基于生成式大语言模型的病历内涵质控智能助手的构建与应用

Construction and practice of medical record connotation quality control intelligent assistant application based on generative large language model

摘要

目的 针对传统病历质控工具依赖规则引擎、检错率低、语义理解不足等问题,探索生成式大语言模型在病历内涵质控中的应用价值。方法 基于 Qwen2.5-32B-Instruct生成式大语言模型,通过 LoRA 高效微调技术和 Chain of Thought方法构建病历内涵质控助手,集成智能感知、实时反馈和提示词优化功能,覆盖主诉一致性、诊断依据充分性等核心质控任务;通过实际应用并观察其运行效果,评估该助手的有效性。结果 病历内涵质控助手上线前后3个月数据统计显示,质控条数提升24.6%,病历质控申诉率较未上线前降低53.9%,误报率降低62.6%。结论 Qwen2.5-32B-Instruct生成式大语言模型能显著提升病历内涵质控的准确性和效率,为医疗质量改进提供智能化支持。

Abstract

Objective To address the issues of traditional medical record quality control tools, such as reliance on rule engines, low error detection rates, and insufficient semantic understanding, this study aims to explore the application value of large language models in the quality control of the connotation of medical records. Methods Based on the Qwen2.5-32B-Instruct generative large language model, a Medical Record Connotation Quality Control Assistant was constructed using the LoRA efficient fine-tuning technology and the Chain of Thought method. It integrates functions of intelligent perception, real-time feedback, and prompt optimization, covering core quality control tasks such as the consistency of the chief complaint and the sufficiency of diagnostic evidence; the effectiveness of the assistant was evaluated through practical application and observation of its operation effect. Results Statistics on the data of 3 months before and after the launch of the Medical Record Connotation Quality Control Assistant showed that the number of quality control items increased by 24.6%, the medical record quality control appeal rate decreased by 53.9% compared with that before the launch, the false positive rate decreased by 62.6%. Conclusion Generative large language models can significantly improve the accuracy and efficiency of the quality control of the connotation of medical records, providing intelligent support for the improvement of medical quality.

关键词

生成式大语言模型 / 病历内涵质控 / 智能医疗 / LoRA 微调 / 语义推理

Key words

generative large language model / quality control of the connotation of medical records / intelligent healthcare / LoRA fine-tuning / semantic reasoning

参考文献

[1] 厉伟民, 陈翔, 李斐铭, 等 . 基于电子病历系统的临床医疗质量实时控制[J]. 中华医院管理杂志, 2012, 28(5): 347-351.

[2] LI W M, CHEN X, LI F M, et al. EMR—based real—time monitoring of clinical care quality[J]. Chin J Hosp Adm, 2012, 28(5): 347-351.

[3] 周书名 . 从医疗纠纷诉讼角度看病历档案质量问题及提出对策[J]. 兰台内外, 2023(36): 59-61.

[4] ZHOU S M. Analysis of the quality issues of medical record archives from the perspective of medical dispute litigation and countermeasures[J]. Inside Outside Lantai, 2023(36): 59-61.

[5] 国家卫生健康委办公厅 . 关于印发卫生健康行业人工智能应用场景参考指引的通知(国卫办规划函〔2024〕420 号)[R/OL]. (2024—11—06) [2024—11—14].https://www.nhc.gov.cn/guihuaxxs/c100133/202411/3dee425b8dc34f739d63483c4e5c334c.shtml.

[6] Office of the National Health Commission. Notice on issuing reference guidelines for artificial intelligence application scenarios in the health industry (national health office planning letter [2024] No.420)[R/OL]. (2024—11—06) [2024—11—14].https://www.nhc.gov.cn/guihuaxxs/c100133/202411/3dee425b8dc34f739d63483c4e5c334c.shtml.

[7] 毛淋淇, 杨静, 徐婉瑛 . 基于院科两级质控管理体系的终末病历质量督查模式探讨[J]. 中国卫生质量管理, 2021, 28(9): 32—34, 41.

[8] MAO L Q, YANG J, XU W Y. Quality supervision mode of final medical records based on hospital and department quality control management system[J]. Chin Health Qual Manag, 2021, 28(9): 32—34, 41.

[9] 宋雪君 . 电子病历在医疗质量控制管理中的应用与研究[J]. 现代计算机, 2023, 29(6): 88-90.

[10] SONG X J. Application and research of electronic medical record in medical quality control management[J]. Mod Comput, 2023, 29(6): 88-90.

[11] 孟岩, 孙雪梅, 王雪, 等 . 人工智能技术应用于电子病历质控的研究与思考[J]. 中国卫生质量管理, 2021, 28(12): 63-65.

[12] MENG Y, SUN X M, WANG X, et al. Research and thinking on the application of artificial intelligence technology in quality control of electronic medical records[J]. Chin Health Qual Manag, 2021, 28(12): 63-65.

[13] 崔少国, 陈俊桦, 李晓虹 . 融合语义及边界信息的中文电子病历命名实体识别[J]. 电子科技大学学报, 2022, 51(4): 565-571.

[14] CUI S G, CHEN J H, LI X H. Named entity recognition for Chinese electronic medical record by fusing semantic and boundary information[J]. J Univ Electron Sci Technol China, 2022, 51(4): 565-571.

[15] 苏龙翔, 翁利, 李文雄, 等 . 大型语言模型在重症医学中的应用与挑战[J]. 中华医学杂志, 2023, 103(31): 2361-2364.

[16] SU L X, WENG L, LI W X, et al. Applications and challenges of large language models in critical care medicine[J]. Natl Med J China, 2023, 103(31): 2361-2364.

[17] 谢思静, 文鼎柱 . 基于联邦分割学习与低秩适应的 RoBERTa 预训练模型微调方法[J]. 数据采集与处理, 2024, 39(3): 577-587.

[18] XIE S J, WEN D Z. Fine—tuning method for pre—trained model RoBERTa based on federated split learning and Low—Rank adaptation[J]. J Data Acquis Process, 2024, 39(3): 577-587.

[19] 李亚玲, 蔡京京, 柏洁明 . 生成式大模型引发的隐私风险及治理路径[J]. 智能科学与技术学报, 2024, 6(3): 394-401.

[20] LI Y L, CAI J J, BAI J M. Privacy risks induced by generative large language models and governance paths[J]. Chin J Intell Sci Technol, 2024, 6(3): 394-401.

基金资助

四川省2025年重点产业链科技攻关项目(2025YFRG0004)