Retrospective analysis of vascular surgery complication rates compared to commonly used risk index calculators.

Abstract

Risk index calculator models are widely used to estimate perioperative risk of surgical procedures, directly impacting surgical planning and patient outcomes. This study assessed the accuracy of the National Surgical Quality Improvement Program Surgical Risk Calculator (NSQIP-SRC), Revised Cardiac Risk Index (RCRI) calculator, and Vascular Quality Initiative Cardiac Risk Index (VQI-CRI) calculators in predicting adverse event rates for common vascular procedures. A retrospective cohort of all carotid endarterectomy (CEA), endovascular abdominal aortic aneurysm repair (EVAR), infrainguinal bypass, open abdominal aortic aneurysm (AAA) repair, and suprainguinal bypass procedures completed in a single hospital system between January 2020 and January 2023 was assembled. Preoperative demographics, medical history, and postoperative adverse events were collected through chart review. Preoperative data were entered into each calculator to produce a predicted risk of adverse events. These predicted adverse event rates were compared with the actual adverse event rates observed; comparison was conducted via χ² goodness-of-fit tests. We included 952 procedures: 348 CEA, 33 open AAA repair, 218 EVAR, 258 infrainguinal bypass, and 95 suprainguinal bypass procedures. In conglomerate, the RCRI significantly overestimated 30-day postoperative cardiac arrest or myocardial infarction (P < .0001); the VQI-CRI also significantly overestimated in-hospital postoperative myocardial infarction (P = .0035). The NSQIP-SRC significantly underestimated severe complications (P < .0001), any complications (P < .0001), urinary tract infection (P < .0001), renal failure (P = .0344), and return to the operating room (P < .0001), while significantly overestimating surgical site infection (P = .0027) and discharge to a nursing/rehabilitation facility (P < .0001). By procedure, the RCRI significantly overestimated risk for CEA (P < .0001) and EVAR (P = .0014), while significantly underestimating risk for open AAA repairs (P = .0259). The VQI-CRI significantly overestimated risk for infrainguinal bypass (P = .0038). The NSQIP-SRC significantly underestimated severe complications for EVAR (P = .0002), infrainguinal bypass (P < .0001), and suprainguinal bypass (P < .0001). The NSQIP-SRC significantly underestimated any complications for EVAR (P = .0091), infrainguinal bypass (P = .0015), and suprainguinal bypass (P = .0049). Significant variability in the accuracy of the NSQIP-SRC, RCRI, and VQI-CRI calculators was found for the prediction of adverse events for common vascular procedures. The degree of inaccuracy per calculator varied significantly based on the indexed procedure. Updated calculators inclusive of modern vascular surgery procedural data may better capture accurate predictions of adverse events to best inform vascular surgical planning.