| 引用本文: | 宁怿轩,杜青青.ChatGPT在消化系统疾病用药咨询中的应用研究[J].中国现代应用药学,2025,42(8):155-161. |
| ningyixuan,duqingqing.Application of ChatGPT in medication consultant for digestive diseases[J].Chin J Mod Appl Pharm(中国现代应用药学),2025,42(8):155-161. |
|
| 摘要: |
| 摘要：目的 ChatGPT的出现为药学服务提供了新的思路和工具。本研究的目的是探索ChatGPT在消化系统疾病用药咨询中的应用前景，同时为ChatGPT用于药学服务相关工作提供参考及借鉴。方法 以《消化系统疾病用药咨询标准化手册》为依据构建问题库，对ChatGPT进行提问，采用双人核对的方式对ChatGPT输出的答案进行评估和分析，同时比较ChatGPT 3.5和4的差异。结果ChatGPT3.5回答的正确率为35.41%，不完整率为33.3%，错误率为31.25%。ChatGPT4.0回答的正确率为77.1%，不完整率为8.3%，错误率为14.6%。ChatGPT4.0的正确率高于3.5。在不同疾病类型的回答中，ChatGPT 3.5对于腹泻用药咨询的正确率最高，为62.5%，消化不良用药咨询的正确率最低，为22.2%；ChatGPT4.0对于腹泻及其他用药咨询的正确率最高，为100%，消化不良用药咨询的正确率最低，为55.6%。不同问题类型的回答中，ChatGPT 3.5对于适应证咨询的正确率最高，为66.7%，用法用量咨询的正确率最低，为13.3%；ChatGPT 4.0对于疗程、治疗方案、相互作用和注意事项咨询的正确率均为100%，不良反应咨询正确率最低，为62.5%。 结论 ChatGPT有应用于消化系统疾病用药咨询的潜力，ChatGPT4.0优于3.5，但其准确性仍需进一步加强。ChatGPT应用于其他药学服务工作的获益和风险有待进一步评估。 |
| 关键词: ChatGPT 用药咨询 消化系统疾病 人工智能 药学服务 |
| DOI： |
| 分类号: |
| 基金项目:2021年度重庆市自然科学基金面上项目，cstc2021jcyj-msxmX0218；重庆医科大学附属第二医院“青年人才育苗基金” |
|
| Application of ChatGPT in medication consultant for digestive diseases |
|
ningyixuan1, duqingqing2
|
|
1.Chongqing Medical University;2.the Second Affiliated Hospital of Chongqing Medical University,Chongqing
|
| Abstract: |
| ABSTRACT:OBJECTIVE The emergence of ChatGPT provided new ideas and tools for pharmaceutical care. This study aimed to explore the application prospect of ChatGPT in medication consultant for digestive diseases, and provide reference for ChatGPT in pharmaceutical care. METHODS A question database was constructed based on the Standardization Manual of Medication Consultant for Digestive Diseases, and the answers from ChatGPT were evaluated and analyzed by two-person check, and the differences between ChatGPT 3.5 and 4 were compared. RESULTS The correct rate of ChatGPT3.5 was 35.41%, incomplete rate was 33.3%, and error rate was 31.25%. The correct rate of ChatGPT4 was 77.1%, incomplete rate was 8.3%, and error rate was 14.6%. The accuracy of ChatGPT4.0 is higher than ChatGPT3.5. Among the answers of different disease types, the highest correct rate of ChatGPT3.5 was 62.5% for diarrhea medication consultation and the lowest correct rate was 22.2% for indigestion medication consultation. The highest accuracy of ChatGPT4.0 was 100% for diarrhea and other medication consultation, and the lowest accuracy was 55.6% for dyspepsia medication consultation. Among the answers of different questions, the highest correct rate of ChatGPT 3.5 was 66.7% for indication consultant and the lowest correct rate was 13.3% for usage and dosage consultant. The accuracy of ChatGPT 4.0 was 100% for consultant on course of treatment, treatment plan, interaction and precautions, and the accuracy of consultant on adverse reactions was the lowest (62.5%). CONCLUSION ChatGPT has the potential to be used in medication consultant for digestive diseases. ChatGPT4.0 is better than 3.5, but its accuracy still needs to be further strengthened. The benefits and risks of applying ChatGPT to other pharmaceutical care work need to be further evaluated. |
| Key words: medication consultant digestive diseases artificial intelligence pharmaceutical care |
