External validation of a novel cancer-associated venous thromboembolism risk assessment score in a safety-net hospital

Background: Cancer-associated thrombosis (CAT) is a leading cause of death in patients diagnosed with cancer. However, pharmacologic thromboprophylaxis use in cancer patients must be carefully evaluated due to a 2-fold increased risk of experiencing a major bleeding event within this population. The electronic health record CAT (EHR-CAT) risk assessment model (RAM) was recently developed, and reports improved performance over the widely used Khorana score. Extensive RAM external validation is crucial to determine accuracy across diverse patient populations prior to clinical utilization. Objectives: To externally validate EHR-CAT using data from 2103 patients with cancer at the Boston Medical Center (BMC), New England's largest safety-net hospital, and to compare this RAM with the Khorana score. Methods: We conducted a retrospective study of BMC cancer patients diagnosed between January 2014 and December 2022 using data from the BMC tumor registry and EHR system. We validated the RAM using measures of discrimination and calibration. Results: The EHR-CAT score exhibited a strong ability to discriminate the risk of CAT (C statistic, 0.67), which was substantially higher than the classic Khorana score (C statistic, 0.58). This increased discrimination power reflects the 20% of patients that were reclassified into high or low risk by the expanded score. Model calibration was also strong in this dataset. Conclusion: In our external validation, the recently published EHR-CAT score showed clear and improved separation of patients at high and low risk for CAT. The utilization of this expanded CAT score could facilitate improved targeting of at-risk cancer patients for prophylactic therapy.