Abstract
Risk analysis models are becoming more important in various aspects of the clinical setting. We have used the logistic EuroSCORE as a risk analysis model, but there is divergence between the model and actual clinical reality in our country. The Japan Score is a risk model based on the Japan Adult Cardiovascular Surgery Database and it is considered to be better reflect from Japanese clinical results. We compared the logistic EuroScore (ES) and Japan Score (JS) and their predictive accuracy, using our clinical results. Between October 2006 and June 2011, 733 operations suitable for evaluation by the Japan Score were performed at our institute. Isolated coronary artery bypass grafting (CABG) was performed in 151 cases, valve surgery (Valve) in 346 cases and aortic surgery (Aorta) in 236 cases. In these cases we calculated 30-day mortality using the EuroSCORE and JapanSCORE and compared the results and prediction accuracy, by calculating the receiver operating characteristic curve (ROC curve) and the area under the ROC curve (AUC). We also calculated 30-day mortality and morbidity by the JapanSCORE and analyzed it by the same method. In the entire group, logistic 30-day mortality by ES and JS was 7.28 and 4.05% respectively. The AUC was 0.740 and 0.806, while 30-day mortality and morbidity calculated by JS was 17.72% and the AUC was 0.646. In the CABG group the 30-day mortality by ES and JS was 5.7 and 3.18% respectively, the AUC was 0.636 and 0.770, the 30-day mortality and morbidity was 13.37% and the AUC was 0.631. In the Valve group 30-day mortality by ES and JS was 6.00 and 3.79% respectively. The AUC was 0.715 and 0.794, 30-day mortality and morbidity was 17.54% and the AUC was 0.606. In the Aorta group 30-day mortality was 10.17 and 4.99% respectively. The AUC was 0.720 and 0.827. The 30-day mortality and morbidity was 20.83% and the AUC was 0.640. The 30-day mortality calculated by JS was significantly lower than that of ES (p<0.001). The prediction accuracy of both of the ES and the JS was satisfactory but the prediction accuracy of JS was better than that of the ES. The prediction accuracy of the logistic 30-day mortality and morbidity were not as accurate as 30-day mortality. JS was a good risk analysis model not only for prediction of surgical results but also for improving surgical outcome.
