层次聚类分析在高校附属医院科研绩效评估中的应用

Hierarchical clustering analysis application in research performance appraisal of university affiliated hospital

目的优化医院科研绩效评估,挖掘不同类型临床科室的科研特点及不足,提升科研绩效评估的有效性。方法采用统计学描述概括2017年医院科研绩效评分整体情况;采用层次聚类分析对75个临床科室科研绩效评分进行聚类与特征分析。结果SCI论文(405分)、国家级课题(181分)及国内核心论文(106分)3项占总平均分75.2%,为主要科研业绩指标。75个科室最佳聚类数为6类,树状图显示6类科室之间具有显著的差异性。6类科室类型分别为学术会议导向型、国内核心期刊论文导向型、SCI论文导向型、高级别课题导向型、省部级及横向课题导向型、学术交流导向型。其导向指标平均分占各自类别总平均分百分比分别为:59.4%、42.0%、66.7%、57.0%、61.8%和52.3%。结论本样本中,相比K-均值法,层次聚类分析法的结果具有更好的特征性与解释性,科研绩效评估得到了进一步优化。不同类型科室的指标特点显著,薄弱点突出,为科研管理者对策研究提供了有效的指导。

Objective The research aims to optimize the hospital research performance appraisal, clarify the scientific characteristics and possible shortages of different clinic departments in various types, to enhance the effectiveness of research performance appraisal. Methods Descriptive statistics were used to generalize hospital research performance appraisal in 2017. Hierarchical clustering was used to cluster and analyze performance appraisal characteristics of 75 clinic departments. Results SCI papers (405), National projects (181) and National core journal papers (106) took 75.2% of total average score. The optimal solution of the cluster was 6 types for 75 departments and the dendrogram illustrated significant varieties among the 6 types. Six departments’ types were academic-conference-oriented, national-paper-oriented, SCI-paper-oriented, advanced-project-oriented, provincial/ horizontal-project-oriented and attending-conference-oriented. The percentages of the oriented indicators that took their total average scores were 59.4%, 42.0%, 66.7%, 57.0%, 61.8%, 52.3%. Conclusions Compared with K-means method, the results of hierarchical clustering equipped with better characteristics and interpretative power. Research performance appraisal has been further optimized. Departments in different types showed significant characteristics and weaknesses, which provides managers with effective guidance on countermeasures.