Temporal validation of the electronic health record cancer-associated thrombosis model in patients with low bleeding risk.

Authors

Jennifer La, Omid Jafari, Kaelyn Nannini, Nhan V Do, Mary T Brophy, Nathanael R Fillmore, Ang Li

Published in

Research and practice in thrombosis and haemostasis. Volume 10. Issue 5. Pages 106830. Epub Jul 14, 2026.

Abstract

The electronic health record cancer-associated thrombosis (EHR-CAT) model was derived to predict cancer-associated venous thromboembolism (VTE). Prior validations often relied on cancer registry data with reporting delays.
We aimed to perform an external temporal validation of the EHR-CAT model among patients at low risk of bleeding using contemporary Veterans Affairs electronic health record data.
We conducted a retrospective cohort study in patients with active cancer who received systemic therapy within the national Veterans Affairs (VA) health care system from 2017 to 2024. The key inclusion criteria included: (1) having readily extractable data for EHR-CAT calculation without cancer registry data, (2) index date at first or subsequent lines of therapy, (3) exclusion of adjuvant endocrine monotherapy with low risk for VTE, and (4) application of anticoagulant clinical trial exclusion criteria to minimize bleeding risk. Patients from previously published studies were excluded. Systemic therapies were defined using the VA Corporate Data Warehouse, and one line was randomly selected. VTE outcomes were determined using validated VTE-Bidirectional Encoder Representations from Transformers natural language processing from clinical notes. Time-dependent C-statistic and calibration slope were assessed.
Among 65,341 patients with cancer receiving systemic therapy (42,865 first-line and 22,476 second-line or higher), median age was 71.9 years, and 95% were male. Most common cancers included prostate (16.1%), lung (14.8%), lymphoma (11.5%), and bladder (11.1%). The VTE incidence at 6 months was 5.4%. VTE was observed in 1.9%, 4.0%, 4.7%, 6.5%, 9.6%, and 12.3% of patients with EHR-CAT scores 0-, 1, 2, 3, 4, and 5+, respectively (C-statistic 0.67). Calibration slope was 0.876.
The EHR-CAT model, when applied across different lines of therapy, demonstrated similar performance as in previous studies.

PMID:
42568708
Bibliographic data and abstract were imported from PubMed on 08 Aug 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.