Accuracy of National Surgical Quality Improvement Program Risk Calculator Among Elderly Patients Undergoing Pancreas Resection

Introduction: A preoperative goals-of-care discussion is essential in maintaining the autonomy of older adults who require surgery. The purpose of this study was to determine the accuracy of the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) risk calculator and its association with age for patients who underwent pancreatectomy. Methods: Using the American College of Surgeons NSQIP database, patients who underwent pancreatectomy between 2012 and 2015 were identified. Age was categorized into three groups: 18-64, 65-79, and 80-89 y. Analysis of variance and Pearson correlation coefficients were employed to assess differences between age categories in predicted and actual mortality and morbidity. Covariate-adjusted logistic regression models were employed to evaluate associations while accounting for potential confounders. Results: A total of 17,906 patients were included. The correlation between actual and predicted mortality was low (r = 0.14, P < 0.001). This correlation was weakest for the age category 80-89 y (r = 0.04, P = 0.07) and strongest for 65-79 y category (r = 0.14, P > 0.001). The correlation was weakest among patients who underwent pancreatoduodenectomy (r = 0.06, P = 0.08) and in this group mortality was overestimated for older adults in the age group 80-89 (actual mortality: 3.2% versus predicted mortality: 5.6%, P = 0.08). After adjusting for covariates, the interaction term between age and predicted mortality (P = 0.0021) indicated that the relationship between predicted and actual mortality is significantly influenced by patient age. Conclusions: The NSQIP risk calculator appears to overestimate mortality and morbidity risk for elderly patients undergoing pancreatoduodenectomy. These predictions should be used with caution in preoperative goals-of-care discussions with patients aged 80 y and older.