombosis (DVT) is a common and serious complication following hip fractures.

Purpose of the Study:

• To develop and validate machine learning models for predicting preoperative DVT risk in elderly patients with hip fractures.
• To identify key clinical parameters that contribute to DVT risk assessment.

Main Methods:

• Retrospective collection of clinical data from 538 elderly hip fracture patients.
• Development of logistic regression, light-gradient boosting machine, and support vector machine (SVM) models.
• Feature selection using Least Absolute Shrinkage and Selection Operator (LASSO) regression and model performance evaluation using AUC, sensitivity, Brier score, calibration curves, and decision curve analysis (DCA).

Main Results:

• LASSO identified seven key predictors: fracture type, time from injury to admission, white blood cell count, red blood cell count, C-reactive protein, D-dimer, and prothrombin time.
• The SVM model achieved the highest performance with an AUC of 0.8525 in the test set and 0.8360 in the temporal validation set.
• SHapley additive explanations (SHAP) analysis highlighted prothrombin time as the most influential predictor.

Conclusions:

• The developed SVM model demonstrates potential as an interpretable tool for assessing preoperative DVT risk in elderly hip fracture patients.
• The model can provide a quantitative reference to aid clinicians in perioperative DVT monitoring and early warning.
• A web-based calculator based on the SVM model was created for clinical utility.