Implementation of STEM education: A bibliometrics analysis from case study research in Scopus database

In response to the long-term challenges of skills shortages and mismatches, many countries worldwide have made significant investments in STEM education. However, to obtain a comprehensive understanding of the progress of global research on the implementation of STEM education, it is essential to compile case studies on STEM education. Therefore, we conducted a bibliometric analysis of 750 publications from Scopus database to assess the bibliographic content of case studies on STEM education from 2006 to 2022. Our main findings show that the first case study on STEM education dates back to 2006, and research in this direction has grown enormously and continuously over the last five years (counting for 72.9%). The most influential articles in the field are cited mainly in the background citations section and help provide an initial perspective for readers to understand the insights discussed later in the study. The most popular journals publishing case studies in STEM education are of good quality, ranking highly in Scopus and Web of Science categories. The analysis results show three main research directions: STEM education in higher education, STEM education expands to STEAM, and STEM education activities in K12-education.


INTRODUCTION
In response to the long-term challenges of skills shortages and mismatches, many countries worldwide, including the USA, have invested in STEM education to preserve future employment opportunities (Friedman, 2005).This emphasis on promoting STEM education to enhance workforce capacity has become widespread among educators and policymakers in developed and emerging countries, such as the UK, Germany, Japan, Korea, and Israel (Drori, 2013;English, 2016;Marginson et al., 2013;Sanders, 2009) Over time, STEM education has been extensively studied and developed, creating a widely-accepted definition and theoretical framework.Common themes from this body of research include a focus on mathematics and science, an emphasis on the relationship between these subjects, and a preference for authentic learning contexts.Additionally, STEM education frequently adopts a learner-centered approach and incorporates process engineering design into learning activities (Ha et al., 2020;Huong et al., 2021).
The case study method is a widely used research approach in academia for researchers who are interested in conducting qualitative research (Baskarada, 2014;Yazan, 2015).A case study is a research method that explores a phenomenon in some specific context (its natural occurrence) through various data sources.It discovers many different prisms that reveal many

Phuong et al. / Implementation of STEM education
2 / 14 aspects of the phenomena (Baxter & Jack, 2008;Kaarbo & Beasley, 1999).Case studies show many advantages in real-world research.One of those advantages is that it helps researchers to quickly approach the nature of social reality, allowing generalization of a case or being able to generalize from a case to a class and focus on the subtlety and complexity of the case as it is in real life (Brown & Rodgers, 2002).Case studies often answer the research questions 'how' and 'why', which are more explanatory and focus on phenomena that occur in reallife contexts.This type of research identifies the complexity and inclusion of social truths.
A collection of case studies on STEM education would facilitate a comprehensive review of the worldwide research landscape pertaining to the practical implementation of STEM education.Such an analysis is currently absent from the literature and would furnish valuable quantitative insights to scholars who wish to explore this topic in future research.However, most of the available quantitative studies on STEM education (Gil-Doménech et al., 2020;Ozkaya, 2019), the role of STEM education (Jamali et al., 2022), analysis of the global landscape on STEM education (Zhan et al., 2022), or focus on STEM quantification in middle school education (Huong et al., 2021), higher education (Kundu et al., 2022), STEM education in specific territories such as ASEAN (Ha et al., 2020).Various techniques are available for conducting reviews, such as citation network analysis, content analysis, and topic mapping.Among these, bibliometric analysis is a valuable tool for quantitatively assessing scientific activity by examining the form and content of scientific literature (Broadus, 1987).The bibliographic analysis results provide insight into several key areas: (1) the annual growth rate of publications related to case studies in STEM education and corresponding citations, (2) the countries and scholars with the most impactful case studies in STEM education, (3) the most popular journals for publishing STEM education case studies, (4) the most cited articles among case studies in STEM education, and (5) analysis of the most significant research topics related to the implementation of STEM education worldwide.

