Aims. In a previous study we have devised a method for monitoring the performance of cardiac surgeons (Lovegrove et al, Lancet 1997;350:1128-30). The aim of this study was to extend that work by developing a method for assessing whether apparently high peri-operative mortality could have occurred by chance, and displaying this graphically. Methods. Peri-operative mortalities for individual patients are calculated from a locally derived and validated risk score. From these we have calculated the probabilities that different levels of mortality would occur for a series of cases with known pre-operative risks. These have been used to construct nested prediction intervals for overall mortality, indicating how likely given performance is to have occurred by chance. Results. The VLAD plot below is for a series of 547 operations performed by a single surgeon. Overall mortality is better than expected but well within the 50% prediction intervals consistent with this being a chance occurrence. (Graph Presented).