OBJECTIVES: To develop and validate an "in house" risk model for predicting perioperative mortality following elective AAA repair and to compare this with other models. DESIGN: Multivariate logistics regression analysis was used to identify risk factors for perioperative-day mortality from one tertiary institution's prospectively maintained database. MATERIALS AND METHODS: Consecutive elective open (564) and endovascular (589) AAA repairs (2000-2010) were split randomly into development (810) and validation (343) data sets. The resultant model was compared to Glasgow Aneurysm Score (GAS), Modified Customised Probability Index (m-CPI), CPI, the Vascular Governance North West (VGNW) model and the Medicare model. RESULTS: Variables associated with perioperative mortality included: increasing age (P = 0.034), myocardial infarct within last 10 years (P = 0.0008), raised serum creatinine (P = 0.005) and open surgery (P = 0.0001). The areas under the receiver operating characteristic curve (AUC) for predicted probability of 30-day mortality in development and validation data sets were 0.79 and 0.82 respectively. AUCs for GAS, m-CPI and CPI were poor (0.63, 0.58 and 0.58 respectively), whilst VGNW and Medicare model were fair (0.73 and 0.79 respectively). CONCLUSIONS: In this study, an "in-house" developed and validated risk model has the most accurate discriminative value in predicting perioperative mortality after elective AAA repair. For purposes of comparative audit with case mix adjustments, national models such as the VGNW or Medicare models should be used.
OBJECTIVES: To develop and validate an "in house" risk model for predicting perioperative mortality following elective AAA repair and to compare this with other models. DESIGN: Multivariate logistics regression analysis was used to identify risk factors for perioperative-day mortality from one tertiary institution's prospectively maintained database. MATERIALS AND METHODS: Consecutive elective open (564) and endovascular (589) AAA repairs (2000-2010) were split randomly into development (810) and validation (343) data sets. The resultant model was compared to Glasgow Aneurysm Score (GAS), Modified Customised Probability Index (m-CPI), CPI, the Vascular Governance North West (VGNW) model and the Medicare model. RESULTS: Variables associated with perioperative mortality included: increasing age (P = 0.034), myocardial infarct within last 10 years (P = 0.0008), raised serum creatinine (P = 0.005) and open surgery (P = 0.0001). The areas under the receiver operating characteristic curve (AUC) for predicted probability of 30-day mortality in development and validation data sets were 0.79 and 0.82 respectively. AUCs for GAS, m-CPI and CPI were poor (0.63, 0.58 and 0.58 respectively), whilst VGNW and Medicare model were fair (0.73 and 0.79 respectively). CONCLUSIONS: In this study, an "in-house" developed and validated risk model has the most accurate discriminative value in predicting perioperative mortality after elective AAA repair. For purposes of comparative audit with case mix adjustments, national models such as the VGNW or Medicare models should be used.
Authors: Bjoern D Suckow; Philip P Goodney; Jesse A Columbo; Ravinder Kang; David H Stone; Art Sedrakyan; Jack L Cronenwett; Mark F Fillinger Journal: J Vasc Surg Date: 2017-12-28 Impact factor: 4.268