METHODOLOGY
Bibliometric analysis was first introduced in the 1960s by Pritchard (1969) and has since been extensively utilized to examine scientific advancements across various fields, both at local and global levels.In this particular study, the methodology for data collection and analysis, as established by Ha et al. (2020), was employed to conduct bibliometric analysis.The data for this study were gathered from Scopus database, which is currently the largest and most widely used scientific database available.
As shown in Figure 1, all records in this database containing related keywords 'STEM education' and 'case study' in their title, summary, or keyword were searched.Keywords related to STEM are referenced by the research of Ha et al. (2020) and Huong et al. (2021), in addition, publications containing keywords related to the verb "stem" or the noun "stem cell" in the title, abstract, and keywords are also removed.The search query was then limited to articles, book chapters, conference papers, and reviews in the field of social science.Subsequently, we identified 806 records from Scopus database.The authors performed an abstract review of all of these publications to remove irrelevant research.The definitive collection of 750 articles was analyzed using the Biblioshiny and VOSviewer tools.
Biblioshiny is an open-source tool that facilitates a comprehensive cartographic analysis of scientific literature.It has been designed and developed using R programming environment, which offers significant flexibility and seamless integration with other graphics and statistical packages.The tool is highly adaptable, allowing quick updates and incorporating new features (Aria & Cuccurullo, 2017).VOSviewer is a software application designed to construct and visualize bibliometric networks.In addition, VOSviewer provides text-mining capabilities that facilitate the development and presentation of co-occurrence networks comprised of significant terms extracted from the scientific literature (van Eck & Waltman, 2010).The necessary information on the relevant authors, affiliations, and journals is also added from Scopus database.The author also provided details regarding the citation count of the most impactful articles as identified by Semantic Scholar (https://www.semanticscholar.org/).

General Information About Publication Collection
Table 1 presents the key findings from the analysis of 750 publications, which comprise 543 articles (72.4%), 146 conference papers (19.4%), 53 book chapters (7.1%), and eight reviews (1.1%).Notably, these publications were sourced from a total of 384 different sources.Articles using case studies in STEM education were published in 2006.However, in the next ten years, the number of studies on the same topic and research methods did not increase much.The growth model shown in Figure 2 is similar to the growth model of STEM education research in middle school (Huong et al., 2021) or STEM education in ASEAN (Ha et al., 2020).The anthology has been written by 2,223 authors (each document has 3.3 co-authors on average).83 single authors are publishing 84 singleauthor papers.At the time of this study, the publication collection had amassed 4,719 citations, resulting in an average citation count of 6,317 per publication.The h-index of the collection was calculated to be 30, indicating that among the 750 documents evaluated for the h-index, 30 had been cited at least 30 times.However, it is noteworthy that 209 papers (27.87% of the collection) had not received any citations.Nearly three-quarters of publications (555 publications) have less than average citations.Other information on the annual citation and its cumulative number for the period 2017-2021 is shown in Figure 2 and Table 2.

Most Productive Countries and Global Network of Cooperation
Table 3 presents the top ten most productive countries based on the number of publications and citations accumulated.The USA outnumber all other countries (n=433, accounting for 57.7% of total publications).Not only that, but the country also excels in total citations with 3397 citations (72.0% of total citations), equivalent to an average number of citations per document of 7.8.The second country in terms of publications is the UK (n=41, 5.5% of the total).With an impressive average of 8.7 citations per document and 357 citations, this country has the highest citation count among the countries examined in this study.The next are Turkey and Australia, with the same number of publications at 36 (5.5%) and the number of citations at 242 and 188, respectively.The two countries with 18 publications are Canada and China.The remaining countries on the list with corresponding publications are Hong Kong (n=18), Spain (n=18), Ireland (n=14), South Africa (n=11), and Portugal (n=11).Of the remaining countries on the list, only Canada has a citation count of 107; the rest have fewer than 60 citations.The USA's superiority in publications compared with other countries is quite similar to the results of Assefa and Rorissa (2013) on the same topic or Marín-Marín et al. ( 2021) in the study of STEAM in education.Overall, except for the top three countries, the rest of the countries have a lower average citation count than the average of the entire collection of publications on case studies in STEM education.
According to data from Scopus, 84 countries have publications related to case studies in STEM education.Figure 3 displays the transnational cooperation network comprising 36 partner countries.Each node in Figure 3 represents a country, while the buttons' colors denote the clusters of links between the countries.The node size corresponds to the number of publications, and the thickness of the lines connecting the nodes reflects the strength of the transnational collaboration.The countries in this network have at least three publications related to the published research topic.The network built in Figure 3 shows three main clusters: Red, green, and blue.The countries listed in Table 3 all appear in this network.In the above analysis, the USA has a much higher number of publications than the rest, but transnational research collaboration is more expansive than in this country.The cooperation between the remaining countries is also prosperous and diverse.

