基于真实世界研究的淋巴瘤研究数据库建设与应用  

Application and design of lymphoma dataset based on real-world research

作　　者：米岚[1] 吴梦[1] 冯非儿 杜婷婷[1] 热依扎·努尔苏力坦 林梦梦 牛明芳 宋玉琴[1] 谢彦[1] 朱军[1] Mi Lan;Wu Meng;Feng Feier;Du Tingting;Reyizha Luersulitan;Lin Mengmeng;Niu Mingfang;Song Yuqin;Xie Yan;Zhu Jun(Key Laboratory of Carcinogenesis and Translational Research(Ministry of Education),Department of Lymphoma,Peking University Cancer Hospital&Institute,Beijing,China;Beijing Goodwill Hessian Health Technology Co.,Ltd.)

机构地区：[1]北京大学肿瘤医院暨北京市肿瘤防治研究所,淋巴肿瘤内科,恶性肿瘤发病机制及转化研究教育部重点实验室,北京100142 [2]北京嘉和海森健康科技有限公司,100007

出　　处：《中华医学科研管理杂志》2023年第1期18-23,共6页Chinese Journal of Medical Science Research Management

基　　金：中国初级卫生保健基金会-"超越计划":淋巴瘤临床数据库公益项目。

摘　　要：目的针对临床数据数量庞大和质量差的现状,本研究旨在以建立淋巴瘤研究数据库为例,探索高质量研究数据库的建立路径以及在真实世界研究的作用。方法汇总研究领域专家意见,参考相关指南和标准,建立标准医学知识库;回顾性抽取2005年2月-2021年12月期间就诊于北京大学肿瘤医院淋巴瘤患者电子诊疗数据,采用深度学习、自然语言处理等方式,搭建"基于电子病历系统的淋巴瘤数据库-生物样本信息库-延伸遗传信息库"的动态智能信息整合与处理系统。结果研究数据库在满足了临床科研人员的研究需求的同时,实现医院病历数据和生物样本信息数据的申请、审批、溯源和分析全过程留痕管理。数据库中核心科研变量总数为668个,结构化变量占46.0%。截至2021年12月25日,数据库中共有淋巴瘤患者68687人,男女患者人数比值为8/9,就诊次≥3次的患者占比为23.0%。此外,研究者可在数据库中根据目标条件叠加检索,显示命中的就诊记录,建立研究队列,进行统计建模,挖掘数据信息。结论通过整合管理流程和利用自然语言人工智能新技术建立循证等级高的数据库,有助于医院信息系统的互联互通与资源共享,从而达到为开展真实世界研究提供可靠详实数据的目的。Objective Considering the large amount and poor quality of clinical data,this study aims to explore the establishment of high-quality research database and its role in real-world research by taking the establishment of lymphoma research database as an example.Methods The expert opinions in the field of lymphoma were collected,and the relevant guidelines and standards were referenced to establish a standard medical knowledge dataset.The electronic diagnosis and treatment data of lymphoma patients treated in Peking University Cancer Hospital from February 2005 to December 2020 were retrospectively extracted,the deep Learning,natural language processing were adopted to build a dynamic intelligent information integration and processing system of"lymphoma database based on electronic medical record system-biological sample information database-extended genetic information database".Results The research database not only meets the research needs of clinical researchers,but also realizes the management of traces in the whole process of application,approval,traceability and analysis of hospital medical record data and biological sample data.The total number of research variables in the database was 668,and the structured variables accounted for 46.0%.On December 25,2021,there were 68687 lymphoma patients in the database,the ratio of male to female patients was 8/9,and the proportion of patients with≥3 visits accounted for 23.0%.In addition,researchers can superimpose searches in the database according to the target conditions,display the targeted medical records according to research hypothesis,and then establish a research cohort,conducting statistical modeling,and mining data information.Conclusions By integrating management processes and using new natural language artificial intelligence technology to establish a high-level evidence-based database,it is helpful for the interconnection and resource sharing of hospital information systems,so as to achieve the purpose of providing reliable and detailed data for real-wo

关 键 词：真实世界研究 电子病历 淋巴瘤 数据库 自然语言处理 人工智能 

分 类 号：R733.1[医药卫生—肿瘤]