Citation trends in ophthalmology articles and keywords in mainland China, Hong Kong, and Taiwan since 2013 using temporal bar graphs (TBGs): Bibliometric analysis

Background:

We selected authors from mainland China, Hong Kong, and Taiwan (CHT) to examine citation trends on articles and keywords. The existence of suitable temporal bar graphs (TBGs) for displaying citation trends is unknown. It is necessary to enhance the traditional TBGs to provide readers with more information about the citation trend. The purpose of this study was to propose an advanced TBG that can be applied to understand the most worth-reading articles by ophthalmology authors in the CHT.

Methods:

Using the search engine of the Web of Science core collection, we conducted bibliometric analyses to examine the article citation trends of ophthalmology authors in CHT since 2013. A total of 6695 metadata was collected from articles and review articles. Using radar plots, the Y-index, and the combining the Y-index with the CJAL scores (CJAL) scores, we could determine the dominance of publications by year, region, institute, journal, department, and author. A choropleth map, a dot plot, and a 4-quadrant radar plot were used to visualize the results. A TBG was designed and provided for readers to display citation trends on articles and keywords.

Results:

We found that the majority of publications were published in 2017 (2275), Shanghai city (935), Sun Yat-Sen University (China) (689), the international journal Ophthalmology (1399), the Department of Ophthalmology (3035), and the author Peizeng Yang (Chongqing) (65); the highest CAJL scores were also from Guangdong (2767.22), Sun Yat-Sen University (China) (2147.35), and the Ophthalmology Department (7130.96); the author Peizeng Yang (Chongqing) (170.16) had the highest CAJL; and the enhanced TBG features maximum counts and recent growth trends that are not included in traditional TBGs.

Conclusion:

Using the Y-index and the CJAL score compared with research achievements of ophthalmology authors in CHT, a 4-quadrant radar plot was provided. The enhanced TBGs and the CJAL scores are recommended for future bibliographical studies.

Key points:

• A discussion was held regarding the importance of temporal bar graphs designed in this study.

• Since no appropriate visualizations were used to measure their research achievements (RAs) in mainland China, Hong Kong, and Taiwan, their RAs are rare in the literature, particularly using enhanced temporal bar graphs.

• Combining the Y-index with the CJAL scores on a radar plot is an innovative, modern method for comparing RAs between countries, regions, institutes, departments, and authors that have not yet been demonstrated in the literature.

1. Introduction

Few bibliometric studies have been conducted in the field of ophthalmology in recent years.^[1–3] The h-index is widely used in the scientific community around the world^[4] in bibliometrics. Although the h-index has been widely used as an indicator of individual research achievement (IRA),^[5] 2 major disadvantages have been noted, including all coauthors contributing equally to article bylines,^[6–8] and the integer nature of the h-index making it difficult to identify IRAs among entities.^[9] IRAs must be measured using the fair author-weighted scheme^[6,7,10] (i.e., taking author contributions into account) and modified indicators (e.g., x-index^[11] and Y-index^[12,13]).

1.1. The Y-index and the combining the Y-index with the CJAL scores (CJAL) score used to measure IRAs

The Y-index^[12,13] was proposed for evaluating IRAs using the number of publications in the positions of corresponding and first (co-first) authors (denoted by RP and FP, respectively, with equal credits in article byline). However, previous studies^[12,13] have not illustrated how to draw the radar diagram,^[14,15] particularly based on both quantity^[12,13] and article quality (e.g., CJA score based on article quality, article category, journal quality, and author order^[16]). IRAs should not be measured solely on the basis of publications (e.g., the Y-index = RP + FP as the radius shown in the first quadrant).^[12,13] To complement the Y-index, the CJA score^[16] was applied to examine the minimum requirements for faculty appointment and promotion in Taiwan medical schools.^[16] The drawback of the CJA score is that no such article citations are included. Therefore, we intended to plot the Y-index^[12,13] together with the CJAL score^[15] combined with a 4-quadrant radar plot and the L-index,^[17] and compare IRAs in visualizations for authors, institutes, and regions. Thus far, no such developments, but the one,^[15] involving much information in a view, have been applied to bibliometrics.

