Abstract
OBJECTIVES: Acute type A aortic dissection (ATAAD) carries high mortality, with emergency surgery being the cornerstone of treatment. The German Registry for Acute Type A Aortic Dissection (GERAADA)-score is advocated in guidelines to predict 30-day mortality after ATAAD surgery. This study investigates its performance in a Dutch cohort, with an emphasis on malperfusion definitions, age groups, and sex. METHODS: Adults undergoing emergency surgery for ATAAD at 5 Dutch centres (2007-2024) were included in a multicentre database. External validation of the GERAADA-score was performed with these data, using discrimination (area under the curve [AUC]) and calibration (Brier score, Hosmer-Lemeshow test, and calibration plots). A logistic regression with GERAADA variables was fitted on the study population, and assumptions were checked. Subgroup analyses were conducted based on sex, age groups, and malperfusion definitions (including imaging and clinical definitions). RESULTS: A total of 1,146 patients underwent emergency surgery for ATAAD. Observed early mortality was 16.9% (n = 194). Of 1,130 patients included in the external validation cohort, 92.2% had low-intermediate risk (GERAADA-score ≤ 30%). The GERAADA-score showed moderate discrimination (AUC = 0.649, 95% confidence interval = 0.604-0.694), with a higher AUC for younger patients (50-59 years). The malperfusion definition including ischaemia confirmed by imaging showed the best discriminative power. Calibration was good (Hosmer-Lemeshow, P = .754, Brier score = 0.131). Logistic regression identified age, catecholamine use, ventilation support, and coronary and peripheral malperfusion as independent risk factors for 30-day mortality, with signs of multicollinearity between preoperative catecholamine use and resuscitation. CONCLUSIONS: In the Dutch setting, the GERAADA-score demonstrated moderate discriminative power and good calibration across relevant subgroups. Adaptations of the GERAADA-score, including conducting a haemodynamic instability variable, may be considered to avoid redundant predictions and boost reproducibility.