2.1 Search protocol
With respect to the objective of the present research, a structured literature review (SLR) is carried out. SLR is a technique used to extract information, in order to identify trends and future developments in any research field [135, 110, 90]. In recent times, SLR is a widely used approach to summarize, categorize and identify the gaps in the previous scientific literature, thanks to the availability and transparency of the involved articles [90]. Although there are many techniques to make a literature review of previous scientific articles [132, 40], the authors have chosen the SLR approach because it ensures the replicability of the study, in terms of transparency and clarity [90]. The replicability is made possible following certain stages, such as the explanation of research questions, search protocol, and assessment linked to articles, framework, and analysis to carry out [121].
SLR consists of four stages. As stated by Massaro et al., [90], defining the research questions (RQs) is the first stage to perform the SLR. The research questions are linked to the state of the art, the main research block related to the topic of the studies, and the research developments. Therefore, it is possible to formulate the following RQs:
RQ1: What is the state of the art in academic research about healthcare efficiency?
RQ2: What are the main research blocks related to this subject area?
RQ3: What are the recommended future developments about this research topic?
In particular, the first research question aims to describe the state of the art of the research topic in the current scientific literature and sheds light on how researchers take into account the analysed topic. More in detail, to answer this question, the authors intend to examine the influential characteristics of studies, considering authors, journals, citations, and countries, and also the clustering and the classification of studies into different research areas.
Then, the second research question intends to detect the main research block related to the efficiency in healthcare sector.
Finally, the third and last question aims to make recommendations on future developments about the research topic of this study, considering new insights and more implications for managers and policy-makers.
The second stage of the SLR is the identification of the articles that composed the sample, in order to perform an extensive literature review. To make up the sample the authors have searched the articles on the Scopus database. Scopus appears the most used and complete database, in which there are more than 20,000 peer-reviewed scientific journals [93], and it includes almost 97% of the articles indexed in the Web of Science (WoS) database [112]. For this purpose, the Scopus database appears to be the best way to conduct a comprehensive SLR.
The third stage of the analysis consists of different types of descriptive analysis, in order to explore the timeframe of articles’ publication and its evolution (number of articles published over time, number of citations per article over time), the geographical analysis (number of articles identified per countries, number of citations per countries); journals (number of articles per journal; number of citation per journal); authors ranking (number of citations; citations per year); relevant keywords and research areas (frequency of keywords, groups of keywords occurrences; bibliographic coupling, detection of relevant research areas).
The final stage of the analysis, is the identification of core contents and the determination of the research agenda, thanks to the considerations that may be drawn from the study.
Practically, to define the sample on the Scopus database, the authors have built a search string through the use of keywords and parameters. The search query of the present analysis considers the search of keywords within the title, abstract, and keywords. Specifically, the keywords used are: “Healthcare” AND “Efficiency” OR “Healthcare” AND “Efficiency” AND “Determinants” OR “Healthcare” AND “Efficiency” AND “Effects” OR “Healthcare” AND “DEA” OR “Healthcare” AND “SFA”. However, given the broad scope of the topic, it is possible to run into some risks. In this sense, SLR is committed to reducing these risks, especially when protocols are used. From this perspective, as stated by Petticrew and Roberts [110], a standardized protocol intends to reduce the false negative (sources not identified) and to increase the false positives (sources irrelevant). In the present study, the authors perform the analysis based on the PRISMA framework, to reduce as much as possible, the bias that incurs during the analysis (see Fig. 1). To carry out the specification of the sample, the authors run in a replicable and transparent way. After the introduction of the search string with the aforementioned keywords, a total of 20,501 articles are found. Data are collected in April 2023. Successively, the authors have limited the research to source and document type, publication stage, subject area, and language. They have considered only English articles published in the final stage of journals, under the area of “Business, Management and Accounting” and “Economics, Econometrics and Finance”. All the authors have read the article’s title and abstract to define the final sample to be used for the analysis. More in detail, after the application of the following stages, the identification of the sample is carried out: firstly, after the search with keywords, 20,501 articles are identified from the Scopus database. Secondly, 20,402 articles are excluded because they don’t follow the criteria established. Thirdly, 702 articles are excluded after reading of title and abstract, because the articles don’t match with the general topic of the research. Finally, 292 are the articles used both within the descriptive and content analysis.
The analyses conducted are divided into two broad categories: descriptive statistics and clustering and content analysis. The first category includes evaluation linked to the articles’ evolution, and the geographical examination, to explore where and how scientific literature on the topic is developed [89]. Another analysis is performed to measure the citation and the citation per year (CPY). In addition, keywords and bibliographical coupling analysis are carried out, to define the most relevant and used ones [122]. Specifically, for keywords the authors make a co-occurrence analysis [140] that is the relatedness of articles based on the number of articles in which they occur together. In the present analysis, the authors have included all papers in which keywords occur 5 times at least. Concerning bibliographic coupling, the result is the clustering of papers that share the same references. In this case, the authors have chosen to consider the articles with at least 10 citations, due to the sample size, through fractional counting [85]. Finally, the authors have carried out the content analysis to identify the gaps and trends of previous research, and successively to define future developments.
The analysis is conducted on VOSviewer, which is a software used to perform and visualize bibliometric analysis [139].