Multicenter study of general anesthesia. III. Predictors of severe perioperative adverse outcomes.

Little information is available about the incidence of severe adverse outcomes, and even less information is available about the identification and quantification of independent predictors of severe perioperative adverse outcomes. The purpose of this study was to identify and quantitate independent predictors of severe perioperative adverse outcomes in a prospective randomized clinical trial of general anesthesia in 17,201 patients. Twenty-nine prognostic variables for 15 severe outcomes in 847 patients were tested by multiple stepwise logistic regressions from which 20 significant (P less than 0.05) predictors were identified. A history of cardiac failure or myocardial infarction less than or equal to 1 yr; ASA physical status 3 or 4; age greater than 50 yr; cardiovascular, thoracic, abdominal or neurologic surgery; and the study anesthetics were significant predictors of "any severe outcome, including death." There were 17 significant predictors for 10 severe cardiovascular outcomes in 608 patients, including a history of ventricular arrhythmia, hypertension, cardiac failure, myocardial ischemia, myocardial infarction less than or equal to 1 yr or myocardial infarction greater than 1 yr, and smoking; ASA physical status; age; cardiovascular, thoracic, abdominal, eyes-ears-nose-throat/endocrine, neurologic, musculoskeletal, or gynecologic surgery; and the study anesthetics. There were 9 significant predictors for 4 severe respiratory outcomes in 163 patients, including a history of cardiac failure, myocardial ischemia, or chronic obstructive pulmonary disease; obesity; smoking; male gender; ASA physical status; abdominal surgery; and the study anesthetics. Colinearity between related prognostic variables (such as disease and ASA physical status) was assessed using progressively segregated groups of variables in eight stepwise logistic regressions. We conclude that the comprehensive stepwise logistic regression of 29 prognostic variables reported here provides a valid estimate of the risks of severe perioperative outcomes associated with general anesthesia.