基于德尔菲法的研究生教育'分流退出'路径探索 Research on the Path of 'Diversion Withdrawal' in Graduate Education Based on Delphi Method

目的 探索研究生教育“分流退出”路径，为“分流退出”制度的实施提供参考。 方法 依据首都医科大学研究生工作手册相关规定制订“分流退出”条目，采用德尔菲法开展两轮函询调查。通过计算条目均值、变异系数、满分比等指标，确定函询条目，形成“分流退出”制度条目，在此基础上，绘制“分流退出”路径图。 结果 本研究于2023年11月共进行两轮函询，函询专家共12名，问卷有效回收率均为100%，最终形成了“分流退出”管理条文，包括退出和分流两个维度，其中退出涵盖日常管理、培养过程和学位管理三方面共22项条目，分流即转科学学位，共4项条目。第1轮专家权威系数为0.892，第2轮专家权威系数为0.896，专家的权威程度较高，评价结果可靠。两轮肯德尔协调系数分别为0.264和0.306（P＜0.05），提示咨询专家意见较为一致，评价结果可信。第2轮函询问卷，各个评价条目的算术均值均＞3.50，且变异系数＜0.25，条目予以保留。满分比在80%以上的条目有10个。条目中最低平均分为3.833分，最高平均分为满分5分。 结论 本研究函询结果科学可靠，探索了研究生教育“分流退出”路径，能够为研究生教育“分流退出”制度实施提供参考，对保证制度落实、保障研究生教育质量，具有重要意义。 Abstract: Objective To explore the path of “diversion withdrawal” in graduate education and provide a reference for implementing the “diversion withdrawal” system. Methods According to the relevant regulations of the work manual of the graduate students of Capital Medical University, the items of “diversion withdrawal” were established, and the Delphi method was used to carry out two rounds of correspondence investigation. By calculating the mean value, coefficient of variation, full score ratio, and other indicators, the items were determined and the system items of “diversion withdrawal” were formed. On this basis, the path diagram of “diversion withdrawal” was drawn. Results In this study, two rounds of correspondence were conducted in November 2023, with 12 experts consulted, and the effective recovery rate of the questionnaire was 100%. Finally, the management provisions of “diversion withdrawal” were formed, including two dimensions of withdrawal and diversion, in which withdrawal covered 22 items in three aspects of daily management, training process, and degree management, while diversion, namely transfer degree, included 4 items. The expert authority coefficient of the first round was 0.892, and that of the second round was 0.896, indicating a high degree of expert authority and reliable evaluation results. The Kendall’s concordance coefficients of the two rounds were 0.264 and 0.306, respectively (P3.50, and the coefficient of variation was