Related Experiment Video
Updated: Dec 9, 2025
A New Hybrid Quantitative Evaluation Model for Axillary Junctional Hemorrhage in Swine
Published on: December 6, 2024
Development and Implementation of an In-Hospital Bleeding Risk Model for Percutaneous Coronary Intervention
Jacob A Doll1, Colin I O'Donnell2, Meg E Plomondon2
1VA Puget Sound Health Care System, Seattle, WA, United States of America; University of Washington, Seattle, WA, United States of America.
Insights
A new bleeding risk model for percutaneous coronary intervention (PCI) accurately predicts in-hospital bleeding. This Veterans Affairs model shows improved accuracy over existing tools, aiding clinical decisions and patient care.
Area of Science:
- Cardiology
- Interventional Cardiology
- Health Informatics
Background:
- Bleeding is a frequent complication of percutaneous coronary intervention (PCI).
- Poor bleeding risk prediction leads to worse outcomes and increased healthcare costs.
- Improved prediction can guide strategies like radial access to reduce bleeding.
Purpose of the Study:
- To develop and validate a novel bleeding risk prediction model for patients undergoing PCI.
- To compare the discriminatory ability of the new model against the existing National Cardiovascular Data Registry (NCDR) CathPCI bleeding risk model.
Main Methods:
- A logistic regression model was developed using demographic, clinical, and procedural data from 107,451 patients in the Veterans Affairs Clinical Assessment, Reporting, and Tracking (VA CART) program.
- The model predicted major in-hospital bleeding events post-PCI.
- Model performance was compared to the NCDR CathPCI bleeding risk model.
Main Results:
- The study identified 12 variables associated with bleeding risk, including emergency PCI, cardiogenic shock, and atrial fibrillation.
- Factors like radial access and higher body surface area were linked to reduced bleeding risk.
- The VA CART model demonstrated superior discrimination (c-index 0.756) compared to the NCDR CathPCI model (c-index 0.707), particularly in high-risk patients.
Conclusions:
- The VA CART bleeding risk model accurately predicts post-PCI in-hospital bleeding using readily available variables.
- This model offers improved predictive performance over existing tools for the veteran population.
- Implementation facilitates point-of-care risk stratification and enhances quality assessment through better risk adjustment.
Background:
Bleeding is a common complication of percutaneous coronary intervention (PCI) that is associated with worse clinical outcomes and increased costs. Improved pre-procedural bleeding risk prediction could promote strategies that have been shown to reduce post-PCI bleeding, including increased adoption of radial access.
Methods:
We studied patients in the Veterans Affairs Clinical Assessment, Reporting, and Tracking (VA CART) program receiving PCI in VA hospitals. Logistic regression was performed to develop a model for major in-hospital bleeding using demographic, clinical, and procedural variables. The discriminatory ability of the model was compared to the existing National Cardiovascular Data Registry (NCDR) CathPCI bleeding risk model.
Results:
Among 107,451 patients treated from 2008 to 2019, 5218 (4.86%) experienced an in-hospital bleeding event. Twelve variables were associated with bleeding risk. Predictors of bleeding included emergency or salvage status, cardiogenic shock, NSTEMI, Atrial fibrillation, elevated INR, and peripheral vascular disease, while radial access, greater body surface area, and stable or unstable angina were associated with lower risk of bleeding. The developed model had superior discrimination compared with the NCDR CathPCI model (c-index 0.756, 95% CI 0.749-0.764 vs. 0.707, 95% CI 0.700-0.714, p < 0.001), especially among the highest risk patients. A web-based tool has been created to facilitate calculation of bleeding risk using this model at the point of care.
Conclusion:
The VA CART bleeding risk model uses baseline clinical and procedural variables to predict post-PCI in-hospital bleeding events and has improved discrimination compared to other available models in this patient population. Implementation of this model can facilitate risk stratification at the point of care and permit improved risk-adjustment for quality assessment.
