GAI问诊模型在诊断学教学中的实践探索与效果分析——与传统床旁教学的对照研究
杨泽利 , 时晓芳 , 梁文霞 , 汪羚利
中国医学教育技术 ›› 2026, Vol. 40 ›› Issue (4) : 436 -441.
GAI问诊模型在诊断学教学中的实践探索与效果分析——与传统床旁教学的对照研究
Practice exploration and effectiveness analysis of GAI consultation models in Diagnostics teaching——A comparative study with traditional bedside teaching
本研究旨在探讨生成式人工智能问诊模型在诊断学教学中替代传统床旁教学(基于病房患者)的可行性及其效果。研究选取首都医科大学延庆教学医院2023、2024级“3+2”助理全科医师规范化培训的医学生为研究对象,共73 人。其中,2023级学生及2024级甲组学生共38 人设为对照组;2024级乙组及丙组学生共35 人为试验组。试验组采用基于Claude 2的生成式人工智能(generative artificial intelligence,GAI)问诊模型进行问诊训练,对照组则采用传统床旁教学(基于病房患者)进行问诊实践。通过教师评价和学生自评两种方式,比较两组教学效果。研究结果显示,教师评价结果显示,试验组学生在问诊流程[(91.40±2.89)分 vs. (89.34±4.12)分]和病史采集[(91.31±2.35)分 vs. (88.55±3.45)分]两个维度的得分均高于对照组,差异具有统计学意义(均P<0.05)。在临床思维方面,两组得分亦存在差异[(90.34±2.21)分vs. (89.29±3.30)分,P=0.048]。其中,试验组在人文关怀维度的提升最为突出,得分显著高于对照组[(97.86±1.46)分 vs. (93.71±2.88)分,P<0.001]。学生自评结果显示,两组在专业知识掌握、临床思维能力、医患沟通能力等方面两组差异无统计学意义(均P>0.05)。研究结果表明,基于Claude 2的GAI问诊模型能够有效提升学生在问诊流程规范性、病史采集完整性及人文关怀行为演练方面的表现,学生对该教学方法及效果的满意度也更高,其整体教学效果与传统床旁教学相当,可作为现有教学资源的有益补充进行推广应用。
This study aimed to evaluate the feasibility and effectiveness of replacing traditional bedside teaching (based on ward patients) with a generative artificial intelligence consultation model in Diagnostics education. A total of 73 medical students enrolled in the “3+2” assistant general practitioner standardized training program from the class of 2023 and 2024 at Capital Medical University Yanqing Teaching Hospital were selected. Students from the class of 2023 and group A of the class of 2024 (n=38) served as the control group, while groups B and C of the class of 2024 (n=35) constituted the experimental group. The experimental group received consultation training using a GAI (generative artificial intelligence) consultation model based on Claude 2, whereas the control group underwent traditional bedside teaching with ward patients. Teaching outcomes were assessed through both teachers’ evaluation and students’ self-evaluation. Teachers’ evaluation demonstrated that the experimental group outperformed the control group in consultation procedure [(91.40±2.89) vs. (89.34±4.12)] and history-taking [(91.31±2.35) vs. (88.55±3.45)], with both differences reaching statistical significance (both P<0.05). A between-group difference was also observed in clinical reasoning [(90.34±2.21) vs. (89.29±3.30), P=0.048]. Notably, the most pronounced improvement in the experimental group was found in humanistic care [(97.86±1.46) vs. (93.71±2.88), P<0.001]. However, students’ self-evaluation revealed no statistically significant differences between the two groups in terms of professional knowledge mastery, clinical reasoning ability, or doctor-patient communication skills (all P>0.05). According to the results, the GAI consultation model based on Claude 2 effectively improved students’ performance in consultation procedure standardization, history-taking completeness, and humanistic care practice, while also receiving higher student satisfaction. Its overall teaching effectiveness was comparable to that of traditional bedside teaching, suggesting that it can serve as a valuable supplementary tool to existing teaching resources for broader application.
| [1] |
曹湘玉, 胡柯, 黄民江, |
| [2] |
|
| [3] |
段志军, 金艳玲, 杜建玲, |
| [4] |
高志荣, 李青松. 标准化病人在诊断学实训问诊教学中的应用和体会[J]. 中国高等医学教育, 2019(9): 101-102. |
| [5] |
王朝英. 基于5P教学模式的英语语音教学设计研究[J]. 教育进展, 2023, 13(10): 7308-7313. |
| [6] |
王成红, 唐晓鸿, 张柯, |
| [7] |
李翔, 陈晓云, 张慧群, |
| [8] |
王贤良, 葛昭, 赵志强, |
| [9] |
|
| [10] |
|
| [11] |
宋雨欣, 李子怡, 丁宁.生成式人工智能对医学教育的影响探析[J]. 中国医学教育技术, 2024, 38(3): 281-286. |
| [12] |
张雷, 王安生, 刘戈, |
| [13] |
|
| [14] |
|
| [15] |
Anthropic. Model card and evaluations for claude models[R/OL]. San Francisco: Anthropic, 2023. [2025-04-29]. |
| [16] |
|
| [17] |
陈湘, 邓然, 吴川清. 生成式人工智能大型语言模型在医学教育实践的探讨[J]. 临床急诊杂志, 2024, 25(6): 310-314. |
| [18] |
|
| [19] |
|
| [20] |
黄泽平, 张文君, 李思敏, |
| [21] |
丛慧源, 玄明实, 朱丹丹, |
北京大学第三医院教学研究专项课题(2024bysyxykzp02)
/
| 〈 |
|
〉 |