Related Experiment Videos
Risk-prediction models for mortality after coronary artery bypass surgery: application to individual patients
Pankaj Madan1, MacArthur A Elayda2, Vei-Vei Lee3
1Department of Cardiology, The Texas Heart Institute at St. Luke's Episcopal Hospital, Houston, Texas, United States.
International Journal of Cardiology
|March 6, 2010
Summary
Predictive models for cardiac surgery mortality lose accuracy over time. Recalibrating existing models or developing new ones are equally effective for contemporary patient outcomes.
Area of Science:
- Cardiovascular Surgery
- Medical Informatics
- Health Outcomes Research
Background:
- A risk-assessment model for predicting mortality after coronary artery bypass surgery was developed using 1990s patient data.
- The study evaluated the accuracy of this older model in contemporary patients and compared it with a new model and a recalibrated older model.
Purpose of the Study:
- To assess the temporal validity of an established coronary artery bypass surgery mortality prediction model.
- To compare the performance of a newly developed model and a recalibrated older model against the original model in contemporary patient cohorts.
Main Methods:
- Three mortality prediction models were created: an "old" model (1993-1999 data), a "new" model (2000-2004 data), and a "recalibrated" old model (using 2000-2004 data).
- Model performance was evaluated using area under receiver-operator characteristic (ROC) curves for discrimination and by comparing observed versus predicted mortality rates for precision.
- Models were tested on data from 2000-2004 and 2005-2007 to assess temporal applicability and compare recalibration versus remodeling strategies.
Main Results:
- The original "old" model demonstrated good discrimination (ROC, 0.80) but overpredicted mortality in later patient cohorts (2005-2007).
- The "new" and "recalibrated" models showed comparable and good discriminatory ability (ROC, 0.81 and 0.79, respectively) with good concordance between observed and predicted mortality for recent data.
- Both recalibration and remodeling strategies proved effective in maintaining prediction accuracy for contemporary patients.
Conclusions:
- Predictive models for cardiac surgery mortality require regular updates as their precision diminishes within a few years of development.
- Both recalibrating existing models with updated data and developing entirely new models are equally effective strategies for accurately predicting outcomes in current patient populations.