从医疗记录中提取结构化数据的双阅读/录入系统及其应用  

A New Model for Processing Data from Medical Records

作　　者：罗立刚 胡佳佳 王晓哲 张天泽 李丽平 

机构地区：[1]零氪科技(北京)有限公司,北京100080

出　　处：《药物流行病学杂志》2017年第6期406-409,共4页Chinese Journal of Pharmacoepidemiology

基　　金：吴阶平医学基金会医疗大数据开发与应用专项基金项目(编号:320.6750.16062)

摘　　要：在医疗实践中产生了大量的保存于纸质归档系统或电子医疗记录系统中的临床和药物数据,并形成了生物医学大数据(BBD)。如何充分利用BBD中的非结构化数据是医疗保健领域大数据研究的一大挑战。本文介绍了一种可从非结构化医疗记录中抽取并创建半结构化数据库的双阅读/录入系统(double-reading/entry system,DRESS),该系统基于云技术,主要包括Link MR、Link Core和Link QC三个子系统,完成医疗数据的收集和上传、安全保存和管理、半结构化数据的提取以及反复的质控。通过随机选取100份肺癌患者的医疗记录对DRESS进行再现性研究,结果显示对于大多数临床变量,DRESS都具有相当高的再现性。DRESS作为一种把非结构化BBD转成半结构化数据的混合系统是目前一种解决医疗大数据挑战的可行方案。Healthcare provided a huge amount of biomedical data stored in either legacy system （paper-based） format or electronic medical records （EMR） around the world, which were collectively referred to as big biomedical data （BBD）. It was a big challenge for health care that how to take advantage of unstructured data in BBD. This paper intro- duced a double-reading/entry system （DRESS） to extracting structured medical information from unstructured data in medi- cal records. Utilizing the modern cloud-based technologies, we had developed DRESS that included three mainly subsys- tems （LinkMR, LinkCore and LinkQC）, from capturing MRs in clinics, to securely transferring MRs, storing and manag- ing cloud-based MRs, to facilitating both machine learning and manual reading, and to performing iterative quality control before committing the semi-structured data into the desired database. To evaluate the reproducibility of extracted medical data elements by DRESS, we conducted a blinded reproducibility study, with 100 MRs from patients who had undergone surgical treatment of lung cancer in China. The results showed an overall high reproducibility. DRESS represented probably a workable solution to solve the big medical data challenge.

关 键 词：非结构化医疗记录 半结构化数据 双阅读/录入系统 质量控制 再现性 

分 类 号：R181.3[医药卫生—流行病学]