Automated Identification of Complex Percutaneous Coronary Intervention from Cardiac Catheterization Reports Using Large Language Models
- 1Yale University
Abstract
IntroductionManual abstraction of complex percutaneous coronary intervention (PCI) variables from cardiac catheterization reports is labor-intensive and limits scalable cardiovascular research. We evaluated open-weight large language models (LLMs) for automated complex PCI phenotyping.
MethodsWe evaluated three LLMs (Llama 3.3 70B, Meditron-7B, and BioMistral-7B) using manually annotated catheterization reports from three hospitals within Yale New Haven Health. Models identified PCI reports and extracted six complex PCI features: 3 vessels treated, [≥]3 lesions treated, bifurcation PCI with two stents, chronic total occlusion, [≥]3 stents, and total stent length [≥]60 mm.
ResultsAmong 1,412 clinical notes, 596 were PCI reports. Llama 3.3 70B outperformed the smaller domain-specific models across most tasks. For PCI identification, Llama 3.3 70B achieved 100.0% sensitivity, 93.8% specificity, 96.4% accuracy, and 95.9% F1 score. Among 590 evaluable PCI reports (excluding 6 indeterminable cases due to missing variables) for complex PCI classification, Llama 3.3 70B achieved 97.7% sensitivity, 80.1% specificity, 57.6% positive predictive value, 99.2% negative predictive value, 83.9% accuracy, and 72.5% F1 score. Performance was higher for explicitly documented variables, including stent number and length, and lower for variables requiring interpretation across procedural details, including lesion count, bifurcation PCI, and chronic total occlusion PCI. Llama 3.3 70B had the highest accuracy at each site for complex PCI classification but significant site-level heterogeneity was observed.
ConclusionA high-capacity open-weight LLM accurately extracted complex PCI variables from unstructured reports and outperformed smaller domain-specific models. These findings support the potential use of locally deployable LLMs for scalable automated PCI phenotyping.
Related
- Artificial Intelligence‐Assisted Automated Optical Coherence Tomography Analysis for Percutaneous Coronary Intervention Planning: From Plaque Analysis to Precision Lesion Preparation and Stent Optimization
- Deep dynamic coronary roadmap guidance in percutaneous coronary intervention: a deep learning approach
- From coronary computed tomography angiography to the cath lab: computed tomography-guided percutaneous coronary intervention of a high-risk left anterior descending artery lesion
- The impact of the SYNTAX score on the incidence of in-hospital cardiac arrest in patients with acute coronary syndrome undergoing percutaneous coronary intervention: a retrospective cohort study
- Assessing Calcified Coronary Lesions to Guide Percutaneous Coronary Intervention: From Angiography to Intracoronary Imaging
BibTeX
@misc{bhatt2026automated,
title = {{Automated Identification of Complex Percutaneous Coronary Intervention from Cardiac Catheterization Reports Using Large Language Models}},
author = {N. Bhatt and F. Warner and J. Miao and R. Thakker and G. Joodi and P. Cantero-Schaffer and C. Huang and H. Krumholz and K. Murugiah},
howpublished = {medRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.05.26359802},
url = {https://www.medrxiv.org/content/10.64898/2026.08.05.26359802v2},
note = {CC BY},
}
RIS
TY - UNPB
TI - Automated Identification of Complex Percutaneous Coronary Intervention from Cardiac Catheterization Reports Using Large Language Models
AU - N. Bhatt
AU - F. Warner
AU - J. Miao
AU - R. Thakker
AU - G. Joodi
AU - P. Cantero-Schaffer
AU - C. Huang
AU - H. Krumholz
AU - K. Murugiah
PY - 2026
DA - 2026/08/05
PB - medRxiv
DO - 10.64898/2026.08.05.26359802
UR - https://www.medrxiv.org/content/10.64898/2026.08.05.26359802v2
AB - IntroductionManual abstraction of complex percutaneous coronary intervention (PCI) variables from cardiac catheterization reports is labor-intensive and limits scalable cardiovascular research. We evaluated open-weight large language models (LLMs) for automated complex PCI phenotyping. MethodsWe evaluated three LLMs (Llama 3.3 70B, Meditron-7B, and BioMistral-7B) using manually annotated catheterization reports from three hospitals within Yale New Haven Health. Models identified PCI reports and extracted six complex PCI features: 3 vessels treated, [≥]3 lesions treated, bifurcation PCI with two stents, chronic total occlusion, [≥]3 stents, and total stent length [≥]60 mm. ResultsAmong 1,412 clinical notes, 596 were PCI reports. Llama 3.3 70B outperformed the smaller domain-specific models across most tasks. For PCI identification, Llama 3.3 70B achieved 100.0% sensitivity, 93.8% specificity, 96.4% accuracy, and 95.9% F1 score. Among 590 evaluable PCI reports (excluding 6 indeterminable cases due to missing variables) for complex PCI classification, Llama 3.3 70B achieved 97.7% sensitivity, 80.1% specificity, 57.6% positive predictive value, 99.2% negative predictive value, 83.9% accuracy, and 72.5% F1 score. Performance was higher for explicitly documented variables, including stent number and length, and lower for variables requiring interpretation across procedural details, including lesion count, bifurcation PCI, and chronic total occlusion PCI. Llama 3.3 70B had the highest accuracy at each site for complex PCI classification but significant site-level heterogeneity was observed. ConclusionA high-capacity open-weight LLM accurately extracted complex PCI variables from unstructured reports and outperformed smaller domain-specific models. These findings support the potential use of locally deployable LLMs for scalable automated PCI phenotyping.
ER -