基于大模型的电子病历与CDA标准模式映射方法研究

A Study on Mapping Method of Electronic Medical Records to CDA Standard Based on Large Language Models

目的构建一套通用的、自动化的数据标准映射方法,解决不同医疗机构之间临床数据交流时数据转换的问题。方法采用基于大模型的多层级映射方法,来实现不同电子病历与CDA标准文档之间的表名和字段名的映射,并在原始标准病历和CDA标准上进行实验验证。结果实验结果表明,该方法可以自动映射74.4%的数据元到CDA标准数据元,准确率达到69.6%。结论基于大模型的多层级映射方法可用于不同的临床电子病历转换场景,与传统人工以及基于知识图谱的方法相比,该方法更加通用灵活,效率更高。

Objective In order to address the problem of data conversion during clinical data exchange between different clinical institutions,it is necessary to establish a universal and automated data standard mapping method.Methods We propose a multi-level mapping method based on large language models to achieve the mapping of table names and data elements between different electronic medical records and the CDA standard,and experiments are conducted on the original standard electronic medical records and the CDA standard.Results Our experiments demonsrate that the method automatically mapped 74.4%of the data elements to CDA standard data elements,achieving an accuracy rate of 69.6%.Conclusion A multi-level mapping method can be used in various clinical electronic medical record conversion scenarios.Compared to traditional manual and knowledge graph-based methods,this method is more universal,flexible,and efficient.