June 6, 2023

The impact of COVID-19 on online medical education: a knowledge graph analysis based on co-term analysis | BMC Medical Education

Time distribution of the literature

The number of papers and the trend of changes can measure the level and dynamics of scientific research results in the field, which is very important for predicting future development trends. As shown in the Fig. 1, within the field of medical education, the number of online education-related papers published surged between 2019–2020, and since the literature was retrieved in October 2022 and some research papers in 2022 have not yet been published, we can reasonably speculate that the number of publications in 2022 should be significantly higher than that in 2021. Although this is influenced to some extent by information technology advances such as 5G, the role of COVID-19 in this should not be overlooked. We believe that the COVID-19 pandemic has led to a huge impact on offline teaching due to blockades of varying degrees around the world, and the huge practical demand for technological advances in online education cannot be ignored. It is foreseeable that research on online medical education will continue to increase in the future.

Fig.1

Chronological distribution of medical online education literature before and after the outbreak

Visualization of important keyword clusters analysis

An analysis of keywords in the literature for the decade prior to the epidemic (January 2010 to December 2019) and the post-epidemic period (January 2020 to June 2022) led to the following conclusions.

Pre-epidemic keyword cluster analysis

Scholars around the world before and after the epidemic maintained a certain level of attention to online teaching of clinical research, but after the epidemic, with the rapid development of Internet technology and the trend of the current social situation, online teaching that can communicate remotely has received extensive attention from schools [13, 14].

A total of 10,942 keywords from the decade before the epidemic constructed a keyword network for clinical research teaching research in the decade before the epidemic. Based on the keywords in the 21 main thematic clusters, keywords with a frequency of ≥ 3 were filtered and keywords with “covid-19”, “education” and their related synonyms, i.e. the terms used in the literature search for this study, were removed. We obtained 1822 keywords that were the most prominent, dominant and relevant. The visualization of the keywords drawn using VOSviewer is shown in Fig. 2.

Fig. 2
figure 2

Visual clustering analysis of popular research keywords before the outbreak

Based on the “total link strength” indicator, it is easy to see that the following ten keywords are among the keywords with a strong link strength, which are not listed in this paper due to space, but are shown in Table 1. The term “medical student” was the most relevant term in the online teaching field during the new epidemic, with a total link strength of 2420. In addition, the keywords “care”, “students” and “impact” all ranked in the top two, three and four in terms of frequency of occurrence with an association intensity of more than 1900, respectively. The phrase “curriculum” is the most frequently used term in the field of teaching and learning. The phrase “curriculum” was the fifth most frequently associated term with a strength of 1634.In terms of subject clusters, the subject clusters are ranked according to the strength of relevance and frequency of occurrence of the keywords they contain, with the first largest cluster containing 235 items, shown in red. The most frequent phrase in this cluster is “medical students”, with 338 occurrences and 2420 links.

Table 1 10 keywords based on total link strength before the epidemic

The second largest cluster is shown in blue in the graph and contains 157 items. The most frequent phrase in this cluster is “care”, with 321 occurrences and 2257 links, making it the most frequent phrase in this cluster.

The third largest cluster contains 132 items and is shown in pink in the graph. The most frequent phrase in this cluster is “impact”, with 248 occurrences and 1835 links.

The fourth largest cluster contains 117 items, highlighted in brown in the graph. Knowledge” is the most frequent phrase in this cluster, with 213 occurrences and 1578 links.

The fifth largest cluster, containing 109 items, is highlighted in light green in the graph. The most frequent phrase in this cluster is “attitudes”, with 209 occurrences and 1522 links.

Post-epidemic keyword cluster analysis

A total of 6283 keywords were used to construct a keyword network for clinical research and teaching research during the novel coronavirus outbreak. Based on the keywords in the 14 main thematic clusters, keywords with a frequency of ≥ 3 were filtered and keywords with “covid-19”, “education” and their related synonyms, i.e. the terms used in the literature search for this study, were removed. We obtained 1004 of the most prominent, dominant and relevant keywords. The visualization of the keywords drawn using VOSviewer is shown in Fig. 3.

Fig.3
figure 3

Visual clustering analysis of popular research keywords after the outbreak

Based on the “total link strength” indicator, it is easy to see that the following ten keywords are among the keywords with strong linkage, which are not listed in this paper due to space, but are shown in (Table 2). The term “Anxiety” was the most relevant term in the online teaching field during the new epidemic with a total link strength of 2138. In addition, the terms “Depression”, “Impact” and “Mental health” all ranked as the most relevant terms with a total association intensity of over 1500. The keywords “Depression”, “Impact” and “Mental health” are in the top two, three and four places respectively in terms of frequency of occurrence. The phrase “Stress” is in fifth place with an association strength of 1460.

