Leveraging American Society of Anesthesiologists Physical Status Classification and Surgeon Risk Estimates to Stratify Surgical Risk: A Prospective Observational Study

The American Society of Anesthesiologists Physical Status Classification (ASA PS class) is generated by the anesthesiologist before surgery. It is correlated with postoperative complications but does not integrate surgery-specific considerations or intraoperative events. We sought to combine ASA PS class with surgeon-generated risk estimates to create an easily deployed and accurate postsurgical risk stratification tool.

Surgeons at one academic center were surveyed before surgery to evaluate perceived risk of postsurgery complications. ASA PS class, presurgery clinical features, and clinical postsurgery outcomes were abstracted from an institutional database and the electronic health record. Binomial regression models predicting overall 30-d morbidity were trained using presurgery clinical features, ASA PS class, and surgeon risk estimates, alone and in combination.

Surgeon risk estimates were collected from 11 surgeons for 286 patients undergoing 68 procedure types. One hundred seventy-five (61.89%) patients had ASA PS class 3 or higher. One hundred twenty (41.96%) patients were estimated to be at higher than average risk before surgery. The overall complication rate was 27.27%. ASA PS class and surgeon risk estimates predicted surgery complication with area under the receiver operating characteristic curve (AUC) 0.79 (95% confidence interval [CI] 0.71-0.86) and AUC 0.71 (95% CI 0.63-0.78), respectively. Combining ASA PS class and the surgeon risk estimate resulted in model discrimination (AUC 0.84, 95% CI 0.78-0.89) similar to that of a clinical data-based model (AUC 0.84, 95% CI 0.78-0.88). Subgroup analysis showed that attending surgeons are better able to predict postsurgery complications than senior trainees; risk estimates from both groups were improved by combination with the ASA PS class.

ASA PS class and surgeon risk estimates are independently predictive of overall 30-d morbidity. Taken together, these assessments resulted in improved anticipation of postsurgery complications with model discrimination on par with a traditional clinical data-based model. Judgment-derived assessments alone can be used to accurately predict a patient's postsurgery risk. Future research should identify scenarios where clinician judgment is especially valuable for postsurgery risk stratification and how to best integrate clinician judgment with risk stratification systems to encourage routine use of these tools and promote optimal postsurgery management.