American college of surgeons national surgical quality improvement program risk calculator analysis demonstrates inferiority of retrospective complications data for patients undergoing distal pancreatectomy

Abstract

Objective: To assess the accuracy of the ACS NSQIP calculator for estimating risk of complications after distal pancreatectomy for benign and malignant disease. Methods: Data regarding postoperative complications after pancreatectomy were collected from a prospectively maintained pancreas surgery consortium database at three academic medical centers. The ability of the NSQIP calculator to estimate risk of postoperative complications was assessed using the c-statistic (0.7 or above for reasonable models). Results: 254 adult patients who underwent distal pancreatectomy with or without splenectomy were included. Numbers and rates of complications were as follows: pneumonia 14, 5.6% (median predicted risk, c-statistic: 2.6%, 0.652); cardiac complications 14, 5.6% (0.70%, 0.543); surgical site infections 11, 4.4% (11.2%, 0.613); urinary tract infection 13, 5.2% (3.8%, 0.630); venous thromboembolism 11, 4.4% (0.450.3%); renal failure 14, 5.6% (0.6%, 0.590); readmission 58, 23.1% (0.602, 15.1%); return to the OR 12, 4.8% (2.7%, 0.558); death 1, 0.04% (0.5%, 0.555). Median length of stay was 5.0 (Range: 0–40) and predicted median length of stay was 7.0 (Range: 511). Conclusion: The ACS NSQIP Risk Calculator estimates the risk of pneumonia (c-statistic 0.652) reasonably well for patients undergoing distal pancreatectomy but generally underestimates risk of other postoperative complications. Retrospective collection of complication data (used to model the calculator) may not accurately reflect the true volume of complications after distal pancreatectomy.