Summary of the used input and output categories.

The efficiency and productivity evaluation process commonly employs Data Envelopment Analysis (DEA) as a performance tool in numerous fields, such as the healthcare industry (hospitals). Therefore, this review examined various hospital-based DEA articles involving input and output variable selection approaches and the recent DEA developments. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology was utilised to extract 89 English articles containing empirical data between 2014 and 2022 from various databases (Web of Science, Scopus, PubMed, ScienceDirect, Springer Link, and Google Scholar). Furthermore, the DEA model parameters were determined using information from previous studies, while the approaches were identified narratively. This review grouped the approaches into four sections: literature review, data availability, systematic method, and expert judgement. An independent single strategy or a combination with other methods was then applied to these approaches. Consequently, the focus of this review on various methodologies employed in hospitals could limit its findings. Alternative approaches or techniques could be utilised to determine the input and output variables for a DEA analysis in a distinct area or based on different perspectives. The DEA application trend was also significantly similar to that of previous studies. Meanwhile, insufficient data was observed to support the usability of any DEA model in terms of fitting all model parameters. Therefore, several recommendations and methodological principles for DEA were proposed after analysing the existing literature.

效率与生产率评估流程通常将数据包络分析（Data Envelopment Analysis，DEA）作为绩效评估工具，广泛应用于包括医疗保健行业（医院）在内的诸多领域。有鉴于此，本综述梳理了各类面向医院场景的DEA相关研究，涵盖投入产出变量选择方法及DEA领域的最新进展。本研究采用系统评价与元分析优先报告条目（Preferred Reporting Items for Systematic Reviews and Meta-Analyses，PRISMA）方法，从Web of Science、Scopus、PubMed、ScienceDirect、Springer Link及Google Scholar等多个数据库中，筛选得到2014至2022年间发表的89篇包含实证数据的英文文献。此外，本研究参考既往研究成果确定DEA模型参数，并通过叙述性分析梳理各类变量选择方法。本研究将上述方法划分为四大类别：文献综述法、数据可得性导向法、系统方法论及专家判断法。随后，针对上述方法可采用单一独立策略或与其他方法联用的组合策略开展应用。综上，本综述聚焦医院场景下的各类方法，其研究结论存在一定局限性：针对其他领域或基于不同视角开展DEA分析时，可采用其他方法或技术确定投入产出变量。DEA的应用趋势与既往研究整体相似；同时，现有数据不足以支撑任意DEA模型适配全部参数的可用性验证。因此，本研究在系统梳理现有文献后，提出了若干DEA领域的研究建议与方法学原则。