Accuracy evaluations of three ruptured abdominal aortic aneurysm mortality risk scores using an independent dataset

To date, no single scoring system for predicting 30-day mortality in patients with ruptured abdominal aortic aneurysms (rAAAs) has been endorsed by any vascular society or proven to definitively predict treatment futility. Three recently developed scoring systems for predicting 30-day mortality in patients with rAAA have been validated by their respective institutional data. The purpose of this study was to evaluate the accuracy of these rAAA mortality risk scores using an independent community hospital dataset.

Consecutive patients presenting with rAAA at Saint Joseph Hospital between January 1, 2009, and February 28, 2017, were used for validation. Logistic regression analysis was used to evaluate the association between risk score and odds of death. Confidence intervals were calculated using the Wilson method. Comparisons were made between models by calculating the area under the receiver operating characteristic (AUC) curves.

Complete data from 38 patients was used for accuracy evaluation. The AUCs for the Dutch Aneurysm Score, Harborview Medical Center score, and Vascular Surgery Group of New England (VSGNE) score were 0.762, 0.792, and 0.860, respectively, for all patients. When evaluating 30-day mortality for patients undergoing ruptured endovascular aneurysm repair, the scores were 0.802, 0.893, and 0.927, respectively. The difference between scores did not reach statistical significance. All three indexes significantly associated with the mortality rate using logistic regression.

Each risk score accurately predicted 30-day mortality using the independent dataset. The results suggest that the VSGNE score is the most accurate; however, differences in accuracy between each scoring system did not reach statistical significance. The Harborview Medical Center scoring system is based only on preoperative variables. Although the VSGNE score had the highest AUC in this analysis, it is dependent on intraoperative variables. The authors favor a single risk assessment tool, with consensus vascular societal approval, that incorporates preoperative variables and includes a tool for the prediction of treatment futility.