基于真实世界诊疗数据的失眠症状群分类方法研究  

Classification of Insomnia Symptom Cluster based on Real-world Medical Data

作　　者：李凌香 李亚茹 秦宇宁 邸露瑶 刘艳骄[5] 张润顺[5] 黄俊山[6] 齐向华[7] 王松龄[8] 郭蓉娟[9] 王平 周雪忠[2] 李洪皎[3] LI Lingxiang;LI Yaru;QIN Yuning;DI Luyao;LIU Yanjiao;ZHANG Runshun;HUAN Junshan;QI Xianghua;WANG Songling;GUO Rongjuan;WANG Ping;ZHOU Xuezhong;LI Hongjiao(Affiliated Hospital of Inner Mongolia Minzu University,Tongliao 028007,China;School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China;Institute of Basic Research in Clinical Medicine,China Academy of Chinese Medical Sciences,Beijing 100700,China;Shanxi University of Chinese Medicine,Xian 712046,China;Guang′anmen Hospital,China Academy of Chinese Medical Sciences,Beijing 100700,China;Fujian Academy of Chinese Medical Sciences,Fuzhou 350003,China;Affiliated Hospital of Shandong University of Traditional Chinese Medicine,Ji′nan 250011,China;Henan Province Hospital of Traditional Chinese Medicine,Zhengzhou 450053,China;Dongfang Hospital Beijing University of Chinese Medicine,Beijing 100078,China;Institute of Geriatrics,Hubei University of Chinese Medicine,Wuhan 430065,China)

机构地区：[1]内蒙古民族大学附属医院,通辽028007 [2]北京交通大学计算机与信息技术学院,北京100044 [3]中国中医科学院中医临床基础医学研究所,北京100700 [4]陕西中医药大学,西安712046 [5]中国中医科学院广安门医院,北京100053 [6]福建省中医药研究院,福州350003 [7]山东中医药大学附属医院,济南250011 [8]河南省中医院,郑州450053 [9]北京中医药大学东方医院,北京100078 [10]湖北中医药大学老年医学研究所,武汉430060

出　　处：《世界中医药》2022年第5期635-644,共10页World Chinese Medicine

基　　金：国家自然科学基金面上项目(81673964);首都卫生发展科研专项项目(2022-1-4301)。

摘　　要：目的:本研究借鉴《伤寒论》“方证相应”思想,引入国际通用的症状群概念,采用相对固定和高度概括的症状群,将证候进一步客观化,并分析比较有效人群与无效人群的失眠症状群分类及特点,可为分析挖掘失眠症状群提供方法学参考。方法:基于真实世界失眠临床诊疗数据,严格按照疾病公认的疗效评价标准,采用倾向性评分匹配方法消除混杂因素,筛选出有效和无效病例;采用文本挖掘方法对有效与无效病例的辨证论治信息进行处理,提取高质量规范化症状谱,再将高频症状导入孔明灯软件进行隐结构分析。结果:隐结构分析法可较好地实现失眠症状群的分类,为症状群的客观性和稳定性提供了相对成熟的模型和算法,不同失眠症人群具有不同的症状群分类和特点,为证候的客观化提供了依据。结论:症状群是证候的主要信息载体,用相对固定且具有高度概括性的症状群,补充替代“证候”进行辨证论治,或许可克服辨证论治由于医师个体经验和学术流派不同而导致的稳定性和一致性较差的问题,为相关研究提供方法学参考。Objective:According to the idea of“correspondence of prescription and syndrome”in Treatise on Cold Damage,this study introduced the international concept of symptom cluster,which was relatively fixed and highly generalized,to objectively reflect syndrome.In addition,the types and characteristics of insomnia symptom clusters of effective and ineffective population were analyzed.This study is expected to provide methodologies for exploring symptom clusters of insomnia.Methods:Based on the real-world clinical data of insomnia and the widely accepted efficacy evaluation standard,the confounding factors were eliminated by propensity score matching and the effective and ineffective cases were screened.The text mining method was used to process information on the symptoms from effective and ineffective insomniacs and extract high-quality standardized symptom set.Then,the Lantern app was employed for latent tree analysis of the high-frequency symptoms.Results:Latent tree analysis classified insomnia symptoms into clusters and provided a mature model and algorithm for the objectivity and stability of symptoms clusters.In addition,different insomnia populations had different types and characteristics of symptom clusters,which laid a basis for the objectification of syndrome.Conclusion:Symptom clusters contain the main information of syndrome and the use of relatively fixed and highly generalized symptom clusters to supplement and replace“syndrome”for syndrome differentiation and treatment might be able to eliminate the instability and inconsistency among different doctors due to the difference in individual experience and academic school.This study is expected to provide methodology for related research.

关 键 词：症状群 真实世界研究 失眠症 方法学 

分 类 号：R21[医药卫生—中医学]