Characteristics and Outcomes of Over 1 Million Veterans With Heart Failure Phenotyped Using Artificial Intelligence Approaches: the National DCVA-HF Registry

Sijian Zhang; Venkatesh K Raman; Yijun Shao; Samir Patel; Yan Cheng; Terri Elizabeth Workman; Helen M Sheriff; Gregg C Fonarow; Phillip H Lam; Paul A Heidenreich; Charity J Morgan; Hans J Moore; Jose D Vargas; Natalie M Vassall; Cherinne Arundel; Pamela E Karasik; Charles Faselis; Michael S Heimall; David Atkins; Stefan D Anker; Javed Butler; Gerasimos S Filippatos; Wen-Chih Wu; Qing Zeng-Treitler; Ali Ahmed

doi:10.1016/j.cardfail.2026.03.019

Characteristics and Outcomes of Over 1 Million Veterans With Heart Failure Phenotyped Using Artificial Intelligence Approaches: the National DCVA-HF Registry

J Card Fail. 2026 Apr 25:S1071-9164(26)00189-2. doi: 10.1016/j.cardfail.2026.03.019. Online ahead of print.

Authors

Sijian Zhang¹, Venkatesh K Raman², Yijun Shao³, Samir Patel³, Yan Cheng³, Terri Elizabeth Workman³, Helen M Sheriff³, Gregg C Fonarow⁴, Phillip H Lam⁵, Paul A Heidenreich⁶, Charity J Morgan⁷, Hans J Moore⁸, Jose D Vargas², Natalie M Vassall³, Cherinne Arundel⁹, Pamela E Karasik³, Charles Faselis¹⁰, Michael S Heimall¹¹, David Atkins³, Stefan D Anker¹², Javed Butler¹³, Gerasimos S Filippatos¹⁴, Wen-Chih Wu¹⁵, Qing Zeng-Treitler³, Ali Ahmed¹⁶

Affiliations

¹ Veterans Affairs Medical Center, Washington, DC, USA.
² Veterans Affairs Medical Center, Washington, DC, USA; Georgetown University, Washington, DC, USA.
³ Veterans Affairs Medical Center, Washington, DC, USA; George Washington University, Washington, DC, USA.
⁴ University of California, Los Angeles, CA, USA.
⁵ Georgetown University, Washington, DC, USA; Medstar Washington Hospital Center, Washington, DC, USA.
⁶ Veterans Affairs Medical Center, Palo Alto, CA, USA; Stanford University, Palo Alto, CA, USA.
⁷ Veterans Affairs Medical Center, Washington, DC, USA; University of Alabama at Birmingham, Birmingham, AL, USA.
⁸ Veterans Affairs Medical Center, Washington, DC, USA; Georgetown University, Washington, DC, USA; George Washington University, Washington, DC, USA; Uniformed Services University, Washington, DC, USA.
⁹ Veterans Affairs Medical Center, Washington, DC, USA; Georgetown University, Washington, DC, USA; George Washington University, Washington, DC, USA.
¹⁰ Veterans Affairs Medical Center, Washington, DC, USA; George Washington University, Washington, DC, USA; Uniformed Services University, Washington, DC, USA.
¹¹ HealthWell Foundation, Germantown, MD, USA.
¹² Charité Universitätsmedizin, Berlin, Germany; Wroclaw Medical University, Wroclaw, Poland.
¹³ Baylor Scott and White Research Institute, Dallas, TX, USA; University of Mississippi, Jackson, MS, USA.
¹⁴ National and Kapodistrian University of Athens, Athens, Greece; Attikon University Hospital, Athens Greece.
¹⁵ Veterans Affairs Medical Center, Providence, RI, USA; Brown University, Providence, RI, USA.
¹⁶ Veterans Affairs Medical Center, Washington, DC, USA; Georgetown University, Washington, DC, USA; George Washington University, Washington, DC, USA. Electronic address: ali.ahmed@va.gov.

PMID: 42035910
DOI: 10.1016/j.cardfail.2026.03.019

Abstract

Background: Major heart failure (HF) registries are limited by manual chart abstraction and are limited to hospitalized patients. Using artificial intelligence (AI) approaches and electronic health record (EHR) data from the Veterans Affairs (VA) health care system, we have assembled the National District of Columbia (DC) VA-HF Registry, which includes both ambulatory and hospitalized patients with HF. In the current study, we descriptively compared these patients with those identified using International Classification of Diseases (ICD) codes.

Methods: We identified 1,416,512 veterans with at least 1 ICD code for HF in the VA national EHR between 1999 and 2017. The first date of an ICD code for HF was considered the index HF date. We used validated encounter-based AI approaches based on machine learning and natural language processing models, and ICD code-based approaches on the basis of ≥1 hospitalization or ≥2 outpatient encounters due to HF, to assemble the AI-HF and ICD-HF cohorts, respectively. The 2 cohorts were compared by using absolute standardized differences, with values ≥10% indicating clinical significance. All analyses were descriptive, and no inferential or causal claims were made.

Results: The AI and ICD approaches assembled 1,031,970 and 614,828 patients, respectively, after a mean of 0.4 and 1.0 years from the index HF date. Patients in the AI-HF (vs the ICD-HF) cohort had a mean age of 71.4 (vs 70.5) years, 98.1% (vs 98.0%) were men, 13.9% (vs 16.1%) were African American, and 7.7% (vs 13.4%) had index HF hospitalizations, with absolute standardized differences of 8%, 1%, 6%, and 18%, respectively, which were <10% for 66 other baseline characteristics. The 1-year post-index HF hospitalizations occurred in 10.7% and 15.6% of patients in the AI-HF and ICD-HF cohorts, respectively. The 1‑year mortality rate was lower in the ICD-HF cohort, reflecting expected immortal‑time bias due to later cohort qualification.

Conclusions: The findings of this descriptive study demonstrate that the AI approach assembled a substantially larger cohort that included most patients identified by using the ICD code approach, suggesting broad consistency between the 2 approaches. Future external validation is needed to determine its potential use as a robust tool for improving patient care, health-services operations, and clinical research in HF.

Keywords: Artificial Intelligence; Veterans; heart failure; phenotype; registry.

Published by Elsevier Inc.