Hong Yu, Li Li-Hua, Kuo Ting-Hao, Lee Yi-Tzu, Hsu Cheng-Chih
Department of Chemistry, National Taiwan University, 10617, Taipei, Taiwan.
Department of Pathology and Laboratory Medicine, Taipei Veterans General Hospital, 11217, Taipei, Taiwan.
J Am Soc Mass Spectrom. 2025 Jun 4;36(6):1264-1276. doi: 10.1021/jasms.5c00009. Epub 2025 May 9.
Early recognition of septic shock is crucial for improving clinical management and patient outcomes, especially in the emergency department (ED). This study conducted serum metabolomic profiling on ED patients diagnosed with septic shock (n = 32) and those without septic shock (n = 92) using a high-resolution mass spectrometer. By implementing a supervised machine learning algorithm, a prediction model based on a panel of metabolites achieved an accuracy of 87.8%. Notably, when employed on a low-resolution instrument, the model maintained its predictive performance with an accuracy of 84.2%. These results demonstrate the potential of metabolite-based algorithms to identify patients at high risk of septic shock. Our proposed workflow aims to optimize risk assessment and streamline clinical management processes in the ED, holding promise as an efficient routine test to promote timely intensive interventions and reduce septic shock mortality.
早期识别感染性休克对于改善临床管理和患者预后至关重要，尤其是在急诊科（ED）。本研究使用高分辨率质谱仪对诊断为感染性休克的急诊科患者（n = 32）和未患感染性休克的患者（n = 92）进行了血清代谢组学分析。通过实施监督式机器学习算法，基于一组代谢物的预测模型准确率达到了87.8%。值得注意的是，当在低分辨率仪器上使用时，该模型仍保持其预测性能，准确率为84.2%。这些结果证明了基于代谢物的算法在识别感染性休克高危患者方面的潜力。我们提出的工作流程旨在优化急诊科的风险评估并简化临床管理流程，有望成为一种有效的常规检测方法，以促进及时的强化干预并降低感染性休克死亡率。
