SP 11.04 Prognostic Models for Mortality in Acute Mesenteric Ischemia: A Systematic Review and Critical Appraisal of Published Prediction Models
Daniela Jardan, Sergiu Timofeiov, Gabriel Andi Iordache, Marius Emilian Hojbota, Radu Florin Popa, Andrei MoscaluAbstract
Introduction
Acute mesenteric ischaemia (AMI) mortality remains high with short- and long-term mortality of 59.6% and 68.2%, respectively. No validated tool exists for mortality prediction. Several models have been proposed; their methodological quality and clinical applicability remain uncertain. This review aims to evaluate published mortality prediction models in AMI.
Methods
A systematic search of MEDLINE (Ovid), EMBASE, PubMed, Scopus, Web of Science, the Cochrane Library, ProQuest, EBSCO, and grey literature was performed from inception to 22ndAugust 2025. Studies developing multivariable mortality prediction models in adult AMI patients in acute-care settings, with defined predictor timing and outcome assessment, were included. Data extraction followed CHARMS checklist. Risk of bias and applicability were assessed using PROBAST+AI. Models were analysed independently.
Results
From 30,125 records, 13 studies (9-retrospective, 4-prospective) reporting 15 prediction models were included (8,146 patients). Mean age ranged from 67 to 74 years. Eleven studies included mixed AMI populations. Where reported, arterial occlusive disease accounted for 12.8–59.5%, non-occlusive mesenteric ischaemia 3.6–66.4%, and mesenteric venous thrombosis 0.8–9.2%. Mortality ranged from 17% to 58%. Logistic regression was used in 9 models and machine-learning methods in 2. Discrimination was reported for 11 models (AUC 0.67–0.93). Calibration was assessed in 8 models (mostly using Hosmer–Lemeshow test). No model underwent external validation. All models demonstrated high risk of bias, driven by analytical limitations and small sample sizes, applicability concerns were generally low.
Conclusion
Current mortality prediction models for AMI are methodologically limited and lack robust validation, restricting clinical usefulness.