Announcement
Thank you for attending THSNA 2026. The virtual meeting is now closed.
Thank you for attending THSNA 2026. The virtual meeting is now closed.
Presentation Details
| Validation of the DIOAC-GIB Model for Predicting Gastrointestinal Bleeding in Anticoagulated Patients: Comparison to the Alfalfa Models for Warfarin and DOACs Abdelrahman G.Malone1, Daniel C.Malone1, Ainhoa Gomez.-Lumbreras1, Daniel M.Witt1, Guilherme Del Fiol2. 1Department of Pharmacotherapy, College of Pharmacy, University of Utah, Salt Lake City, UT, USA.23Department of Biomedical Informatics, School of Medicine, University of Utah, Salt Lake City, UT, USA |
Abstract
Background: Gastrointestinal bleeding (GIB) is among the most common and clinically significant complications associated with oral anticoagulant (OAC) therapy. Existing bleeding-risk tools—including HAS-BLED, DOAC Score, and ATRIA—provide broad estimates of major bleeding risk but offer limited specificity for GIB. The Alfalfa-Warfarin-GIB and Alfalfa-DOAC-GIB models estimate GIB risk among patients on oral anticoagulants. While they account for nonsteroidal anti-inflammatory drugs (NSAIDs) use, they do not differentiate between specific NSAIDs such as ibuprofen and ketoprofen. A novel model called the Drug Interactions in Oral AntiCoagulants for GastroIntestinal Bleeding (DIOAC-GIB) incorporates patient and medication-specific risk estimates derived from more than 500 published studies. Objectives: To validate the DIOAC-GIB model in warfarin and direct oral anticoagulant (DOAC) cohorts and compare discrimination, calibration, and operating characteristics against the Alfalfa-Warfarin-GIB and Alfalfa-DOAC-GIB models. Methods: A retrospective cohort study was conducted using data from the electronic health record (EHR) at the University of Utah Health system. Adults (≥18 years) who initiated warfarin or DOAC therapy between 2021 and 2023 were included, with a 12-month baseline period prior to OAC initiation. The primary outcome was the first GIB event occurring within one year after OAC initiation, identified using ICD-10 diagnostic codes. Predictor variables included age; comorbidities; history of prior GIB or major bleeding; renal and hepatic impairment; thrombocytopenia; alcohol use; and medication exposures, including NSAIDs, serotonin/norepinephrine reuptake inhibitors (SSRI/SNRI), systemic corticosteroids, aspirin and antiplatelet agents, and gastrointestinal protective agents. Model performance was evaluated using discrimination, binary threshold behavior, and risk-stratification patterns across score bands. For threshold-based analyses, we applied the recommended cutoffs of DIOAC-GIB ≥4 and Alfalfa score ≥2.5 to identify elevated GIB risk for both warfarin and DOAC cohorts. Results: A total of 17,122 new OAC users met inclusion criteria, comprising 1,717 (10.0%) on warfarin. One-year GIB incidence was 4.7% in warfarin users and 3.5% in DOAC users. In the warfarin cohort, DIOAC-GIB achieved an AUC of 0.62 (95% CI 0.56–0.69), which was comparable to the Alfalfa-Warfarin-GIB AUC of 0.64 (95% CI 0.58–0.71). At recommended thresholds, DIOAC-GIB demonstrated substantially higher specificity than Alfalfa-Warfarin-GIB (0.76 vs 0.48), while Alfalfa showed higher sensitivity (0.73 vs 0.42). In the DOAC cohort, DIOAC-GIB again showed comparable discrimination with an AUC of 0.62 (95% CI 0.59–0.64) versus 0.60 (95% CI 0.57–0.62) for the Alfalfa-DOAC-GIB model. For DOACs, DIOAC-GIB demonstrated higher sensitivity (0.48 vs 0.34), with modest tradeoffs in specificity (0.72 vs 0.80). Across both anticoagulant groups, observed GIB rates increased consistently with higher DIOAC-GIB scores, indicating stable and clinically meaningful risk stratification. Among DOAC users, GIB incidence rose from 2.8% in the lowest DIOAC-GIB score category to 9.7% in the highest; among warfarin users, rates increased from 3.6% to 37.5%, demonstrating strong calibration in real-world data. Conclusions: The DIOAC-GIB model demonstrated stable predictive performance and clinically meaningful risk stratification. By incorporating detailed medication-specific risk coefficients—including individual NSAIDs, SSRIs, antiplatelets, corticosteroids, and gastroprotective therapy, DIOAC-GIB provides a nuanced, medication-responsive framework for estimating GIB risk in anticoagulated patients.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.