1.2. Improved temporal bar graph (TBG) required in bibliometrics

A trend analysis of article citations is also commonly used in bibliographic studies to describe the development of topical entities over time.^[18,19] Thus, a good bibliographic study provides a chronological view of the field (e.g., how the number of studies has evolved, how the topic has evolved, and how the outlet has changed). A TBG,^[17,20–22] for interpreting the evolution of article entities, is one method of interpreting the influence of the most influential entities.

TBG plays an important role in any bibliographic study and should not be stopped at this point. Instead, TBG can be used as a starting point to explore bibliographic data using bibliometric methods, such as predicting citations in future articles. Therefore, a bibliographic study should use more sophisticated, multivariate statistical analyses (e.g., cocitation analysis) to derive results instead of simple article and citation counts in tables, as in those 100 top-cited studies^[23–26] or the one^[27] with 15 tables and 27 figures in an article. An example of an article that follows this approach is those^[28–33] that identify a relevant and interesting “story” supported by their bibliographic data and bibliometric analysis on trend analysis.

TBG^{[17,19–21,34]} traditionally has several limitations that should be addressed, including the lack of information about trend stages (e.g., increasing, ready to rise, decreasing, and slowdown). Due to this, the burst spot and burst strength alone are not sufficient (i.e., performing only the TBG tool of Sci2^[35] or keywords only in CiteSpace^[36]). Therefore, the improved TBG is proposed for this study. This study differs from a previous study^[15] in demonstrating the improved application of TBG to bibliometrics.

1.3. Ophthalmology in mainland China, Hong Kong, and Taiwan

The field of ophthalmology has experienced rapid advancements in artificial intelligence (AI).^[37,38] For example, fundus images have become a research hotspot due to their ease of acquisition and rich biological information.^[39] AI research in health care, specifically machine learning and deep learning, has definite clinical relevance in ophthalmology.^[40,41] The yearly growth of AI in ophthalmology publications has been 18.89% over the last 10 years,^[42] indicating that AI in ophthalmology is a very attractive topic in science. Studies on AI have been conducted in the clinical screening, diagnosis, and prognosis of eye diseases, and the research results have gradually been applied to clinical practice, supported by evidence from publications and contributions from ophthalmology authors. The TBG should be used to examine the trend of AI keywords in ophthalmology.

The results of a survey on the attitudes of medical workers in China toward AI in ophthalmology indicated that medical workers had a higher understanding level of AI in ophthalmology than other professional technicians, making it necessary to promote training in ophthalmic AI, among other professional technicians.^[43] Many medical workers did not have experience in the use of AI in ophthalmology, but the general acceptance of AI in ophthalmology was relatively high. Emerging AI articles in the field of ophthalmology need to be verified.

A number of academic publications on the rise in China have prompted the comparison of the results of over 40 studies conducted in mainland China, Hong Kong, and Taiwan (CHT).^[43] The IRAs of ophthalmology authors in CHT should be compared rather than publications alone, as was done in previous studies.^[44] The research funding opportunities and grants are very different in CHT. The IRAs would be distinctly disparate, and verification is needed.

1.4. Study aims

We proposed 2 hypotheses: the CJAL scores of ophthalmology authors differ based on regions in CHT, and the trend of AI keywords and the worth-reading articles in ophthalmology can be displayed using the TBG. Through the use of visualizations, this study aimed to verify the hypotheses.

2. Methods

2.1. Data sources

Searching the Web of Science (WoS) core collection for terms such as ([CU = China OR CU = Taiwan OR CU = Hong Kong] AND SG = Ophthalmology AND PY>=2013), we excluded letter articles, editorials, and comments. On October 2, 2022, a total of 6695 results were obtained in terms of one of the first and corresponding authors from the CHT. The study data are included at the link.^[45]

As this study did not involve examination or treatment of patients or review of patient records, it was exempt from review and approval by our research ethics committee.

2.2. Four approaches used in this study

2.2.1. Enhanced TBG is required.

CiteSpace (College of Computing and Informatics, Drexel University) ^[36] is commonly used for keyword citations using TBG. The disadvantage of this approach is that readers are provided with less information regarding maximum counts and growth trends on the TBG. Figure 1 shows the comparison of the 2 TBGs (i.e., TBG produced in CiteSpace,^[46] and made in this study). The bubble sizes can provide information regarding the burst strength,^[22] growth trends (e.g., increasing, ready to rise, decreasing, and slowdown^[47]) and maximum counts that should also be displayed on the TBG, as shown in Figure 1.

