从真实世界数据到临床研究数据的标准转化研究.docxVIP

PAGE2 / NUMPAGES2 从真实世界数据到临床研究数据的标准转化研究 摘? 要 / Abstract ? 临床研究中电子病例报告表(eCRF)的数据收集，传统上由临床研究协调员(CRC)阅读电子病历(EMR)数据将相关内容手动录入至电子数据采集(EDC)系统。为了减轻CRC的负担，目前已有研究在探索将EMR源数据经过数据标准化转换直接变为研究数据集的方法。EMR中大量的非结构化文本数据导致了数据提取困难，无法直接用于临床研究。本文首先探讨了国内对于真实世界数据应用于临床研究数据标准化的需求及困难，开发了一种数据标准化方法。本方法可以基于EMR源数据，通过数据标准化的方式自动填充临床数据交换标准协会(CDISC)标准的eCRF，并满足监管部门的数据递交要求。本方法采用了我国常见的数据标准、人工智能领域的自然语言处理技术，以及提升数据质量的创新型数据采集模式。其数据转化过程的核心是根据最简化的数据模型制定文本数据标签指南，提高了使用自然语言处理算法的效率，优化了其与临床数据模型的互操作性，以及辅助提取研究中所需要的标准术语库。 ? For the data collection of electronic case report form (eCRF) in clinical research,the clinical research coordinator (CRC) traditionally reads the electronic medical record (EMR) and manually enters its relevant contents into the electronic data collection system (EDC).In order to reduce the burden of CRC,methods has been explored to directly transform EMR source data into a research dataset through data standardization and transformation.The large amount of unstructured text data in EMR leads to difficulty in data extraction,which prevents data from being directly used in clinical research.This study discusses the domestic needs and difficulties of real-world data standardization,and develops a data standardization framework to solve the difficulties.The data standardization framework developed can be used to automatically fill the eCRFs based on the CDISC standard using EMR source data while satisfying regulatory requirements for data submission authorities.The framework considers China's common data standards,natural language processing technology in the field of artificial intelligence,and innovative data acquisition mode to improve data quality.The core aspects of the data transformation process in the standardization framework include the formulation of text data label guidelines according to the simple data models,improvement of the efficiency of natural language processing algorithms,optimization of interoperability with clinical data models and capture of standard terminologies used in cli