Preoperative risk prediction and intraoperative events in cardiac surgery.

OBJECTIVE: To examine the relationship between preoperative risk prediction and intraoperative events. METHODS: A total of 3118 patients operated in 1999 and 2000 at our institution were analysed, all of whom had their EuroSCORE collected prospectively. The intraoperative variables studied were consultant or trainee operating, long bypass time, long ischaemic time, return on bypass in theatre and use of intra-aortic balloon pump at the end of the procedure. The outcomes are reported as hospital mortality, prolonged length of stay in the intensive therapy unit (pLOS-ITU, >48 h) and death or pLOS-ITU. Risk models were constructed by logistic regression for predicting these three outcomes. RESULTS: With the exception of prolonged cross-clamp time, all variables analysed were independently predictive of a negative outcome. Trainee operating had an apparent protective effect. All risk models performed well. The area under the receiver operating characteristic (ROC) curve (95% CI) increased from 0.857 (0.81, 0.90) for EuroSCORE to 0.874 (0.83, 0.92) for the risk of death model. Similarly, the area under the ROC curve for the pLOS-ITU model increased from 0.687 (0.642, 0.732) to 0.734 (0.691, 0.777) and for the death or pLOS-ITU model from 0.717 (0.677, 0.756) to 0.757 (0.719, 0.795). CONCLUSIONS: Knowledge of adverse intraoperative events enhances preoperative risk prediction. This type of analysis could be used for identifying "near miss" outcomes in adult cardiac surgery.