J Int Med Res. 2026 Aug;54(8):3000605261471674. doi: 10.1177/03000605261471674. Epub 2026 Aug 13.
ABSTRACT
ObjectiveTo assess the predictive accuracy of the comparison of methods for thromboembolic risk assessment with clinical perceptions and awareness in real-life patients with cancer-associated thrombosis (COMPASS-CAT) risk model against the established Khorana score for predicting in-hospital venous thromboembolism in patients with solid tumors.MethodsThis retrospective cohort study analyzed 482 inpatients with breast, lung, colorectal, or ovarian cancer. Venous thromboembolism risk was stratified using the COMPASS-CAT and Khorana scores. The primary outcome was objectively verified in-hospital venous thromboembolism. Discriminative performance was evaluated using receiver operating characteristic curve analysis.ResultsOf the 482 patients, 53 (11.0%) developed in-hospital venous thromboembolism. The COMPASS-CAT model (score ≥7) demonstrated higher specificity (98.13% vs. 81.77%) and positive predictive value (82.22% vs. 35.00%) compared with the Khorana score (score ≥2). The COMPASS-CAT model exhibited significantly better overall discrimination, with an area under the curve of 0.913 versus 0.695 for the Khorana score (p < 0.001). Univariate logistic regression revealed that a COMPASS-CAT score ≥7 was strongly associated with elevated venous thromboembolism risk (odds ratio = 20.66, 95% confidence interval: 9.96-42.83). A Khorana score ≥2 was also associated with an increased risk, but the magnitude was weaker (odds ratio = 3.26).ConclusionThe COMPASS-CAT model exhibited a promising discriminative trend over the Khorana score for identifying the risk of in-hospital venous thromboembolism. Its notably higher specificity and positive predictive value suggest that it may serve as a helpful supplementary tool for targeted thromboprophylaxis in the inpatient oncology setting.
PMID:42593353 | DOI:10.1177/03000605261471674