Background: Clinicians treating coronary revascularization patients need to be able to identify those who require more intensive medical therapy or follow-up. However, predictors of survival after coronary revascularization are often reported in terms of hazard ratios, which are accurate but difficult to convert to concrete values. We sought to develop a more practical and user-friendly method of predicting long-term survival in revascularization patients. Methods: We used a decision-tree induction algorithm to retrospectively examine all-cause mortality during 3-year follow-up in 3331 consecutive patients with multivessel or single proximal left anterior descending coronary artery disease who underwent an isolated first revascularization by either coronary stenting or coronary artery bypass graft between 1995 and 1999. Results: Recursive partitioning of the derivation cohort by the algorithm indicated that the best single predictor of long-term mortality was history of congestive heart failure, followed by age greater than 65 years and the presence of renal insufficiency. With these three variables, patients were readily stratified into low-, intermediate-, and high-risk groups whose 3-year mortality risks ranged from 2.0% to 18.8%. Logistic regression revealed nine significant predictors of 3-year mortality, including two interaction terms. Areas under the receiver operation characteristic curve for prediction of 3-year mortality were not significantly different between the decision tree and the logistic regression models [0.72 (95% confidence interval, 0.69 to 0.75) vs. 0.76 (95% confidence interval, 0.73 to 0.80)]. Conclusions: Long-term mortality risk in coronary revascularization patients can be estimated from three predictors that are easily obtained in clinical settings.
- Coronary arteriosclerosis
- Myocardial revascularization
ASJC Scopus subject areas
- Cardiology and Cardiovascular Medicine