基于既有医疗数据构建研究型数据库的方法学探讨及实例解读(二):数据治理的方法

Methodological Discussion and Case Interpretation of Construction Research Database Based on Existing Medical Data(Ⅱ):Methods of Data Governance

真实世界研究可利用临床实际产生的数据,通过科学合理的研究设计及严谨的数据治理工作,形成真实世界证据,与随机对照试验形成互补的关系。数据治理包括数据链接、数据提取、数据核查、数据清理4个步骤。其中,数据链接的核心在于患者识别码的治理,应确定唯一、无重复的患者ID序列,并进行“纵向锁定”;数据提取的核心是同质化操作,可通过背靠背、培训、预提取、格式限定等方法,减少不同操作人员间的差异;数据核查的难点在于选择恰当的核查范围,可采用抽样核查和全面核查相结合的方式,在保证有效核查的前提下减少核查工作量;数据清理的重点在于对核查出的重复数据、矛盾数据、极端值和缺失值等各种问题数据制定恰当的清理规则。本文以“中西医结合治疗新型冠状病毒感染研究型数据库”为例,对研究中数据治理的方法进行介绍,以供研究人员参考。

Real world research can use the data generated in clinical practice to form real world evidence through scientific and reasonable research design and rigorous data governance,and form a complementary relationship with randomized controlled trials.Data governance includes four steps:data linking,data extraction,data verification,and data cleaning.The core of the data link is the management of the patient ID,which should determine a unique,non-repetitive patient ID sequence and carry out“portrait lock”.The core of data extraction is homogenization operation,which can reduce the differences between different operators through methods such as back-to-back,training,pre-extraction,and format limitation.The difficulty of data verification is choosing the appropriate verification scope.Combination of sampling verification and comprehensive verification can be used to reduce the workload on ensuring effective verification.The focus of data cleaning is to formulate appropriate cleaning rules for problematic data such as duplicate data,contradictory data,extreme values and missing values.This article took the“Integrated Traditional Chinese and Western Medicine Treatment of COVID-19 Research Database”as an example to introduce the method of data governance in the study for the reference of researchers.