基于生成式人工智能大语言模型编写缺血性卒中患者出院用药教育材料的案例研究  

An case study of preparation of educational materials on discharge medication for ischaemic stroke patients based on a generative artificial intelligence large language model

作　　者：王天琳[1] 蔡乐[1] 董钊 何绵旺 杨滢霖 杨帆[1] 陈孟莉[1] WANG Tianlin;CAI Le;DONG Zhao;HE Mianwang;YANG Yinglin;YAN Fan;CHEN Mengli(Department of Pharmacy,Medical Supplies Center of Chinese People's Liberation Army General Hospital,Beijing 100853,China;Department of Neurology,First Medical Centre of Chinese People's Liberation Army General Hospital,Beijing 100853,China)

机构地区：[1]中国人民解放军总医院医疗保障中心药剂科,北京100853 [2]中国人民解放军总医院第一医学中心神经内科医学部,北京100853

出　　处：《临床药物治疗杂志》2025年第1期71-76,共6页Clinical Medication Journal

基　　金：军队课题(21BJZ36)。

摘　　要：目的 探索利用生成式人工智能(GenAI)大语言模型编写缺血性卒中患者出院用药教育材料的有效性和可行性。方法 系统检索PubMed、Embase、中国知网和万方数据知识服务平台数据库,搜集与缺血性卒中相关的临床实践指南和药物治疗建议。利用字节跳动公司推出的零代码人工智能(AI)开发平台构建智能体,通过随机抽取病历信息,结合提示词和对话模板,生成患者用药教育材料。神经内科专业临床药师对生成式人工智能(GenAI)输出内容进行准确性评估和调整。结果 检索得到相关文献978篇,最终纳入38篇文献形成知识库。GenAI智能体根据病历信息和知识库内容,成功生成了用药指导、监测提示和生活方式教育建议。经临床药师评估,GenAI输出内容未发现明显错误,中位输出时间为27.05(25.08~28.63)s,临床药师审核和修订的中位时间为74.50(67.50~102.00)s,完成1份出院用药教育材料的中位总时间为102.35(94.15~130.10)s。结论 基于GenAI大语言模型能够显著提高药师工作效率,为卒中患者提供个性化、易于理解的出院用药教育材料。但AI生成的内容仍需药师细致审核,以确保信息的准确性和适用性。Objective The aim of this study was to evaluate the effectiveness and feasibility of utilizing a generative artificial intelligence(GenAI)large language model for creating educational materials on discharge medications for ischemic stroke patients.Methods The study collected clinical practice guidelines and medication recommendations related to ischemic stroke by systematically searching the PubMed,EMBASE,CNKI and WanFang databases.Intelligent models were constructed using the zero-code AI development platform launched by ByteDance.Patient medication education materials were generated by randomly extracting medical record information and integrating prompt words and dialogue templates.The GenAI-generated content was assessed and adjusted for accuracy by clinical pharmacists specializing in neurology.Results The search yielded 978 relevant literature entries,of which 38 were ultimately included to form a knowledge base.The GenAI models successfully generated medication instructions,monitoring tips,and lifestyle education recommendations based on the medical record information and knowledge base content.As assessed by the clinical pharmacist,the GenAI output content did not contain any significant errors.The median output time was 27.05(25.08-28.63)seconds,while the median time for review and revision by the clinical pharmacist was 74.50(67.5-102.00)seconds.Consequently,the median total time for completing a discharge medication education material was 102.35(94.15-130.10)seconds.Conclusion Generative AI-based large language models can significantly enhance pharmacists'efficiency and provide personalized,easy-to-understand discharge medication education materials for stroke patients.However,AI-generated content still requires meticulous reviewed by pharmacists to ensure the accuracy and applicability of the information.

关 键 词：生成式人工智能 缺血性卒中 用药教育 临床药师 个性化指导 

分 类 号：R95[医药卫生—药学] R743