Related Experiment Video
Updated: Aug 5, 2026
Patient-specific Modeling of the Heart: Estimation of Ventricular Fiber Orientations
Published on: January 8, 2013
Improved Prediction by Dynamic Modeling: An Exploratory Study in the Adult Cardiac Surgery Database of the
Sabrina Siregar1, Daan Nieboer2, Yvonne Vergouwe2
1From the Department of Cardio-Thoracic Surgery, Leiden University Medical Center, Leiden, The Netherlands (S.S., M.I.M.V.); Department of Public Health, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands (D.N., Y.V., E.W.S.); Department of Cardio-Thoracic Surgery, Radboud University Nijmegen Medical Center, Nijmegen, The Netherlands (L.N.); Department of Cardio-Thoracic Surgery, VU Medical Center, Amsterdam, The Netherlands (A.B.A.V.); and Department of Cardio-Thoracic Surgery, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands (J.J.M.T.). s.siregar@lumc.nl.
Updating the EuroSCORE model with new data improves its accuracy for predicting cardiac surgery outcomes. Dynamic modeling enhances calibration, especially for large patient groups, but requires careful application in smaller subgroups.
Area of Science:
- Cardiovascular Surgery
- Medical Statistics
- Health Informatics
Background:
- Static risk prediction models, such as the EuroSCORE, experience declining predictive performance over time.
- Continuous updating of these models is necessary to maintain and improve risk prediction accuracy.
Purpose of the Study:
- To explore and compare various methods for the continuous updating (dynamic modeling) of the EuroSCORE to enhance its predictive performance.
- To assess the impact of different updating strategies on model accuracy in cardiac surgery risk prediction.
Main Methods:
- Utilized a large dataset (n=95,240) of adult cardiac surgery from 2007-2012.
- Applied six distinct methods for updating the logistic EuroSCORE, including recalibration and Bayesian approaches.
- Evaluated model performance using discrimination (Area Under the Curve) and calibration (calibration slope, calibration-in-the-large) metrics.
Main Results:
- All dynamic updating methods demonstrated improved calibration-in-the-large compared to the original EuroSCORE.
- Calibration slope and discrimination remained consistent across most updating methods.
- In small subgroups (e.g., aortic valve replacement), extensive updating with limited data (1 year) resulted in decreased performance.
Conclusions:
- Dynamic modeling offers superior calibration and comparable discrimination to the original EuroSCORE.
- For large patient populations, various updating methods are suitable.
- For smaller subgroups, utilizing multi-year data or a Bayesian approach is recommended for robust risk prediction.
