生成式人工智能构建患者药品说明书的方法研究

Methods for constructing patient medication instructions using generative artificial intelligence

目的建立患者药品说明书的基本框架,研究基于生成式人工智能构建患者药品说明书的方法,生成患者更易了解的药品说明书,促进合理用药。方法比较5个不同国家(地区)患者药品说明书的异同,纳入4个以上国家(地区)患者药品说明书均提及的核心元素,组成患者药品说明书基本框架。基于此框架并利用生成式人工智能方法及提示语策略,提供患者药品说明书范例。结果对美国、欧盟、日本、澳大利亚患者药品说明书进行了比较。综合各自优点建立了包含药品名称、适应证、用法用量、注意事项、不良反应、贮存方法、药品信息、生产厂家等核心元素的患者药品说明书框架。采用“通义千问”2.0模型,生成了普瑞巴林患者药品说明书案例。与原药品说明书相比,患者药品说明书的长度从15395字减少到1052字,可读性较前有明显改善,准确性为77.8%;由高年资药师人工把关后其准确性可达到与原说明书相同。结论构建了患者药品说明书框架,生成式人工智能可高效生成可读性更优的患者药品说明书,药师审定后保证其正确性。

Objective To establish a basic framework for patient medication instructions,study the method of construct⁃ing patient medication instructions based on generative artificial intelligence,generate medication instructions that are easier for patients to understand,and promote rational medication use.Method Compared the similarities and differences in patient medication instructions across five countries,We incorporate core elements mentioned in the medication instructions of four or more countries,to form a basic framework for patient medication instructions.Based on this framework and using generative artificial intelligence and prompt,We provide examples of patient medication instructions.Results The patient medication in⁃structions from the USA,EU,Japan,and Australia were compared.A framework for patient medication instructions was estab⁃lished,incorporating information such as drug name,indications,dosage and administration,precautions,adverse reactions,storage methods,drug information,and manufacturer details.Prompt included directive words,background descriptions,out⁃put requirements,and related materials.Using the QWEN 2.0 model,an example of patient medication instructions to pregab⁃alin was generated.Compared to the original medication instructions,the length of the patient medication instructions was re⁃duced from 15395 words to about 1052 words,showing a significant improvement in readability.The accuracy was 77.8%,and with the oversight of experienced pharmacists,the accuracy could reach the same level.Conclusion We constructed a framework for patient medication instructions.Generative artificial intelligence can efficiently produce patient medication in⁃structions with improved readability,which were then reviewed and approved by pharmacists to ensure accuracy.