Figure 1.

Comparison of 2 TBGs in traditional and newly designed modes. TBGs = temporal bar graphs.

2.2.2. Research achievements in comparison using radar plots.

The most productive entities, including regions/areas, institutes, departments, and authors, were highlighted using a 4-quadrant radar plot.^[15] Additionally, the counts and mean citations of journals, publication years, WoS categories, and research areas are also displayed on a radar plot.

IRAs cannot be measured solely on publications (e.g., the Y-index = RP + FP^[12,13]). The CJA score^[16] (i.e., taking into account article quality) is also insufficient due to the lack of article citations involved in IRAs. The CJAL score^[15] was applied to this study via equations 1 to 3.

$${\displaystyle \begin{array}{l}{\displaystyle CJAscore={\displaystyle \underset{i=1}{{\displaystyle \sum ^n}}}{C}_i\times J_i\times A_i,}\end{array}}$$ (1)

$${\displaystyle \begin{array}{l}{\displaystyle CJALscore={\displaystyle \underset{i=1}{{\displaystyle \sum ^n}}}{C}_i\times J_i\times A_i\times L-inde{x}_i,}\end{array}}$$ (2)

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{L}-\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{d}(\mathrm{l}\mathrm{o}\mathrm{g}(\frac{Citation}{A_n\times Age}+1),0),>=1.}\end{array}}$$ (3)

Four factors are considered in the CJAL score for a published article: the category (C; e.g., review, original article, case report, etc.); the journal “quality” (J; e.g., journal impact factor (JIF), or ranking of the journal); the authorship order denoted by A); and the L-index.^[16] The CJAL score is calculated by multiplying each of these 3 aspects. CJAL scores original research articles higher than other types of manuscripts; co-first authors (denoted RP and FP to compute the Y-index RP + FP^[12,13]) score higher than other collaborators; for the journal’s quality assessment, they use the JIF or Scientific Citation Index (SCI)/Social Sciences Citation Index (SSCI) journal rankings for SCI/SSCI-indexed papers.^[16] SCI/SSCI journal rankings are based on JIF in each research domain; therefore, domain-specific journal rankings are usually not significantly different from those based on JIF.^[15,16]

The choropleth map^[48] was applied to present the CJAL scores in CHT. The term “choropleth map” was coined by John Kirtland Wright in 1938.^[49] The most famous example of choropleth map used was the results of the 2000 US presidential election.^[50] Other examples include showing the disparities in health outcomes across areas related to dengue outbreaks,^[51,52] disease hotspots,^[53] and the Global Health Observatory maps on major health topics.^[54]

2.2.3. Citation trends of keywords and articles in ophthalmology.

Citation trends of keywords and articles were displayed on the enhanced TBGs. The trend of AI keywords and the worth-reading articles in ophthalmology were displayed using TBG.

Keywords were extracted from the coword analysis^[55–57] using Pajek.^[58] The chief keywords are representative in its clusters. The top 5 keywords in the top 10 clusters and the AI keyword were displayed on TBG.

Similarly, the top-cited articles with at least 100 citations were selected to compute their growth trend (denoted by the correlation coefficient between citation counts and a series of numbers from 1–4). Only the increasing growth type was demonstrated on TBG. The growth types refer to previous studies.^[15,45,59]

2.2.4. The most worth-reading articles shown on the dot plot.

The most worthy-reading articles were displayed on a dot plot (namely, the impact beam plot^[60]). The method for drawing the dot plot is deposited at the link.^[45]

2.3. Creating dashboards on Google Maps

All figures were drawn by author-made modules in Excel (Microsoft Corp). We created pages of HTML used for Google Maps. All relevant CJAL scores and Y-indices for each member can be linked to dashboards on Google Maps. The TBG can be zoomed in and out with a link to the website. The method of how to draw the TBG in R is deposited with a PDF file and an MP4 video at the link.^[45]

3. Results

3.1. Research achievements in comparison using radar plots

The geographical distribution of CJAL scores in regions is shown on the choropleth map in Figure 2. The top 3 (i.e., Guangdong, Shanghai, and Beijing) are linked by lines. The darker colors (e.g., the other 2 of Zhejiang and Taiwan) indicate higher CJAL scores in comparison to other regions with a lighter color in Figure 2.

Figure 2.

