Abstract: Venous thromboembolism (VTE) constitutes a severe postoperative complication of colorectal cancer (CRC), significantly increases patient mortality and healthcare burden. In China, the incidence of post⁃CRC VTE reaches 11.2% within one month, while adherence to guideline⁃compliant prophylaxis rates remains low at 10.3%. This highlights the critical need for precise risk assessment tools. While widely utilized generic models such as Caprini may lack CRC specificity, and the Khorana score shows moderate discrimination (C⁃statistic 0.7), emerging CRC⁃specific models (e.g., CRC⁃VTE score, AUC 0.72) and machine learning approaches (e.g., XGBoost, AUC up to 0.908) demonstrate better performance. Nonetheless, the majority of these models are derived from single⁃center retrospective data, which limits their generalizability. Consequently, there is a pressing necessity to develop and validate the dynamic population⁃tailored prediction models to optimize perioperative VTE prevention and reduce related morbidity and mortality. This review systematically examines the risk factors associated with VTE formation following CRC surgery, synthesizes research progress and clinical applications of predictive models, aims to provide evidence⁃based support for optimizing perioperative VTE prevention and treatment strategies in CRC.

Key words: Colorectal cancer, Venous thromboembolism, Risk prediction model, Risk factors, Machine learning