Unanticipated intensive care unit admission (UIA) associated with anesthesia served as an outcome measure to assess the quality of anesthesia care in a large teaching hospital. We characterized the patient population and types of problems associated with UIAs, attempted to identify patterns of care that could have led to specific adverse outcomes, and determined if a specific intervention, pulse oximetry, reduced UIAs. During a consecutive 65-wk period (July 1985-September 1986), 17,093 surgical patients were expected to enter the recovery room and then return to floor care. Seventy-one patients (0.42%) experienced a UIA from either the recovery room or operating room, and the circumstances of their admissions were analyzed in detail. After introduction of pulse oximetry in all anesthetizing locations (not including the recovery room) in the 29th week, the overall rate of UIAs and, specifically, the rate of UIA to rule out myocardial infarction, decreased significantly. Increasing ASA physical status (ASA-PS) and age significantly increased the probability of UIA. UIA patients with ASA-PS III/IV had a significantly higher acuity in the intensive care unit and were far more likely to die during their hospitalization than ASA-PS I/II patients. Retrospective review of UIAs alone did not identify patterns of care requiring remediation, which leads us to question the utility of UIAs as a generic screen for quality assurance.