Risk stratification analysis of operative mortality in heart and thoracic aorta surgery: comparison between Parsonnet and EuroSCORE additive model

OBJECTIVE: Our purpose was to compare the performance of risk stratification model between Parsonnet and European System for Cardiac Operative Risk Evaluation (EuroSCORE) in our patient database. METHODS: From August 1994 to December 2000, 803 consecutive patients have undergone heart and thoracic aorta surgery using cardiopulmonary bypass and scored according to Parsonnet and EuroSCORE algorithm. The population was divided into five clinically relevant risk categories. We compared correlation of predicted mortality and observed mortality between these two models. Score validity was assessed by calculating the area under the receiver operating characteristic (ROC) curve. RESULTS: Overall hospital mortality was 4.5%. In Parsonnet model, predicted mortality was 2.4% for 0-4% risk, 6.7% for 5-9% risk, 12% for 10-14% risk, 17% for 15-19% risk, 25% for 20% plus risk, and 10.4% for overall patients. Observed mortality was 2.4, 0.4, 5.9, 8.7, 11, and 4.5%, respectively. The thoracic aorta and valve cohort indicated poor correlation between predicted and observed mortality compared to coronary cohort. In the EuroSCORE model, predicted mortality was 1.4% for 0-2% risk, 4.0% for 3-5% risk, 6.7% for 6-8% risk, 9.7% for 9-11% risk, 13% for 12% plus risk, and 5.3% for overall patients. Actual mortality was 0, 1.5, 6.8, 11, 21, and 4.5%, respectively. Each of the thoracic aorta, valve, and coronary cohort indicated good correlation between predicted and observed mortality. Areas under the ROC curves were 0.72 in Parsonnet and 0.82 in EuroSCORE. CONCLUSIONS: The EuroSCORE additive model yielded good predictive value for hospital mortality of Japanese patients undergoing not only cardiac but also thoracic aortic surgery.