Bibliometric Analysis of Literature on Acute Respiratory Distress Syndrome Treatments Published Between 2000 and 2019

Acute Respiratory Distress Syndrome (ARDS) is a clinical syndrome characterized by non-cardiogenic pulmonary edema, diffuse endothelial injury, and bilateral pulmonary infiltrates, which can be devastating to critically ill patients in the intensive care unit and places a burden on the healthcare system. The LUNG SAFE study reported that 23.4% of patients receiving mechanical ventilation were ARDS cases and that the hospital mortality rate in patients with severe ARDS was 46.1%.^[1] The first report of the syndrome was published in 1967,^[2] and discussions of how best to define the new clinical entity have been ongoing, from the American-European Consensus Conference definition to the Berlin definition, and the same is true for developments in ARDS treatment. In 1969, a treatment with continuous positive-pressure breathing was proposed,^[3] and Positive End-expiratory Pressure (PEEP) therapy was evaluated in 11 ARDS patients in 1973.^[4] Ventilation with a low tidal volume (6 mL/kg predicted body weight) and a plateau pressure of 30 cm of water or less was shown to improve survival in patients with ARDS,^[5] and these strategies are recommended in clinical practice guidelines. ARDS management strategies are more developed than therapies. Recruitment maneuvers, the technique that sustains high airway pressure for a period of time so as to expand the collapsed alveoli during mechanical ventilation, are still under debate, although they can reverse hypoxemia.^[6] The findings of a recent randomized clinical trial did not support the routine use of recruitment maneuvers and titrated PEEP because of increased 28-day mortality in patients with moderate to severe ARDS.^[7] These examples illustrate the evolution in thinking and clinical practice around the treatment of ARDS, but individual examples cannot provide a comprehensive overview of this evolution. The entire published literature on ARDS, however, contains a detailed history of this evolution in terms of trends, leading journals in the publication of ARDS research, and the investigators and institutions behind this research, along with their interrelationships and influences.

A detailed and comprehensive overview of this nature may provide policy makers and funding agencies with important information to facilitate identification of key research topics, current trends, or priority areas for funding in this field. Although ARDS treatments have been reviewed,^{[8, 9]} only one bibliometric analysis has examined the trends in topics, authors, and research institutions. Wang and colleagues analyzed trends in ARDS research, compared publication years, institutions, journals, and highly cited articles, and divided the articles into clinical and basic research clusters by keywords,^[10] an approach that is not sufficiently detailed and comprehensive. Recognizing the absence of such a published survey, we undertook to perform a bibliographic analysis of papers related to ARDS therapy published over the past 20 years. Bibliometric analysis is a useful method for identifying impactful authors and regions, constructing collaboration networks, and distilling key research topics in particular areas.^[11]

In this study, bibliometric analysis was used to achieve three goals: (1) map the overall layout of ARDS research from the perspective of publication dates, journals, keywords, and citations, (2) identify core research institutions and their direct collaborative networks, and (3) highlight current and potential future areas of research focus.

Literature retrieval was performed in both PubMed and Web of ScienceCoreCollectiononAugust3,2020.Weselected”respiratory distress syndrome, adult” as a subject heading, which was combined in PubMed searches with several subheadings (therapy, diet therapy, drug therapy, nursing, prevention and control, and surgery). We then searched for ARDS treatment-related articles in Web of Science using the same strategy. Both sets of records were merged and duplicates were removed. Articles, reviews, and letters published between 2000 and 2019 (inclusive) were kept for further analysis. Figure 1 shows a flowchart of the literature retrieval strategy. Online Supplement 1 provides more details on the retrieval process. Informed consent was not necessary as our study analyzed published literature without any identifying information about participants in the research reported.

The articles identified in the search were systematically analyzed using the Web of Science website and CiteSpace (5.7.R1, Chaomei Chen).^[12] Web of Science has built-in statistics tools to graph publication trends by year, language, and journal. CiteSpace is a tool for visualizing and analyzing scientific literature by constructing the knowledge architecture by citation relationship between studies in the research field. Among the bibliometric analysis methods, we selected reference co-citation analysis^[13] and analysis of institutional collaboration networks for articles published between 2000 and 2019.^[14] The indicators used in the evaluation were “modularity” and “betweenness centrality,” as follows. The degree to which a network can be divided into several independent blocks is measured by modularity.^[15] Its value ranges from 0 to 1, and a relatively high value means a well-structured network.^[16] Betweenness centrality is a quantitative indicator of the influence of institutions, calculated as the fraction of shortest paths going through a given node that denotes an institution in the cooperative network.^[17] A citation is regarded as a reasoned and solid index of scientific communication. Co-citation analysis enables the identification of the inner structure of research disciplines. A co-citation exists if two references or authors appear in the same bibliography, and is interpreted as the measure for similarity of content of the two references or authors.^[18] In this study, the 50 most-cited articles in successive 3-year intervals (e.g., 2000-2002, 2003-2005) were collected, and all references from each 50-article block were used to create the individual networks. Log-likelihood ratio weighting was used to analyze the contents of each cluster. “Burst keyword detection,” a computing technique used to identify mutations in information, was used to determine new research foci in the field.^[19] A keyword burst is characterized by the intensity of the burst and by its duration.

