Predicting Complications Following Robot-Assisted Partial Nephrectomy with the ACS NSQIP ^® Universal Surgical Risk Calculator

Abstract

PURPOSE: We evaluated the predictive value of the ACS NSQIP® (American College of Surgeons National Surgical Quality Improvement Program®) surgical risk calculator in a tertiary referral cohort of patients who underwent robot-assisted partial nephrectomy. MATERIALS AND METHODS: We queried our prospectively maintained, multi-institutional database of patients treated with robot-assisted partial nephrectomy and input the preoperative details of 300 randomly selected patients into the calculator. Accuracy of the calculator was assessed by the ROC AUC and the Brier score. RESULTS: The observed rate of any complication in our cohort was 14% while the mean predicted rate of any complication using the calculator was 5.42%. The observed rate of serious complications (Clavien score 3 or greater) was 3.67% compared to the predicted rate of 4.89%. Low AUC and high Brier score were calculated for any complication (0.51 and 0.1272) and serious complications (0.55 and 0.0352, respectively). The calculated AUC was low for all outcomes, including venous thromboembolism (0.67), surgical site infection (0.51) and pneumonia (0.44). CONCLUSIONS: The ACS NSQIP risk calculator poorly predicted and discriminated which patients would experience complications after robot-assisted partial nephrectomy. These findings suggest the need for a more tailored outcome prediction model to better assist urologists risk stratify patients undergoing robot-assisted partial nephrectomy and counsel them on individual surgical risks.