Geographical distribution of CJAL scores in regions (the top 3 are linked by lines, and darker colors indicate higher CJAL scores). CJAL = combining the Y-index with the CJAL scores.

In Figure 3, we can see that the majority of publications were published in Shanghai city (935), Sun Yat-Sen University (China) (689), Department of Ophthalmology (3035), and the author of Peizeng Yang (Chongqing) (65), with the Y-index on coordinates in the 4-quadrant radar plot.

Figure 3.

The radar plot is used to display the Y-index and CJAL scores for the top 10 elements in each entity in the CHT. CHT = mainland China, Hong Kong, or Taiwan, CJAL = combining the Y-index with the CJAL scores.

The highest CAJL scores were also from Guangdong (2767.22), Sun Yat-Sen University (China) (2147.35), and Ophthalmology Department (7130.96); the author Peizeng Yang (Chongqing) (170.16) had the highest CAJL.

The majority of publications were published in the journal of Int. J. Ophthalmol. (1399), 2017 (2275), Ophthalmology (6619) in the WoS category and in the research area (6548), with the counts in Figure 4. In contrast, the highest mean citations were published in the journal Sci (17.13), 2020 (21.88), Surgery (18.69) in the WoS category and neurosciences & neurology in the research area (20.50), as shown in Figure 4.

Figure 4.

The radar plot is used to display the counts and mean citations for the top 10 elements in each entity in the CHT. CHT = mainland China, Hong Kong, or Taiwan.

3.2. Citation trends of keywords and articles in ophthalmology

The top 10 clusters are separated using coword analysis and are shown in the top panel of Figure 5. The chief keywords with the highest centrality degree (denoted by the weighted counts) are representative in each cluster. The keyword artificial intelligence is located in cluster 1 (as labeled by prevalence in the green circle). The top 5 keywords in each cluster were extracted and are shown in the bottom panel of Figure 5. They are combined with the AI keyword to display the citation trends on the TBG in Figure 6.

Figure 5.

Coword analysis of keywords Plus in WoS and 10 clusters found in this study. WoS = Web of Science.

Figure 6.

Citation trends of keywords shown on the enhanced TBG we developed in this study. TBG = temporal bar graph.

The AI keyword is displayed at the bottom of Figure 6, with citations of (0, 0, 1, 0, 2, 0, 1, 1, 1, 3, 1) since 2013. There is a maximum count (=3) with a black dot in 2021. A burst spot with a red dot appears in 2021 and 2022 as a result of the next counts being not smaller than the count at the burst spot. In this case, the red bar would extend to the right from the beginning of the burst. There is a green circle for the AI keyword at the extreme right of the TBG, indicating that the growth type is ready to decline (=−0.16 = correlation coefficient against series of numbers from 1–4 since 2019). A light green bubble indicates the burst strength (=1.18), as shown on the bottom-right corner of Figure 6.

In Figure 6, there are 3 keywords with a burst near 2022, including anterior segment dysgenesis, strabismus, and artificial-intelligence, although their trends belong to other types (i.e., ready to decline, slow down, and ready to decline, respectively). The reason may be that the citations were collected until October of 2022, instead of December.

Figure 7 illustrates 3 worth-reading articles that were selected from 86 articles with citations >100 based on the following 3 criteria: an increasing trend over the last 4 years (denoted by type 1 at the utmost right side of the TBG); a higher burst strength (e.g., >0.20); and a maximum count or burst spot near the recent years (e.g., 2022 or 2021 in this study). Three worth-reading articles^[60–62] were selected and are shown with light-green circles labeled on the right-hand side in Figure 7.

Figure 7.

Citation trends of articles with the noteworthy feature shown on the enhanced TBG we developed in this study. TBG = temporal bar graph.

3.3. The most worth-reading articles shown on the dot plot

According to the dot plot in Figure 8, articles with an increasing citation trend are displayed on the impact beam map (IBP).^[59] The 3 papers with the highest citation counts are located on the right side of Figure 8, while the 3 papers^[61–63] with the most noteworthy features can be found at the bottom of Figure 8. Readers are invited to scan the QR code in Figure 8 and click on the dot of interest to be directed to the PubMed article.

Figure 8.

A total of 35 articles with noteworthy features shown on the dot plot we developed in this study.

