真实世界医院集中监测中药不良反应的类不平衡高维预测及其风险因素分类识别  

Class-imbalance Prediction and High-dimensional Risk Factor Identification of Adverse Reactions of Traditional Chinese Medicine with Centralized Monitoring in Real-world Hospitals

作　　者：谢飞彪 彭叶辉[1] 杨伟[2] 唐进法[3] 刘娟[4] 李伟霞[3] 张辉[3] 吴东苑 吴娅丽 冷源铭 向兴华 XIE Feibiao;PENG Yehui;YANG Wei;TANG Jinfa;LIU Juan;LI Weixia;ZHANG Hui;WU Dongyuan;WU Yali;LENG Yuanming;XIANG Xinghua(School of Mathematics and Computational Science,Hunan University of Science and Technology,Xiangtan 411201,China;Institute of Basic Research in Clinical Medicine,China Academy of Chinese Medical Sciences,Beijing 100700,China;Engineering Laboratory of Henan Province for Clinical Evaluation Technology of Chinese Medicine,The First Affiliated Hospital of Henan University of Chinese Medicine,Zhengzhou 450000,China;China Academy of Inspection and Quarantine,Beijing 100123,China;College of Public Health and Health Professions,University of Florida,Gainesville 32601,US;College of Art and Science,Boston University,Boston 02215,US)

机构地区：[1]湖南科技大学数学与计算科学学院,湖南湘潭411201 [2]中国中医科学院中医临床基础医学研究所,北京100700 [3]河南中医药大学第一附属医院,河南省中医药临床评价技术工程实验室,郑州450000 [4]中国检验检疫科学研究院,北京100123 [5]佛罗里达大学公共卫生学院,盖恩斯维尔32601 [6]波士顿大学艺术与科学学院,波士顿02215

出　　处：《中国实验方剂学杂志》2023年第14期114-122,共9页Chinese Journal of Experimental Traditional Medical Formulae

基　　金：中国中医科学院科技创新工程项目(CI2021A04706,CI2021B003);中国中医科学院自主选题项目(Z0643,Z0723);国家重点研发计划项目(2017YFC1700406-2,2018YFC1704306)。

摘　　要：目的:基于真实世界医院集中监测的中药上市后安全性数据,实现中药不良反应的类不平衡高维预测并分类识别影响药品不良反应(ADR)发生的风险因素。方法:采用集成聚类重采样结合正则化Group Lasso回归,对ADR类不平衡的数据进行高维平衡处理,进而集成平衡数据实现ADR预测及其风险因素分类识别。结果:对中药安全性监测数据的示例研究结果显示,建立的ADR预测模型在测试集上的预测正确率、ADR发生的预测灵敏度、ADR未发生的预测特异度及受试者工作特征曲线下面积(AUC)4个指标均达到0.8以上。同时,该方法从600个高维变量中筛选出40个影响ADR发生的保护因素或危险因素,并分类加权识别各类因素对ADR发生的影响,重要风险因素类别依次为既往史、给药方案、合并用药名称、病症情况、合并用药数量、个人资料。结论:在ADR罕见且存在大量临床变量的真实世界数据中,本文实现了精准ADR类不平衡高维预测,并分类识别出关键风险因素及其所属类别的临床重要性,为临床合理用药及联合用药提供风险预警,也为中药上市后安全性再评价提供科学依据。Objective:Accurately predicting the imbalanced Adverse Drug Reactions(ADRs)of traditional Chinese medicine(TCM)and identifying by category the key risk factors affecting the occurrence of ADRs on the Post-marketing safety data of TCM monitored centrally in real world hospitals.Method:Applying the ensemble clustering resampling combined with regularization method Group Lasso to process the class-imbalanced and high dimensional ADRs data on the safety data and then to integrate the balanced datasets to achieve ADRs prediction and risk factors identification by category.Result:A practical example study of the proposed method on a monitoring data of TCM injection performed that the accuracy of the ADRs prediction,the prediction sensitivity,the prediction specificity and the area under receiver operating characteristic curve(AUC)are all above 0.8 on the test set.Meanwhile,40 risk factors affecting the occurrence of ADRs are screened out from total 600 high-dimensional variables.And the effects of risk factors are weighted to identified according to the personal data,past history,disease status,medication information,number of combined drugs and name of combined drugs.Conclusion:Applying the proposed method of class-imbalanced and high-dimensional prediction and risk factors identification by category on the real world data with the problem of extremely rare ADRs and high-dimensional clinical variables achieved high accurate prediction of ADRs and identification of key risk factors with clinical importance of their affiliated category.The application of the method could provide advance ADRs warning for clinical rational medication and drug combination,provide scientific basis for safety reevaluation of post-marketing TCM.

关 键 词：医院集中监测 药品不良反应(ADR) 聚类重采样 丹红注射液 组结构正则化 类不平衡高维预测 风险因素分类识别 

分 类 号：R22[医药卫生—中医基础理论] R28[医药卫生—中医学] R24[社会学—统计学] C8