Most Published Sources
The collection of publications has been published from 384 different sources, including journals, scientific conferences, and book chapters.The top-10 publications that have published the most articles related to case study research in STEM education are presented in Table 4.These top-publications published 152 studies (20.3% of the total) and were cited 1,441 times (equivalent to 30.5% of the total citations).In this list of 11 sources, there are three conferences and eight journals.Education Sciences (MDPI Publishing House) is the journal with the highest number of publications in this field (24 papers), followed by Proceedings of Frontiers in Education Conference FIE (23 papers) and International Journal of STEM Education (Springer Nature Publishing House) is the third (20 articles).The remaining sources publish between 10 and 12 publications.Of the top-8 journals, only Education Sciences is in Scopus Q2 group of journals.Furthermore, this journal is the sole one included in the Emerging Sources Citation Index (ESCI) of the Web of Science (WoS) core collection.The remaining seven journals are all in the journals ranked Q1 according to Scopus's ranking.These are also journals in the ISI category of the WoS core collection.Studies from the most published

Most Effective Authors
The total number of authors who participated in the case study in STEM education recorded in Scopus database is 2,223.Table 5 exhibits the rankings of the ten most prolific scholars, as determined by the number of articles produced.It is noteworthy that a given scholar may leverage affiliations with multiple institutions, however, the data presented in Table 5 pertains exclusively to their most recent publications.In general, the number of publications by leading scholars is not too disparate.There are 7/10 scholars from the USA (Moore, T. J. with seven papers, Guzey, S. S. with six papers, Roehrig, G. H. published five papers, Biswas, G., Glancy, A. W., Kier, M. W., and Siverling, E. A. each with four    (Moore et al., 2015a(Moore et al., , 2015b)).Additionally, these two authors also collaborated on four other studies on the same topic.Sharing the same research trend, Muntean, C. H., and Roehrig, G. H. was first published in 2017 and lasted until 2022; however, we did not record research collaboration between these two authors.The remaining authors have several publications that cannot reflect research trends adequately.