3.4. Online dashboards shown on Google Maps

All the QR codes in figures^[64–67] are linked to the dashboards. Readers are suggested to examine the displayed dashboards on Google Maps.

4. Discussion

We found that the majority of publications were published in 2017 (2275), Shanghai city (935), Sun Yat-Sen University (China) (689), the international journal Ophthalmology (1399), the Department of Ophthalmology (3035), and the author Peizeng Yang (Chongqing) (65); the highest CAJL scores were also from Guangdong (2767.22), Sun Yat-Sen University (China) (2147.35), and Ophthalmology Department (7130.96); the author Peizeng Yang (Chongqing) (170.16) had the highest CAJL; and the enhanced TBG featured maximum counts and recent growth trends that were not included in traditional TBGs.

4.1. Additional information

In this study, the number of publications published by ophthalmology authors has a slowdown trend over time (see the bottom of Fig. 7), based on the CHT only.

It is possible to explore many aspects of research in bibliometrics, including citation analysis, theme exploration, topic analysis, reference citations, research, research achievement, author collaboration, and trends and hot spots in terms and articles.^{[18,47,68–71]} This study examined only research achievement (Figs. 2–4) and citation trends (Figs. 6 and 7). There were 3 features that were illustrated in this study: the 4-quadrant radar plot allows condensing the information of bibliographic data (e.g., authors, journals, countries, and institutes) with a glance rather than having many tables and graphs as in many traditional bibliographical studies^[15,27] (e.g., the one^[27] with 15 tables and 27 figures in an article); Y-index and CJAL scores are successfully combined on a radar plot (e.g., Figs. 3 and 4); and the enhanced TBG was applied to citation trends of keywords and articles (Figs. 6 and 7), distinctly different from the TBG solely applied to keywords in Citespace.^[37] As such, the first hypothesis that institute CJA scores differ based on regions in CHT has been confirmed.

The field of ophthalmology has experienced rapid advancements in AI.^[37,38] For example, fundus images have become a research hotspot due to their ease of acquisition and rich biological information.^[39] AI research in health care, specifically machine learning and deep learning, has definite clinical relevance in ophthalmology.^[40,41] The yearly growth of AI in ophthalmology publications has been 18.89% over the last 10 years^[43] and was verified over the last 4 years in this study (at the bottom of Fig. 6). The second hypothesis has also been confirmed: AI is emerging in ophthalmology, particularly in mainland China.

The third hypothesis that the worth-reading articles in ophthalmology can be displayed on the TBG has been confirmed and shown in the next section.

4.2. Three worth-reading articles with a higher growth rate of citations in recent years

A study written by Spaide et al and published in 2020 was the most noteworthy article.^[61] The article is entitled, “Consensus Nomenclature for Reporting Neovascular Age-Related Macular Degeneration Data: Consensus on Neovascular Age-Related Macular Degeneration Nomenclature Study Group.” The framework for a consensus nomenclature system was developed, and the study group suggests that it should be used in future reported studies of neovascular age-related macular degeneration.

The second article^[62] entitled, “Progression of Myopia in School-Aged Children After COVID-19 Home Confinement” was authored by Wang et al and published in 2021. The prevalence of myopia was calculated for each age group and compared between 2020 and the previous 5 years. In 2020, there was a substantial myopic shift in younger children aged 6 to 8 years compared with previous years, but the prevalence of myopia was higher than in previous years for children aged 9 to 13 years. Home confinement during the COVID-19 pandemic appeared to be associated with a significant myopic shift for children aged 6 to 8 years. However, numerous limitations warrant caution in the interpretation of these associations.

The third article^[63] entitled, “Global Prevalence of Diabetic Retinopathy and Projection of Burden through 2045: Systematic Review and Meta-Analysis” was authored by Teo et al and published in 2021 as well. The authors conducted a systematic review and meta-analysis of population-based studies on diabetic retinopathy (DR), vision-threatening DR, and clinically significant macular edema to estimate global and regional prevalence and burden. The number of adults worldwide with DR, vision-threatening DR, and clinically significant macular edema is projected to increase by 2045. Diabetic retinopathy prevalence is highest in Africa and North America.

4.3. Implications and changes

The study has several distinctive features. First, the enhanced TBG was developed and demonstrated, particularly for any time-series data. Thus, the easy-use TBG can be drawn in Rstudio (Boston, MA URL http://www.rstudio.com/) as shown (see Document of how to conduct this study at the link^[45]).

