Comparison of 4 mortality scores for surgical repair of type A aortic dissection: a multicentre external validation

Abstract OBJECTIVES In the last decades, 4 different scores for the prediction of mortality following surgery for type A acute aortic dissection (TAAD) were proposed. We aimed to validate these scores in a large external multicentre cohort. METHODS We retrospectively analysed patients who underwent surgery for TAAD between 2000 and 2020. Patients were enrolled from 10 centres from 2 European countries. Outcomes were the early (30-day and/or in-hospital) and 1-year mortality. Discrimination, calibration and observed/expected (O/E) ratio were evaluated. RESULTS A total of 1895 patients (31.7% females, mean age 63.72 ± 12.8 years) were included in the study. Thirty-day mortality and in-hospital mortality were 21.7% (n = 412) and 22.5% (n = 427) respectively. The German Registry of Acute Aortic Dissection Type A (GERAADA) score shows to have the best discrimination [area under the curve (AUC) 0.671 and 0.672] in predicting as well the early and the 1-year mortality, followed by the International Registry of Acute Aortic Dissection (IRAD) model 1 (AUC 0.658 and 0.672), the Centofanti (AUC 0.645 and 0.66) and the UK aortic score (AUC 0.549 and 0.563). According to Hosmer–Lemeshow and Brier tests, the IRAD model I and GERAADA, respectively, were well calibrated for the early mortality, while the GERAADA and Centofanti for the 1-year mortality. The O/E analysis showed a marked underestimation for patients labelled as low-risk for UK aortic score and IRAD model I for both outcomes. CONCLUSIONS The GERAADA score showed the best performance in comparison with other scores. However, none of them achieved together a fair discrimination and a good calibration for predicting either the early or the 1-year mortality.