Most Influential Studies
The first study recorded in Scopus database on a case study in STEM education was an exploratory case study examining the self-definition of eight graduate students who participated in a GK-12 program published in May 2006 (Buck et al., 2006).Nevertheless, it is essential to note that this study does not hold significance, as it only garnered eight citations at the time of the investigation.In contrast, Table 6 provides information on the ten most frequently cited works from the 750 studies within the collection.Table 6 includes additional details on the authors and publishers of the selected works.Within this list, nine works are journal articles, while one is a conference paper with 795 citations at the time of this study.Collectively, these 10 works account for 16.8% of the total citation count for the anthology.
The paper with the highest citation count (139 citations) examines the articulation and negotiation of in-practice identities in science by middle school girls desiring a career in STEM-related fields (Tan et al., 2013).The second most cited paper (with over 100 citations) describes the use of flipped classroom modules for large enrolment general chemistry courses as a low-barrier approach to increase active learning and improve student grades (Eichler & Peeples, 2016).This paper has  garnered considerable attention, with an annual citation count of 14.86.Two analytical papers examining black students' and parents' views of science and science careers (Archer et al., 2015) and the effects of STEMrelated after-school program activities on student learning (Sahin et al., 2014) occupy the third and fourth positions, respectively, with 87 and 83 citations.The only conference paper on this list (He et al., 2016) reports on a case study of courseware that integrates modern technology for embedded system courses to incorporate  (Han et al., 2015) and white faculty mentoring students of color (McCoy et al., 2015) received 65 citations each.Only one paper on this list is singleauthored, describing a pedagogical approach for early childhood computer science called "coding as another language" (Bers, 2019).This paper has 58 citations.The other two studies investigate the implementation of integrated STEM curricular units by science teachers through a phenomenological multiple case study (Dare et al., 2018) and active learning in flipped life science courses to promote the development of critical thinking skills (Styers et al., 2018).
Table 6 presents the citation counts for scientific publications based on their position within the article.Cohan et al. (2019) classify citation intents into three distinct types: background information, methods utilized, and comparison of results.The citation data was obtained from Semantic Scholar (https://www.semanticscholar.org).
However, it should be noted that the total number of citations for each category may differ from those listed in the TC column.The citation count is limited to articles for which Semantic Scholar can access the full text (Thuan et al., 2022).As such, the total number of citations for the top-10 most frequently cited articles is as follows: background citations accounted for 396 (82.0%) of the total, methods citations for 49 (10.1%), and results citations for 38 (7.9%).
The most influential articles in research citations cover broad issues, not a particular topic.These studies had three authors per paper or more published between 2013-2019.The most influential studies are mainly published in scientific journals.This is consistent with the view that, in the social sciences, articles in scientific journals are often more interesting than conference papers (Fairclough & Thelwall, 2015).This is a suggestion for researchers when searching for materials in the field of STEM education.Eight of the ten first author's countries of the most cited papers are the USA, and the remaining two first authors are from the UK and South Korea.This is consistent with the analysis of the most published countries, where more than half of the publications come from the USA.
The citations of the most influential articles come mainly from the background of the articles (82.0% of the citations by content have been counted).This index is significantly more significant when compared with the number of background citations in the field of measuring efficiency in higher education (64.2%) (Thuan et al., 2022).It is safe to say that the most influential papers in the field help to provide an initial perspective to the reader's understanding of the insights discussed later in the study of this topic.STEM education topic.These articles also help clarify the importance and reinforce the authors' arguments (Thao et al., 2020).

Main Research Topics
Based on the keywords selected by the author in the articles, the author analyses the main research topics of the case studies on STEM education.The total number of keywords set by the authors was 2,034.We combined synonyms, singular, and plural keywords, and removed keywords that did not show research trends before analysis.
Figure 5 displays the co-occurrence network of 69 keywords, each appearing in at least five papers.The nodes in Figure 5 represent individual keywords, with the size of each node proportional to the frequency of the corresponding keyword.The keywords are classified into four distinct clusters, with the most sizable cluster (denoted in red) centering on the examination of case studies in STEM education at the tertiary level.Issues related to gender (with a female focus), professional development, experience learning, diversity (Holgado et al., 2020;Shehab et al., 2012).In addition, there are also studies related to curriculum design, self-efficacy of students (Craig et al., 2022;McFadden & Roehrig, 2017).The red cluster has the highest keyword concentration, consistent with previous research that confirmed that STEM education is most implemented at the university level (Ha et al., 2020;Kundu et al., 2022).The green cluster: research interest in STEM education expands to STEAM (adds an A-art element).Research issues focus on specific teaching methods and techniques such as problem-based learning, project-based leaning, technology-enhanced learning, game-based leaning, and scaffold.In addition, many studies are interested in developing thinking and creativity for students such as computational thinking, design learning, creativity (Clark & Mahboobin, 2018;Han et al., 2015;Hathcock et al., 2015;Lou et al., 2017).The blue cluster including studies interested in STEM education activities in K12 education, specific issues discussed include curriculum, integrated teaching, pedagogical content knowledge, etc. (Mathis et al., 2017;Wang et al., 2020).Yellow cluster: This group does not focus on a particular topic.Nodes that are pretty far apart are not closely linked.This group includes keywords such as science education, middle school, and engineering.
Figure 6 depicts the yearly research trends in the field by identifying relevant keywords.A line represents the timeline of each keyword, and a bubble indicates the keyword's highest appearance frequency in a given year.The size of each bubble is proportional to the number of publications that contain the respective keyword (Thuan et al., 2022).
It is worth noting that we only included keywords that appeared in a minimum of five publications each 9 / 14 year.Furthermore, up to three keywords with the highest frequency were selected to represent the identified trend.Notably, the keywords STEM/STEM education and "case study" were excluded as they did not provide information on research trends.
Although the time span of this collection is from 2006 to 2022, the trend topics appear only for the period from 2016 to 2022.Keywords that represent the last five years are 2018 (stem integration, diversity, and equity), 2019 (active learning, professional development, and gender), 2020 (higher education, engineering, and technology), Thematic maps illustrating the authors' chosen keywords were generated, as depicted in Figure 7.The maps produced facilitate the visualization of four distinct theme typologies.Particularly noteworthy is the upper right quadrant, which presents "motor themes" that exhibit high centrality and density, indicating their significance in the research field.The upper left quadrant, on the other hand, is characterized by high density but low centrality, signifying highly developed and isolated topics.In contrast, emerging or declining topics are situated in the lower left quadrant, while basic and transversal topics are found in the lower right quadrant (Cobo et al., 2011;Pham-Duc et al., 2020, 2021).
The themes in the lower left quadrant exhibit inadequate development, with some being marginally developed.These articles demonstrate low density and concentration, indicating emerging or disappearing topics.Includes two groups of keywords: technical education, original integration, and engineering design; and the group consists of keywords about computational thinking, robotics, and educational robotics.Themes are essential for this field, but are not developed, including keywords: active learning, science education, professional development, and a group of keywords: engineering, technology, and middle school.Themes that are considered to be niche typically possess strong internal ties, but they exhibit insignificant external ties, rendering them of marginal significance for the field.This is true of certain keywords, such as experiential learning, pedagogy, and teachers.The most important topics of research trends corresponding to the motor themes of the specialty include keywords higher education, gender, diversity, collaboration, and keyword group: teacher education, K-12, and pedagogical content knowledge.
These analyses corroborate the findings of the cooccurrence network analysis in Figure 5 and the trend topics depicted in Figure 6.

