MACHINE LEARNING MODELS FOR PREDICTION OF POSTOPERATIVE VENOUS THROMBOEMBOLISM IN GYNECOLOGICAL MALIGNANT TUMOR PATIENTS

Abstract

Objective: To develop and validate machine learning models for predicting postoperative venous thromboembolism (VTE) in patients with gynecological malignant tumors.

Methods: A retrospective cohort study was conducted involving 245 patients who underwent surgical treatment for gynecological malignancies between January 2015 and December 2020. Data on demographics, medical history, tumor characteristics, surgical details, and perioperative variables were collected. The occurrence of postoperative VTE within 30 days after surgery was the primary outcome.

Results: Out of the 245 patients, 25 (10.2%) developed postoperative VTE. The Gradient Boosting model exhibited the highest performance with an accuracy of 0.92, sensitivity of 0.84, specificity of 0.94, precision of 0.75, F1 score of 0.79, and an AUC-ROC of 0.91. Key predictors identified included history of VTE, tumor stage, duration of surgery, use of perioperative thromboprophylaxis, and preoperative D-dimer levels. The best-performing model was validated on an independent cohort, achieving an accuracy of 0.94, sensitivity of 0.85, specificity of 0.95, precision of 0.78, F1 score of 0.81, and an AUC-ROC of 0.93.

Conclusion: Machine learning models, especially Gradient Boosting, effectively predict postoperative VTE in gynecological malignant tumor patients, allowing for targeted prophylactic strategies. Future work should focus on prospective validation and model integration into clinical practice to enhance patient care.