S. Khaneki1,2, M. R. Bronsert1, W. G. Henderson1, M. Yazdanfar1, A. Lambert-Kerzner1, K. E. Hammermeister1, R. A. Meguid1 1University Of Colorado Denver,Surgery,Aurora, CO, USA 2Hurley Medical Center,Internal Medicine,Flint, MI, USA
Introduction:
The Surgical Risk Preoperative Assessment System (SURPAS) is a parsimonious surgical risk assessment tool integrated into our electronic health record (EHR) for the preoperative prediction of postoperative adverse events. SURPAS applies to >3000 operations in 9 surgical specialties, requires entry of 7 readily available predictor variables, and predicts outcomes of mortality, overall morbidity, unplanned readmission and 8 clusters of common complications. It was developed from the American College of Surgeons’ National Surgical Quality Improvement Program (ACS NSQIP) dataset. The objective of this study was to compare the accuracy of predictions of postoperative mortality and morbidity using SURPAS vs. the ACS NSQIP risk calculator.
Methods:
We calculated predicted preoperative risk of postoperative mortality and morbidity using both SURPAS and the ACS NSQIP risk calculator for 1,006 patients randomly selected from the ACS NSQIP database across 9 different surgical subspecialties. We calculated the relative and absolute mean and median of the risk differences and plotted histograms and Bland-Altman graphs to analyze these differences. We also compared the goodness of fit statistics for expected and observed adverse postoperative outcomes between SURPAS and the ACS NSQIP risk calculator using the c-index, Hosmer-Lemeshow analysis, and Brier scores.
Results:
The SURPAS risk estimates for mortality were slightly higher (Mean=0.64%) than the ACS NSQIP estimates (0.59%) and considerably higher for overall morbidity (10.65% vs 7.73%). The ACS NSQIP risk estimates for morbidity tended to underestimate risk compared to observed adverse postoperative outcomes, particularly for the highest risk patients. Goodness of fit statistics were similar for SURPAS and the ACS NSQIP risk calculator, except for the c-index for mortality (SURPAS c=0.853 vs ACS NSQIP c=0.937), although this finding is probably tentative because there were only 6 deaths. Hosmer-Lemeshow graphs and fit statistics for ACS NSQIP and SURPAS risk estimates vs observed adverse postoperative outcomes are shown for mortality and overall morbidity (Figure).
Conclusions:
The SURPAS risk predictions for mortality and overall morbidity are as good as those of the ACS NSQIP risk calculator. SURPAS has the advantages that it requires only one-third of the number of predictor variables as the ACS NSQIP tool, provides patient risk estimates compared to national averages for patients undergoing the same operation, is integrated into the EHR, and automatically provides a preoperative note in the patient’s medical record and a graphical handout of risks for the patients to take home.
