基于电子病历的表型分析方法及应用综述

目的：对基于电子病历的表型分析流程、主要方法、应用现状进行总结梳理，为后续相关研究提供借鉴。方法：通过关键词检索、参考文献回溯等方法在PubMed数据库中检索并经人工筛选后获得53篇文献，在此基础上总结了表型分析的一般流程，对表型定义及表型算法开发、验证和共享进行了综述，最后归纳了表型分析在临床研究中的应用现状。结果：表型分析的数据来源丰富，包括开放医学资源、结构化和非结构化电子病历数据等，表型算法开发的方法逐渐从基于规则开发向无监督学习开发转变，但目前电子病历数据的利用仍有很多挑战和困难有待解决。结论：今后应继续致力于提高数据的可信度，增加数据来源和种类以提升表型算法的效果，同时还应加强表型分析相关研究的共享和传播、扩大表型分析在临床研究中的应用。

Phenotype methods and applications based on electronic medical records

Objective To provide reference for further research work by analyzing the workflow, methods and applications of phenotyping based on electronic medical records. Methods 53 papers on phenotyping based on electronic medical records were retrieved from the PubMed database by key words searching, reference backtracking and manual screening, with which the general workflow of phenotyping was summarized, the phenotype definition, phenotype algorithm development, validation and sharing were reviewed, and finally the application status of phenotyping in clinical research was summarized. Results The data sources for phenotyping are rich, including open medical resources, structured and unstructured electronic medical records, etc. The development method of phenotype algorithm development was gradually changing from rule-based to unsupervised learning. However, there are still many challenges and difficulties to be solved in the application of electronic medical records. Conclusion Further efforts should be made to improve the reliability of data, increase the variety and types of data sources and improve the effect of phenotype algorithm, and further research can be carried out on strengthening the sharing and dissemination of research related to phenotyping and expanding the application of phenotyping in clinical research.