Table 2 10 keywords based on total link strength after the epidemic

In terms of subject clusters, the subject clusters were ranked according to the strength of relevance and frequency of the keywords included. Anxiety” is the most frequent phrase in this cluster, with 252 occurrences and 2138 links.

The second largest cluster is shown in red in the graph and contains 282 items. The most frequent phrase in this cluster is “Impact”, which has 224 occurrences and 1685 links.

The third largest cluster contains 57 items and is shown in brown in the graph. Pandemic” is the most frequent phrase in this cluster, with 216 occurrences and 1441 links.

The fourth largest cluster contains 141 items, highlighted in green in the graph. Health” is the most frequent phrase in this cluster, with 175 occurrences and 1210 links.

The fifth largest cluster, containing 93 items, is shown in blue. Knowledge” is the most frequent phrase in this cluster, with 140 occurrences and 861 links.

Comparative analysis of keyword clusters before and after the outbreak

By analyzing the top ten keywords before and after the epidemic, it is easy to find that “Students”, “Impact” and “Health” are the main keywords before and after the epidemic. However, the ranking of “Knowledge” has dropped and the position of “Impact” has increased. After the epidemic, keywords such as “Anxiety” and “Stress” gradually emerged to describe psychological stress states, and the focus on students’ health was gradually refined to include mental health.

Visualization of important authors analysis

In the author coupling analysis before and after the epidemic using VOSviewer, the threshold was set to ≥ 7. 17 authors were obtained from 16,989 authors screened before the epidemic, and 18 authors were obtained from 15,676 authors screened after the epidemic. The deeper the curve in the figure indicates the stronger the association. From Figs. 4 and 5, we can see that although the final screening results are similar, some authors are not even shown due to clustering, indicating that some authors are not strongly associated. In general, the authors posting after the epidemic are more closely connected, and for the same topic medicine and online education, the coupling network graph of that author changes greatly after adding the topic of the epidemic, indicating the impact of the change of the research topic on the authors in the related fields.

Fig.4
figure 4

Visual clustering analysis of author coupling network before the outbreak

Fig.5
figure 5

Visual clustering analysis of author coupling network after the outbreak

Then using Citespace to summarize the top 10 authors of the cited articles published, as shown in the figure, before the epidemic, Brent Thoma’s article was the most cited with 18 times; after the epidemic, Chungying Lin’s article was the most cited with 11 times, so for a quick understanding of the field, the articles of these two are more readable (Tables 3 and 4).

Table 3 Top 10 citations by authors of articles published before the outbreak
Table 4 Top 10 citations by authors of articles published after the outbreak

In terms of author centrality, pre-epidemic A Bullock, E Barnes, A Kavadella and A Liepa ranked first with a centrality of 17, while post-epidemic Aimen Khacharem and Achim Jerg ranked first with a centrality of 54, indicating the strong influence of these authors in the field (Tables 5 and 6).

Table 5 Top 10 posting author centrality before the outbreak
Table 6 Top 10 author centrality after the outbreak

Visualising the country density map

In this country density map showing each country’s involvement in online teaching and learning during the novel coronavirus outbreak. Each point in the item density visualisation has a colour that indicates the density of the item at that point. The colours range from blue to green to yellow [15]. The greater the number of items near a point visualized in this density view, the higher the weight of neighboring items, and the closer the color of the point is to yellow. Conversely, the lower the number of items near a point, the lower the weight of neighboring items, and the closer the color of the point is to blue. From this, the most important countries in terms of online teaching and learning research engagement during the novel coronavirus outbreak period can be observed concisely and directly by plotting the visualized country densities.

Prior to the COVID-19 outbreak, most countries had not conducted extensive research in this area (Fig. 6). Among the major countries conducting research, the United States is the country with the greatest involvement and research impact in the field of medical online education, and researchers from the United States have closer ties with Ireland and England, China and Australia, and Canada and Latvia, but most countries are still conducting research independently without forming. The research is still conducted independently in most countries without a cooperative system. After the outbreak of COVID-19, more countries have conducted research on online medical education, but the United States is still the most influential country and has developed close ties with China and England, and countries such as Spain, Germany, Canada, Brazil, and France have also conducted a lot of research in medical online education (Fig. 7).

Fig. 6
figure 6

Visual clustering analysis of country density maps before the outbreak

Fig. 7
figure 7

Visual clustering analysis of country density maps after the outbreak

Collaborative network of organizations that visualize co-authors

The visualization of the institutional collaboration network (nodes are the names of institutions that connect institutions that have collaborative relationships) provides a clear visual representation of the collaborative relationships and key institutions. (Figs. 8 and 9). The size of the nodes provides a visual representation of the centrality of each institution and the density of connections is also directly related to the closeness of the cooperation relationship.

