Abstract 18597: How Accurate is the ‰Eyeball Test? A Comparison of Physicians' Subjective Assessment Versus Statistical Methods in Estimating Mortality Risk after Cardiac Surgery

Abstract

Introduction: In the era of increasing percutaneous options for ischemic and valvular heart disease, estimation of surgical risk has become a key factor in selecting optimal treatment strategies. Surgical risk has historically been estimated by physicians' subjective assessment and more recently by validated statistical risk scores. The aim of this study was to compare the performance of these two methods of risk estimation in cardiac surgery. Methods: We studied 5099 consecutive patients who underwent cardiac surgery at the Minneapolis VA Medical Center between 1993 and 2010. Perioperative mortality risk was estimated statistically by the national VA mortality risk score, and subjectively by cardiac surgeons. These surgeons were blinded to the statistical estimate. All estimates were recorded prior to surgery. Perioperative (30-day) and long-term mortality were ascertained. Results: Mean age was 65 +9 years and 75% of the pts underwent CABG only. Both methods overestimated mortality risk (Physician's estimate 5.6 + 4.4%; statistical estimate 4.3 + 5.1%; observed mortality 3.3%). Correlation between the two methods was modest (C = 0.56, P<0.01). Physician estimate was higher than the statistical estimate and observed mortality in all pts except those in the highest risk category (>10% mortality). Statistical estimate was significantly better than physician’s estimate in discriminating operative death vs. survival (C-index = 0.78 vs. 0.73; p=0.003) and in the long-term (C-index = 0.72 vs. 0.61; p<0.001) (Figure). Conclusions: In patients undergoing cardiac surgery, statistical risk estimate is a more accurate discriminator of operative and long-term mortality than physician’s subjective assessment. However, both methods generally overestimated observed mortality.