A meta-analysis of the predictive accuracy of postoperative mortality using the American Society of Anesthesiologists' physical status classification system.

Abstract

The American Society of Anesthesiologists' physical status (ASA) tool has been applied to determine compensation, risk adjustment and risk prediction, but little is known about the accuracy and generalizability of this tool for prediction of postoperative mortality. We systematically investigated prior published reports of associations between ASA physical status and mortality to test the hypothesis that ASA physical status will have varying accuracy in prediction of postoperative mortality across surgical populations with varying surgical risk of mortality. We used random effects models and metaregression to account for heterogeneity. Combining 77 studies with 165,705 patients, the ASA physical status tool demonstrated the following pooled performance (95 % confidence intervals)--sensitivity 0.74 (0.73, 0.74), specificity 0.67 (0.67, 0.67), and area under summary receiver operating curve 0.736 (0.725, 0.747). Metaregression revealed that study death rates and surgical specialty were significant factors. ASA physical status is a better predictor of postoperative mortality in settings with lower rather than higher death rates.