Department of Artificial Intelligence and Informatics, Mayo Clinic, Rochester, MN, United States.
Mayo Clinic Libraries, Mayo Clinic, Rochester, MN, United States.
Int J Med Inform. 2022 Sep;165:104833. doi: 10.1016/j.ijmedinf.2022.104833. Epub 2022 Jul 14.
We performed a scoping review of informatics core literature about medical practice variation (MPV) as an agile summary of the subject in our field.
The Ovid integrated database was searched between 1946 and 2022 to identify MPV studies published in major informatics journals and conference proceedings. Two reviewers performed relevance screening, with assistance from another independent reviewer for adjudication. The included articles were then thematically analyzed and summarized through discussion among all three reviewers.
A total of 43 articles were included and went through the thematic analysis. About half (n = 21) of the included articles were published in conference proceedings. Five articles reported the effect of MPV on patient outcomes. The variation of interest was most frequently in treatment decisions. In terms of the role informatics played (multiple roles allowed), 39 (90.7%) articles pertained to detection of MPV, 5 were about prevention of MPV and 4 about learning from MPV.
MPV remains a critical issue in health care, yet most informatics research has been focused on simple tasks such as automating the detection of MPV and assessing compliance to decision-support systems, and less focused on addressing the causes of variation or supporting learning from variation.
Our scoping review found that informatics studies have focused on detecting of MPV, especially variability in treatments and deviation from practice guidelines. Technological advances should promote more informatics research focused on explaining and learning from MPV.
我们对医学实践变异性（MPV）的信息学核心文献进行了范围综述，作为对我们领域中该主题的快速总结。
在 1946 年至 2022 年间，我们在 Ovid 集成数据库中搜索了主要信息学期刊和会议论文集中发表的关于 MPV 的研究。两位审稿人进行了相关性筛选，另一位独立审稿人协助裁决。然后，通过所有三位审稿人之间的讨论，对纳入的文章进行主题分析和总结。
共纳入 43 篇文章进行主题分析。纳入的文章中有一半（n=21）发表在会议论文集上。有 5 篇文章报道了 MPV 对患者结局的影响。感兴趣的变异最常发生在治疗决策中。就信息学所扮演的角色（允许多个角色）而言，39 篇（90.7%）文章涉及到 MPV 的检测，5 篇文章涉及到 MPV 的预防，4 篇文章涉及到从 MPV 中学习。
MPV 仍然是医疗保健中的一个关键问题，但大多数信息学研究都集中在简单的任务上，例如自动检测 MPV 和评估对决策支持系统的遵守情况，而较少关注变异的原因或支持从变异中学习。
我们的范围综述发现，信息学研究集中在检测 MPV 上，特别是在治疗方法的变异性和偏离实践指南方面。技术进步应该促进更多的信息学研究，重点是解释和从 MPV 中学习。
