VTE

VTE predicts the risk of venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), within 60 days after surgical resection of major sellar region tumors.

Key Features:

• Data Integration: Integrates baseline characteristics, surgical findings, and postoperative laboratory tests from patient records to inform predictions.
• Machine Learning Models: Applies regression and machine learning algorithms, including linear discriminant analysis which achieved AUC 0.869 (95% CI: 0.840–0.898).
• Risk Stratification: Stratifies patients into risk categories and identifies risk factors such as age >65 years, tumor type (chordoma or craniopharyngioma), craniotomy approach, high-volume cerebrospinal fluid leakage, and extended surgical duration.
• Clinical Relevance: Reports a total 60-day VTE incidence of 3.2% with substantially higher risk in patients over 65 years and those with chordoma or craniopharyngioma.
• Performance Metrics: Reports model performance including sensitivity 92.8%, accuracy 93.6%, and specificity 61.8% for the best-performing model.

Scientific Applications:

• Neurosurgical VTE risk prediction: Predicts postoperative VTE risk for patients undergoing resection of major sellar region tumors to inform clinical assessment.
• Thromboprophylaxis decision support: Supports decision-making for postoperative care and thromboprophylaxis strategies based on individualized risk.
• Early postoperative monitoring: Identifies patients requiring intensified monitoring given a reported peak incidence of VTE within ten days after surgery.

Methodology:

Models were trained on a dataset of 3,818 patients using patient- and procedure-level variables with regression and machine learning algorithms; linear discriminant analysis emerged as the most effective model (AUC 0.869; 95% CI: 0.840–0.898).