Predicting complications following radical cystectomy with the ACS NSQIP universal surgical risk calculator.

Introduction: Radical cystectomy (RC) is a challenging procedure with significant morbidity, though remains the standard of care treatment for many patients with bladder cancer. There has been debate regarding the utility of universal risk calculators to aid in point-of-care prediction of complications in individual patients preoperatively. We retrospectively evaluated the predictive value of the ACS NSQIP universal surgical risk calculator in our patients who underwent RC.
Methods: A prospective cohort of patients undergoing RC was retrospectively reviewed between October 2014 and August 2017. Only patients who underwent a RC for genitourinary cancer without significant deviation from NSQIP surgery codes 51590, 51595, and 51596 (n = 29) were included. The accuracy of the risk calculator was assessed by ROC AUC and Brier scores for both NSQIP and Clavien-Dindo defined complications. Additionally, each NSQIP risk factor was individually assessed for association with postoperative complications.
Results: 223 patients who underwent open or robotic RC (n = 18) were included for analysis. Determined by AUC C-stat and Brier scores, prediction was good for cardiac complications (0.80 and 0.021), fair for pneumonia (0.75 and 0.017), poor for UTI (0.64 and 0.078), 30-day mortality (0.62 and 0.013), any complication (0.60 and 0.19) and serious complication (0.60 and 0.17). There was a significant discordance between the rate of NSQIP predicted vs. Clavien-Dindo observed any and serious complications: 28.8% vs. 67.3%, and 25.3% vs. 11.7%, respectively.
Conclusion: The NSQIP universal surgical risk calculator did not perform with enough accuracy to consider adoption into clinical practice.