Fig. 8
figure 8

Visual clustering analysis of the institutions of co-authors before the outbreak (2) Post-epidemic organizations cluster analysis

Fig. 9
figure 9

Visual clustering analysis of the institutions of co-authors after the outbreak

Pre-epidemic organizations cluster analysis

Based on the centrality index, it is easy to conclude that “University of Toronto” is in the first place with a centrality of 2450. “Harvard University” is the second most centralised institution in the online inter-institutional collaboration network. The No. 3 institution for centrality is “University of California, San Francisco”.

According to the “Ducuments” metric, the highest-ranking institution in the visual agency collaboration network is “University of Toronto” with 115. “University of California, San Francisco” ranked second with 84, followed by “McMaster University” with 38. It is not difficult to find that the average number of “ducuments” of the top ten comprehensive literatures is about 67.

“Total link strength” refers to the Total co-occurrence times of keywords and other keywords (including repeated co-occurrence times). In the organization cooperation network of the visual co-author, it can indicate the close cooperation degree between the organization and different institutions. By analyzing the data, it’s easy to know that “University of Toronto” ranks first with 20,123, “McMaster University” came in second with 17,301, and “Johns Hopkins University” came in third with 11,631. Not surprisingly, the top three are all over 10,000.

Based on the centrality index, it is easy to conclude that “National University of Singapore” is in the first place with a centrality of 3324. “Harvard Medical School” is the second most centralised institution in the online inter-institutional collaboration network. The No. 3 institution for centrality is “Huazhong University of Science and Technology”.

According to the “Ducuments” metric, the highest-ranking institution in the visual agency collaboration network is “Harvard Medical School” with 47. “

University of Toronto “ranked second with 39, followed by “The University of Hong Kong” with 38. It is not difficult to find that the average number of “ducuments” of the top ten comprehensive literatures is about 34.

“Total link strength” refers to the Total co-occurrence times of keywords and other keywords (including repeated co-occurrence times). In the organization cooperation network of the visual co-author, it can indicate the close cooperation degree between the organization and different institutions. By analyzing the data, it’s easy to know that “Harvard Medical School” ranks first with 10,080. “Johns Hopkins University” came in second with 8431, and “University of Toronto” came in third with 8062.

However, considering the three indicators of “Ducuments”, “Citations” and “Total link strength”, it is easy to find that the development of “Harvard Medical School” in terms of publications, centrality, and total link strength with other institutions is more balanced and better. In addition, it is easy to find that the institution “Harvard Medical School” has direct links with many other institutions, which further reflects that this is a very influential institution and it maintains close cooperation with many other institutions. We realize that the outbreak of COVID-19 has produced a dramatic change in the research community in terms of online medical education. Taking this change into account in a timely manner when selecting a partner institution can go a long way in helping researchers find the right partner institution.

Visualization of collaborative journal clustering networks

The clustering network drawn by the journals to which the visual references in this study belong (the nodes are the names of the journals, and the associated journals are connected by curves), and the clustering analysis is performed according to the main research keywords of the journals. By looking at the graphs and analyzing the data, you can explore the interconnections between journals. Specifically, the following conclusions are obtained:

The journals included in the study before the epidemic were divided into 21 clusters through cluster analysis. (Fig. 10) The journal clustering network contains a total of 770 nodes and 2647 lines. There are about 20 journals with citations above 238. By sorting the number of journal citations, the following graph (Table 7) can be obtained. This study found that the journal ACAD MED ranked first with 1326 citations, JAMA-J AM MED ASSOC was a little behind in second place with 1119 citations, and MED TEACH was third with 1080 applications. In addition, by observing the data, it can be seen that the top three citations have more than 1000 citations. In addition, MED EDUC and BMC MED EDUC are ranked fourth and fifth, respectively, with very excellent journal citations.

Fig. 10
figure 10

Visualizing the clustering network of journals before the epidemic

Table 7 Top 10 journals in citation counts in the clinical learning before the epidemic

The journals included in the study after the epidemic were divided into 19 clusters through cluster analysis. (Fig. 11) By ranking the number of journal citations, the following graph can be obtained (Table 8). The journal THE LANCET was found to be in the first place with 784 citations, Public Library of Science (PLOS ONE) was slightly behind the first place in the second place with 782 citations, International Journal of Environmental Research and Public Health (INT J ENV RES PUB HE) was in the third place with 747 applications. In addition, by looking at the data it can be seen that the top three citation numbers are above 700. However, a large drop in the number of citations occurs from the fourth position.

Fig. 11
figure 11

Visualizing the clustering network of journals after the epidemic

Table 8 Top 10 journals in citation counts in the clinical learning after the epidemic

A comprehensive comparison between before and after the epidemic showed that 60% of the journals remained in the top 10 citation numbers after the epidemic. The analysis included “JAMA-J AM MED ASSOC” “BMC MED EDUC” “NEW ENGL J MED” “BMJ-BRIT MED J” “J MED INTERNET RES” “LANCET”.