Time-zero multimodal risk stratification for in-hospital mortality in acute aortic dissection: two-centre development, temporal and external validation, and occult shock phenotyping.
Authors
Affiliations (3)
Affiliations (3)
- Department of Ultrasonography, The First Affiliated Hospital, Air Force Medical University, Xi'an, China.
- Department of Health Statistics, Air Force Medical University, Xi'an, China.
- Department of Cardiac Surgery, The First Affiliated Hospital, Air Force Medical University, Xi'an, China.
Abstract
Acute aortic dissection (AAD) requires rapid anatomical diagnosis and early risk stratification, yet initial blood pressure alone may miss patients with concealed hypoperfusion. Although computed tomography angiography (CTA) defines dissection anatomy and transthoracic echocardiography (TTE) provides bedside haemodynamic information, the incremental prognostic value of early TTE and the significance of normotensive hyperlactatemia remain uncertain. We aimed to develop and validate an initial 2-hour multimodal model integrating clinical, CTA, and TTE data to predict in-hospital mortality and to describe occult shock in normotensive AAD patients. This retrospective two-centre cohort study included 2,584 adults with first-time AAD who completed CTA and TTE during the initial 2-hour assessment after emergency department triage. Cohorts were development (n=1,678; 2007-2017), internal temporal validation (n=452; 2018-2021), and external geographic/later-era validation (n=454; 2022-March 2024). Candidate variables were restricted to information available during this initial assessment window. Four supervised algorithms were compared, with ablation analyses assessing modality-specific contribution. XGBoost achieved optimism-corrected development area under the curve (AUC) of 0.86 [95% confidence interval (CI): 0.82-0.88], internal validation 0.83 (0.79-0.87), and external validation 0.78 (0.74-0.82). Adding TTE to Clinical + CTA improved discrimination (Delta AUC +0.04, P=0.002) and integrated discrimination improvement (IDI) (+0.021, P=0.01). Normotensive occult shock occurred in 548/2,082 patients (26.3%) and was associated with mortality 15.9% versus 6.7% in metabolically stable normotension. In high-volume aortic centres where CTA and early TTE are routinely available, this initial 2-hour multimodal model showed temporal validation and moderate external discrimination for in-hospital mortality. TTE added prognostic information, particularly for type A disease, and normotensive hyperlactatemia identified a higher-risk subgroup; however, general use, type B performance, local recalibration, and action thresholds require prospective multicentre validation before clinical implementation.