Using text mining to analyze reflective essays from Japanese medical students after rural community placement
Using text mining to analyze reflective essays from Japanese medical students after rural community placement
复制标题
DOI：
10.1186/s12909-020-1951-x
复制
发表时间：
2020-02-06
影响因子：
3.6
通讯作者：
Matsumura, Masami
中科院分区：
文献类型：
--
作者：
Lebowitz, Adam;Kotani, Kazuhiko;Matsumura, Masami
在社区临床实习之后，医学生使用反思性写作来发现他们成为医疗专业人员的旅程的故事。然而，由于评估者的偏见，定性地分析这些写作来概括学习者的经验可能是有问题的。本研究采用过程导向的文本挖掘方法，通过连接扩展的学生反思文章中的关键概念，更好地理解学习者经验的意义。方法采用文本挖掘定量分析的自我评价的文章（n = 47，独特的字数范围43-575），由五年级的学生在日本的一个区域配额系统的大学，专门培训全科医生服务不足的社区。首先，确定了六个高频率的关键词：患者，系统治疗，现场，医院，护理和培训。然后，标准化的关键字频率分析鲁棒性的整体文章长度和关键字量使用单个关键字作为“节点”来计算每一篇文章的每个关键字值。最后，运用主成分分析和回归分析方法对关键词关系进行分析。结果关键字区域的成分负荷最强，表明最共享的方差。对其余三个关键词医院、全身治疗和培训进行多重回归，得出R-2 = 0.45，认为该探索性研究的R-2较高。相比之下，学生的直接患者经验很难概括。结论实习生对实习区环境的印象最强烈，这些印象受到医院工作场所、治疗提供和培训的影响。文本挖掘可以以有效和客观的方式从更大的学生论文样本中提取信息，并识别学习情况之间的模式，以创建学习体验的模型。可能的影响，以社区为基础的临床学习可能是更好地了解学生的经验，为现场戒律受益他们的角色作为导师。
Background Following community clinical placements, medical students use reflective writing to discover the story of their journey to becoming medical professionals. However, because of assessor bias analyzing these writings qualitatively to generalize learner experiences may be problematic. This study uses a process-oriented text mining approach to better understand meanings of learner experiences by connecting key concepts in extended student reflective essays. Methods Text mining quantitative analysis is used on self-evaluative essays (n = 47, unique word count range 43-575) by fifth-year students at a regional quota-system university in Japan that specializes in training general practitioners for underserved communities. First, six highly-occurring key words were identified: patient, systemic treatment, locale, hospital, care, and training. Then, standardized keyword frequency analysis robust to overall essay length and keyword volume used individual keywords as "nodes" to calculate per-keyword values for each essay. Finally, Principle Components Analysis and regression were used to analyze key word relationships. Results Component loadings were strongest for the keyword area, indicating most shared variance. Multiply regressing three of the remaining keywords hospital, systemic treatment, and training yielded R-2 = 0.45, considered high for this exploratory study. In contrast, direct patient experience for students was difficult to generalize. Conclusions Impressions of the practicing area environment were strongest in students, and these impressions were influenced by hospital workplace, treatment provision, and training. Text mining can extract information from larger samples of student essays in an efficient and objective manner, as well as identify patterns between learning situations to create models of the learning experience. Possible implications for community-based clinical learning may be greater understanding of student experiences for on-site precepts benefitting their roles as mentors.