Second, numerous bibliographical studies used visualizations to explore only 1 aspect of research (e.g., citation analysis, theme exploration, topic analysis, reference citations, research, research achievement, author collaboration; and trend and hot spot in terms and articles^{[18,47,68–71]}). It was not found that such research was conducted in visual displays to verify their hypotheses so that research could focus on a specific area rather than providing many tables and graphs (e.g., the one^[27] with 15 tables and 27 figures in an article) to explore the knowledge of interest and noninterest to readers.

Third, combining the Y-index and CJAL scores on a radar plot is an innovative, modern method for comparing IRAs between countries, regions, institutes, departments, authors, publication years, journals, WoS categories, and research areas that have not yet been demonstrated in the literature and deserve to be incorporated into future bibliometric analyses.

Fourth, 4 visual representations are included in the study, including a choropleth map based on CJAL scores, a 4-quadrant radar plot to identify IRA comparisons among entities, a network diagram illustrating the results of coword analysis, and a dot plot to display all articles of interest in a single plot, rather than listing them individually.

Moreover, it would be beneficial if journals (or WoS) provided the practical application of articles worthy of reading on their website in place of the traditional highlighted articles with the highest number of downloads or reads for readers. Several previous studies^[15,47,58] have demonstrated how to calculate hotspots, growth types, and inflection points based on sequential time-series data in recent years. In addition to providing highly cited articles, journals (or WoS) can also test and implement algorithms (MSExcel module at the link,^[45] which include the code produced in R for drawing the TBG) that assist readers in selecting high-impact articles worth reading, rather than simply providing highly cited articles published many years ago.

4.4. Limitations and suggestions

There are a number of issues that need to be addressed in detail in further research. As a first concern, only articles pertaining to the research area of ophthalmology in CHT are included. It is recommended that future studies include a wider range of articles based on co-first authors in ophthalmology using the CJAL score.

The second point is that although the Y-index^[12,13] and the CJAL score^[15] have been considered to be fair measures of IRA contributions, it is assumed that the co-first authors contribute equally to the articles. If authorship does not follow the rule as designed using the author-weighted scheme,^[6,7,10] the results regarding the authors who contributed the most to articles will be biased.

Third, it takes some time to calculate the CJAL score. The advancement in hardware has made this task trivial, similar to computing other bibliometric metrics (e.g., h-/x-index^[5,11]) with a dedicated software program, equally easy and quick.

Fourth, it was proposed in this study to use the CJAL score, including article citations, as is standard in bibliometrics; however, the IRA is determined by many other factors (e.g., when multiplying items in the CJAL^[15]), which need to be taken into account other possible factors (e.g., grant funding) in the future.

Fifth, according to Figure 2, only regions with higher CJAL scores are compared. Readers may also be interested in the list of regions with the Y-index^[12,13] shown on the radar plot. Using the radar plot to display the productive regions should be involved in a future study (see how to draw the radar plot at the link^[45]).

Sixth, it is possible that there are some biases because there are some different authors with the same name or abbreviation who are affiliated with different institutions. Authors are limited to a region in order to prevent identical names across regions, but the results of IRA comparison would be influenced by authors with the same name or abbreviation in the same region.

Finally, although the enhanced TBG is considered useful and applicable, some basic R expertise is required to draw them effectively in R.

5. Conclusion

We compared IRAs in article entities using a radar plot, the Y-index, and the CJAL score. The results of this study confirm the hypothesis that AI is emerging in ophthalmology, and the trend of keywords and the articles that are worth reading in ophthalmology can be displayed on the TBG. This study employed 4 novel visualization techniques. For future bibliographical studies, enhanced TBG and CJAL scores are recommended.

Acknowledgment

We are grateful for the grant provided by An Nan Hospital, China Medical University, Tainan, Taiwan (ANHRF105-07) that enabled us to complete this study.

Author contributions

S-YH and Y-TY developed the study concept and design. TWC, Y-TY and S-CK analyzed and interpreted the data. S-CK monitored the process of this study and helped in responding to the reviewers’ advice and comments. T-WC drafted the manuscript, and all authors provided critical revisions for important intellectual content. The study was supervised by SC.

Investigation: Shu-Chun Kuo.

Methodology: Tsair-Wei Chien.