Ingestion of amatoxin-containing mushrooms can result in acute hepatotoxicity and carries a high risk of progression to liver failure with considerable mortality. Timely identification of patients at risk is essential for improving outcomes. This study investigated admission biomarkers predictive of liver failure in amatoxin poisoning and identified key prognostic indicators through machine learning analysis. A retrospective cohort of 71 patients was stratified into liver injury (n = 36) and liver failure (n = 35) groups. Comparative analysis revealed significant intergroup differences across multiple admission biomarkers. The machine learning model incorporating aspartate aminotransferase (AST), lactate dehydrogenase (LDH), fibrinogen (FIB), activated partial thromboplastin time (APTT), thrombin time (TT), prothrombin time (PT), D-dimer, and fibrin degradation products (FDP) demonstrated high predictive performance. The model achieved an accuracy of 0.98, specificity of 0.98, and sensitivity of 0.94, with an area under the curve (AUC) of 0.97 in the training set; test performance remained robust, with 0.90 accuracy, 0.89 AUC, 0.93 specificity, and 0.85 sensitivity. Quartile analysis further identified admission D-dimer >2.5 μg/mL and prothrombin time >19.2 s as critical risk thresholds for liver failure progression. These findings highlight a panel of early hemostatic and hepatic injury markers-particularly elevated D-dimer and prolonged PT-as effective predictors of liver failure in amatoxin-induced hepatotoxicity, offering valuable guidance for early clinical intervention.Copyright © 2025 The Authors. Published by Elsevier Ltd.. All rights reserved.