A pediatric cardiac surgeon performed 104 neonatal arterial switch operations for transposition of the great arteries with or without ventricular septal defect between June 1987 and February 1993. Initial euphoria on having only one death in the first 52 patients gave way to increasing concern when patients 53, 55, 59, 63, 64, 67, and 68 died. Sensing a problem, the surgeon visited a low-risk institution after patients 55 and 64 had died and then decided to re-train after patient 68 died. One death has occurred since. To find out whether the cluster of failures could have been related to chance alone, to variability of risk factors across time, or to suboptimal performance, we conducted the following analyses: First, identification of trends with the cumulative sum procedure was undertaken and actual mortality compared with the mortality predicted from an equation derived from a multi-institutional study. Second, logistic regression analysis of risk factors was done. If a mechanism of continuous monitoring had been in place, unfavorable trends and a need for change in protocol would have been detected earlier. Retrospective risk factor analysis suggested an excessive risk for patients with origin of the circumflex or left anterior descending coronary arteries from sinus 2 and a protective effect of phenoxybenzamine. However, about half of the risk associated with the cluster of failures was not accounted for by the variables analyzed. There was therefore an indication of suboptimal performance that appears to have been neutralized by retraining.