败血症
药品
雅卡索引
医学
重症监护医学
不利影响
计算机科学
药物与药物的相互作用
精密医学
器官功能障碍
药物反应
临床试验
药理学
梅德林
生物信息学
药物相互作用
机器学习
计算生物学
基线(sea)
药品管理局
主机响应
作者
Zhuo Zhao,Yongqiang Yu,Yi Zhong,Gaozheng Li,Baihua Wang,Zhaoting Ye,Qi Sun,Hong Sun,Zuquan Weng
标识
DOI:10.1142/s0219720026500113
摘要
Sepsis is a type of life-threatening organ dysfunction caused by dysregulated host response to infection. Given the heterogenous nature of sepsis, precise and temporally-optimized drug recommen-dations are critical, yet existing AI-based drug recommendation models often fail to address temporal dynamics in medication administration and in clinical practice. To address this limitation, we propose ICUDrug, a novel time-window-based drug recommendation model that extracts patient-specific clini-cal features at distinct critical time points through focused temporal windows while explicitly modeling drug-drug interaction (DDI) to minimize adverse reactions. ICUDrug achieved performance with Jaccard (0.7068 ± 0.0035), F1-score (0.8166 ± 0.0028), and AUPRC (0.8879 ± 0.0022) at a 24-hour time window, while reducing DDI rate to 0.1914 ± 0.0010, demonstrating significant effectiveness and outperformed baseline models. These results demonstrate the ability of ICUDrug to deliver both clini-cally-relevant timing and safe medication combinations for sepsis management.
科研通智能强力驱动
Strongly Powered by AbleSci AI