How accurate is the eyeball test?: a comparison of physician's subjective assessment versus statistical methods in...

In the era of increasing percutaneous treatment options for heart disease, the estimation of surgical risk has become a key factor in selecting optimal treatment strategies. Surgical risk has historically been estimated by physician's subjective assessment and more recently by statistical risk estimates.

We studied 5099 consecutive patients who underwent cardiac surgery at Minneapolis Veterans Affairs Medical Center between 1993 and 2010. Operative mortality risk was estimated statistically by the Veterans Affairs mortality risk estimate and subjectively by cardiac surgeons before surgery. Observed mortality rate was 3.3% (168 deaths) at 1 month, 7.1% (360 deaths) at 1 year, and 18.5% (942 deaths) at 5 years after surgery. Physician's risk estimate (mean [SD], 5.6% [4.4]) and statistical risk estimate (4.3% [5.1]) had modest correlation (c-index, 0.56; P<0.001). Both methods modestly overestimated operative mortality risk. Statistical risk estimate was significantly better than physician's risk estimate in separating patients who died from those who survived at 30 days (c-index, 0.78 versus 0.73; P=0.003), at 1 year (c-index, 0.72 versus 0.61; P<0.001), and at 5 years (c-index, 0.72 versus 0.64; P<0.001) after surgery. Physician's risk estimate was higher than statistical risk estimate in all subgroups except high-risk patients.

In patients undergoing cardiac surgery, statistical risk estimate is a better method to predict operative and long-term mortality compared with physician's subjective risk estimate. However, both methods modestly overestimate actual operative mortality risk.