Objective: The objective of this study was to determine the ability of current risk estimation models to predict operative mortality in patients undergoing elective aortic surgery. We hypothesize that perioperative events are more likely to predict 30-day mortality compared with these risk estimation models that incorporate only preoperative comorbidity variables.Methods: All patients who underwent elective abdominal aortic procedures (open and endovascular) between 2002 and 2016 were included in this single-center cohort study. Emergent and urgent procedures were excluded. Data were collected on patient demographics, comorbidities, intraoperative course, postoperative course, and 30-day mortality. Matched pairs survivor sampling with a 2:1 ratio was used (two survivors matched to one death by sex, age, and procedure type). Risk estimation model scores (Vascular Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity, British Aneurysm Repair [BAR], Glasgow Aneurysm Score, and Revised Cardiac Risk Index) were calculated and analyzed alongside perioperative factors and CLASSIC grade (severity classification for intraoperative adverse events). Multiple logistic regression with adjustments for covariates was used to assess the relationship between predictors and outcome.Results: A total of 2596 elective procedures were performed during the study period (open, 57.6%; endovascular, 38.9%; advanced endovascular, 1.7%; other, 1.8%). Overall 30-day mortality was 2.0% (n ¼ 53). There was a disproportionate number of deaths in female patients compared with the overall cohort (45.3% vs 21.5%; P < .0001). Intraoperative factors significantly predicted 30-day mortality, including operative time (P ¼ .036), proximal aortic clamp level (P ¼ .001), estimated blood loss (P ¼ .016), CLASSIC grade (P ¼ .0001), and postoperative reintervention rate (P < .00001). The BAR risk model had reasonable ability (C statistic, 0.76) to predict 30-day mortality risk. The other models (Vascular Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity: C statistic, 0.69; Revised Cardiac Risk Index: C statistic, 0.60; Glasgow Aneurysm Score: C statistic, 0.58) performed relatively poorly. A custom risk assessment model including preoperative factors (functional status and smoking history) demonstrated a C statistic of 0.85; however, with the inclusion of intraoperative factors (number of blood transfusions and proximal clamp time), the accuracy of the model greatly increased (C statistic, 0.98).Conclusions: With the exception of BAR, current risk prediction models do not predict 30-day mortality as well as reported in the literature. Although they are not available in preoperative decision-making, intraoperative and postoperative adverse events are significant in predicting 30-day mortality. Creating a risk prediction model that incorporates perioperative events may improve identification of at-risk patients during the postoperative course.
