Login to Access Video or Poster Abstract: MP54-05
Sources of Funding: None

Introduction

The American College of Surgeons National Surgical Quality Improvement Program Risk Calculator (NSQIPRC) is widely used to predict post-op morbidity and mortality in procedure specific models. No validation has been done of its accuracy for urologic surgeries. The goal of this study is to evaluate the ability of NSQIP surgical risk calculator to predict the 30-day complications, readmission, reoperation, and mortality after radical cystectomy for bladder cancer.

Methods

Medical records of all bladder cancer patients who underwent radical cystectomy between 2008 and 2015 were retrospectively reviewed. The data were collected according the NSQIP’s definitions. NSQIP calculator was used to predict the probability of complications for these patients. NSQIPRC predicted complication rates were compared to the observed complication rates in our institute using c-statistics, the area under the receiver operating characteristic (ROC) curve, and Brier’s score to measure the calibration and discrimination.

Results

We identified 439 who underwent radical cystectomy for bladder cancer. One main weakness of the NSQIPRC is its inability of predicting gastroenterological and urinary tract complications, these complications occurred in 136 (30%) and 49 (11%) patients respectively. Even after eliminating these complications from our analysis, the mean NSQIPRC predicted rates were lower than the observed rates of serious complications ( 25% vs 34%), any complications (28% vs 35%), readmission (19% vs 27%), and reoperation (4% vs 6.5%) respectively. The NSQIPRC did not perform well with area under the ROC curve was (0.563) for serious complication, (0.551) for any complication, (0.515) for readmission, and (0.587) for reoperation. However, the calculator had acceptable prediction of mortality (c-statistic 0.729, Brier’s score 0.02), and poor in predicting pneumonia (c-statistic 0.629, Brier’s score 0.022) and deep venous thrombosis (c-statistic 0.625, Brier’s score 0.032).

Conclusions

NSQIP risk factor calculator underestimates the probability of post-op morbidity. It underperforms in predicting the rates of serious complications, any complication, readmission, and reoperation. Nonetheless, it is reasonable in predicting mortality after surgery. Major improvements in data processing and calculator design are needed.

Funding

None