Abstract
BACKGROUND: Current risk assessment models (RAMs) for prediction of venous thromboembolism (VTE) risk in the outpatient cancer population have shown poor predictive value in many of the most common cancers. The Comparison of Methods for Thromboembolic Risk Assessment with Clinical Perceptions and AwareneSS in Real Life Patients-Cancer Associated Thrombosis (COMPASS-CAT) RAM was derived in this patient population and predicted patients at high risk for VTE even after initiation of chemotherapy. We sought to externally validate this RAM. MATERIALS AND METHODS: Patients aged ≥18 years who presented to a tertiary care center between January 1, 2014, and December 31, 2016, with invasive breast, ovarian, lung, or colorectal cancers were included. The COMPASS-CAT RAM was applied using our health system's tumor registry and variables that were identified by International Statistical Classification of Diseases and Related Health Problems-9 and -10 codes of the electronic health record and independent chart review. The primary endpoint at 6-month study follow-up was documented VTE. RESULTS: A total of 3,814 patients were included. Documented VTE at 6-month follow-up occurred in 5.85% of patients. Patients stratified into low/intermediate- and high-risk groups had VTE rates of 2.27% and 6.31%, respectively. The sensitivity, specificity, and negative and positive predictive value of the RAM were 95%, 12%, 97.73%, and 6.31%, respectively. Diagnostic accuracy via receiver operating characteristic curve was calculated at 0.62 of the area under the curve. CONCLUSION: In this large retrospective external validation study of the COMPASS-CAT RAM for VTE in patients with cancer undergoing active treatment, model discrimination was moderate and calibration was poor. The model had good negative predictive value. Further prospective validation studies-especially within 6 months of cancer diagnosis-are needed before the model can be implemented into routine clinical practice for primary thromboprophylaxis of high-VTE-risk patients with cancer with solid tumors. IMPLICATIONS FOR PRACTICE: This study provides further guidance for researchers and clinicians in determining clinical and laboratory risk factors associated with development of venous thromboembolism among the ambulatory population of patients being treated for lung, breast, colorectal, or ovarian cancer. It validates the COMPASS-CAT risk model that was developed in this cancer population and suggests that further prospective validation of the model, with more focus on patients within 6 months of their index cancer diagnosis, would likely enhance the accuracy and usefulness of this model as a clinical prediction tool.