Risk Stratification in Unstable Angina

OBJECTIVES: To validate the Braunwald classification of unstable angina as a predictor of in-hospital cardiac complications; to determine which factors of the Braunwald classification contributed significantly to this prediction; and to devise a method of combining these predictive factors into an overall odds ratio for complications. DESIGN: A validation cohort of consecutive patients followed prospectively for in-hospital cardiac complications including myocardial infarction and death. SETTING: A community-based academic medical center. PATIENTS: A total of 393 patients admitted consecutively to the coronary care and intermediate care units with unstable angina. MAIN OUTCOME MEASURES: Major cardiac complications including death, myocardial infarction, congestive heart failure, cardiogenic shock, and severe ventricular dysrhythmias. RESULTS: Multiple logistic regression analysis identified four clinical factors used in the Braunwald classification that predicted the in-hospital occurrence of major cardiac complications: (1) myocardial infarction within less than 14 days (odds ratio OR, 5.72; 95% confidence interval CI, 1.92 to 16.97); (2) need for intravenous nitroglycerin (OR, 2.33; 95% CI, 1.31 to 4.17); (3) lack of beta-blocker or calcium channel blocker prior to admission (OR, 3.83; 95% CI, 1.55 to 9.42); and (4) baseline ST depression (OR, 2.81; 95% CI, 1.45 to 5.47). Two other clinical factors, diabetes and age, were also significant predictors. Validation of this model using parametric and nonparametric bootstrap techniques revealed excellent agreement between the CIs for adjusted ORs derived from the multiple logistic regression and those derived from the bootstrap. CONCLUSIONS: The classification of unstable angina proposed by Braunwald includes four factors that predict risk of major in-hospital cardiac complications. Specific factors used in this classification can be combined with diabetes and age to better stratify risk of major cardiac complications in this disorder using a simpler model.