Abstract Body: Background: Cancer-associated VTE (CAT) is the second leading cause of death in patients diagnosed with cancer. However, pharmacologic thromboprophylaxis use in cancer patients must be tightly regulated due to a two-fold increased risk of experiencing a major bleeding event within this population. A cancer-associated venous thromboembolism (CAT) risk assessment model (RAM) has recently been developed utilizing patient electronic health records (EHR) that reports improved performance over the prior state-of-the-art Khorana score.
Objective: To externally validate a recently published RAM using EHR data from 2,103 cancer patients at Boston Medical Center, New England’s largest safety-net hospital, and to compare this RAM to the Khorana score.
Methods: We conducted a retrospective study of Boston Medical Center (BMC) cancer patients diagnosed between January 2014 and December 2022 using data from the BMC tumor registry and electronic health record system. We validated the risk assessment model using measures of discrimination and calibration.
Results: The recently published score exhibited a strong ability to discriminate risk of CAT (C statistic, 0.67) substantially higher than the classic Khorana score (C statistic, 0.58). This increased discrimination power reflects the 20% of patients correctly 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 CAT score clearly and improvedly separated patients at high and low risk for CAT. This expanded CAT score could facilitate improved selection of prophylactic therapy among cancer patients.