CONCLUSIONS & FUTURE RESEARCH
This article examines the global research trends concerning the practical implementation of STEM education worldwide by analyzing data from Scopus database.Utilizing bibliometric techniques, the article highlights several key features of the literature on STEM education.The first case studies on STEM education began in 2006, but the scientific results of case studies in STEM education have only been around for the past five years.Case studies of STEM education have continued to increase in the last five years (2019)(2020)(2021)(2022)(2023).This can confirm that case studies of STEM education are still attractive to many scholars around the world.However, there is uneven development among countries around the world.The USA has a more significant number of publications than all other countries (n=433, 57.7% of total publications).The country also excels in total citations with 3397 (72.0% of total citations), equivalent to an average citation count per document of 7.8.However, the research cooperation between the countries revolves around the USA, and the other countries' collaboration is also diverse.
Looking at author contributions by volume, seven of the top ten best-performing authors are from the USA, and the other three are from Ireland and Turkey, respectively.Prominent scholars (top-10) in the field of research have published volumes of 3-10 articles, and collaboration between authors on this topic is neither clear nor compelling.The number of publications by author groups is not large enough to represent research trends.Upon examining the author's contribution based on the number of citations their articles have received, it is evident that authors from the USA still constitute the majority.Specifically, nine of ten primary authors of the most highly cited articles are from the USA.In one article, the other primary author is from South Korea.The most influential articles in the field are primarily cited in the background citations section and help provide an initial perspective for readers to understand the insights discussed later in the study.Research on STEM education topics.These articles also help clarify the importance and reinforce the author's arguments.
The most popular journals publishing case studies in STEM education are of good quality, ranking highly in Scopus and WoS database; out of the top-8 journals, only Education Sciences is in Scopus's Q2 group of journals, which is also the only journal in the ESCI list.The remaining seven journals are all in the journals ranked Q1 according to Scopus's ranking.These are also journals in the ISI category of the WoS core collection.
The reality of implementing STEM education worldwide through analyzing the most critical research topics about case studies in STEM education shows three main research directions.The first research direction focusses on analyzing case studies in STEM education in higher education.Issues related to gender (with a female focus), professional development, experience learning, and diversity.The second research direction interested in STEM education expands to STEAM (adding A-art elements).Research problems focus on specific teaching methods and techniques such as problem-based learning, project-based leaning, technology-enhanced learning, game-based leaning, scaffold.The third research direction is more interested in STEM education activities in K-12-education, specific issues discussed include curriculum, integrated teaching, pedagogical content knowledge.These research directions are still ongoing.continues to be of interest in the last five years when it comes to keywords representing research: 2018 (stem integration, diversity, and equity), 2019 (active learning, professional development, and gender), 2020 (higher education, engineering, and technology), 2021 (science education, middle school, and experiment learning), 2022 (creativity and maker education).
This study presents the world's first bibliographic analysis of case studies focused on STEM education.The research findings provide a comprehensive overview of the actual implementation of STEM education globally, including the development of the research field, primary contributors, and their collaborations.The study also offers valuable quantitative insights for scholars interested in conducting future research in this direction.Given the rapid development of STEM education, researchers frequently report on specific cases, making this analysis particularly valuable for closely monitoring developments in this field.Regular bibliographic analyses should be conducted to stay abreast of the rapidly evolving nature of STEM education research.

Limitations
This study has several limitations previously addressed in the literature (Ha et al., 2020;Huong et al., 2021).First, using only bibliographic data from Scopus database may result in a limited representation of publications in the STEM education field.Second, although we manually filtered irrelevant papers in various Scopus categories, our filtering process may require further improvement, which could lead to omissions.Thirdly, certain crucial pieces of information, such as author names and affiliations, necessitate standardization within Scopus database.As the manual correction is not feasible, this limitation could impact our findings since the accuracy of our analyses relies on the quality of the input information imported from Scopus database.Fourth, some types of analyses, such as gender-based statistics of scholars, could not be performed in this study due to the technical limitations of the Biblioshiny and VOSviewer tools.
In the next three years, the number of articles was approximately 50, with 45 articles in 2015, 53 in 2016, and 47 in 2017.The number of papers published in the last five years has shown a rapid increase, with 60 articles published in 2018, 98 in 2019, 113 in 2020, 135 in 2021, and 141 in 2022 (as shown in Figure 2, bar chart).The annual growth rate recorded from 2006 to 2022 is 30.47%.

Figure 1 .
Figure 1.PRISMA diagram represents narrowing of analysis dataset of case studies in STEM (data collected from Scopus [https://scopus.com/] on 14 December 2022)

Figure 3 .
Figure 3. Transnational cooperation network of 36 partner countries (Source: Authors' own elaboration, using VOSviewer) ; JN: Journal; CO: Conference; NP: Number of publications; TC: Total citations; Q: Quartile; *Data collected from https://scopus.com on 1 January 2023; & ** Data collected from https://mjl.clarivate.com on 1 January 2023 Phuong et al. / Implementation of STEM education 6 / 14 papers), two academics from Ireland (Muntean, C. H., and Muntean, G. M. with five each), and one scholar from Turkey (Altan, E. B. with three papers).Based on citations, the highest is Roehrig, G. H. from University of Minnesota Twin Cities (n=106), this author has an average of 21.2 citations per paper-outperforming the rest, followed by two authors from Purdue University: Guzey, S. S. (n=68) and Moore, T. J. (n=54), one other author with more than 50 citations is Muntean, G. M. from Dublin City University.The remaining authors exhibit a citation number from 14 to 46.The annual publications and citations of the top-10 authors are shown in Figure 4. Two authors, Moore, T. J. and Guzey, S. S., have a similar publication process.The first two papers by these two authors on case studies in STEM education were published in 2015 in Proceedings-Frontiers in Education Conference FIE

Figure 5 .Figure 6 .
Figure 5. Co-occurrence network of 69 author keywords that appear together in at least five papers (Source: Authors' own elaboration, using VOSviewer)

Figure 7 .
Figure 7. Thematic map of research directions related to case studies in STEM education (Source: Authors' own elaboration, using Biblioshiny)

Table 1 .
Main information of the collection Figure 2. General information about publication collection (Source: Authors' own elaboration)

Table 2 .
Information on citation count of the collection

Table 4 .
Top-10 most frequently published sources ranked by number of publications

Table 5 .
Top-10 authors ranked by number of publications produced Note.TC: Total of citations & NP: Number of publications