Figure 1.
Flowchart of literature selection.

A total of 13,933 papers were collected for analysis. Figure 2 graphs the number of articles published by year and indicates a general upward trend in research into ARDS treatments, from 466 articles in 2000 to 918 in 2019.

The articles returned in the search had been published in 15 languages. Of the 13,933 papers, 13,460 (96.61%) were published in English, 221 (1.59%) in German, 121 (0.87%) in French, 89 (0.64%) in Spanish, 15 (0.11%) in Portuguese, and 27 (0.2%) in other languages.

A total of 1,592 scholarly journals published articles on ARDS treatment during the period of interest. The 10 journals that published the most articles (Table 1) accounted for 28.1% of all documents published (3,914 articles). The journals publishing the most ARDS research were Critical Care Medicine, Intensive Care Medicine, American Journal of Respiratory and Critical Care Medicine, and Critical Care.

The researchers who published the literature included in this study were affiliated with a total of 9,468 institutions. Table 2 lists the 15 institutions most associated with publishing ARDS treatment research, which accounted for 28.64% of the total articles. The University of Toronto was associated with the most publications, followed by the University of California, San Francisco, and Harvard University. Extensive collaborations were observed between institutions (Figure 3). The University of Toronto, the University of California, San Francisco, Harvard University, and the University of Washington had high betweenness centrality.

We used all references from the 50 most-cited publications in each successive 3-year grouping to construct each individual network, and then synthesized the individual networks for a total of 243,830 references. We performed reference co-citation analysis to generalize clusters and construct a knowledge map in timeline view, which grouped publications on ARDS treatment into nine major clusters (Figure 4).

The network in this study had a modularity of 0.717, which is typically considered relatively high. We focused on the major clusters that were sorted and tagged by the number of co-cited references. A total of 10 clusters were included in the analysis (Table 3). Three of these clusters (key terms “acute lung injury,” “long-term outcome,” and “extracorporeal membrane oxygenation”) have been active in recent years.

The timeline view of reference Co-citation analysis (Figure 4) shows that the duration of research into extracorporeal membrane oxygenation was the longest, lasting from 2006 to 2018. The use of nitric oxide in treatment was no longer a topic of research papers after 1999. Some references were included in multiple clusters, acting as a link. Table 4 lists the core articles in the main Co-citation clusters.

Figure 2.
Numbers of articles on acute respiratory distress syndrome published over time.

Keyword co-occurrence analysis returned 59 keywords. The keyword citation burst (citation increase over a given period) identified 15 keywords with strong citation bursts over the past decade (Table 5). Extracorporeal membrane oxygenation, listed in both full and abbreviated forms, was an area of active research between 2013 and 2018. Meta-analysis and randomized controlled trials also had high burst strength, especially meta-analysis in recent years. Two active keywords related to basic ARDS research were NF-kappa B, a key nuclear transcription factor regulating inflammatory responses in ARDS,^[20] and oxidative stress. Recruitment, protective ventilation, and noninvasive ventilation were often-used keywords connected to ARDS treatment between 2013 and 2016. Keywords relating to pathogenic factors such as risk, infection, pneumonia, and sepsis are still in use.

Bibliometric analysis is a useful method for identifying impactful institutions, constructing collaboration networks, and identifying research topics from published literature. In addition, bibliometric analysis may assist investigators in predicting research trends in medical science. We have presented a visual analysis of the research literature into the treatment of ARDS, which revealed the following epistemological characteristics. The body of literature has grown over time, with Critical Care Medicine and Intensive Care Medicine publishing multiple papers on the topic. Many of the researchers publishing on the topic were affiliated with the University of Toronto, St. Michael’s Hospital, the University of Washington, and the University of California, San Francisco. Co-citation analysis grouped the references into eight major co-citation clusters indicating research interests within the field of ARDS therapy. The 15 keywords thus identified may be helpful in formulating future research.

The growth in the number of publications indicates that an increasing number of researchers were studying ARDS treatments, frequently in collaboration with researchers in other institutions. Nearly one-third of the literature was published in only 10 journals, indicating that ARDS treatment was a major focus of research published by these journals.

Reference co-citation analysis was used to assess the co-citation relevance between papers and to categorize the data into major clusters representing research interests. In our study, the reference network was grouped into eight clearly defined co-citation clusters. Active research in this field appears to be dominated currently by extracorporeal membrane oxygenation and acute lung injury. Extracorporeal membrane oxygenation, as an approach to cardiopulmonary support, has been shown to improve the survival of patients with severe respiratory failure^[21] and was used in influenza A (H¹N¹)-associated ARDS.^[22] Related topics of research discussion were criteria for extracorporeal membrane oxygenation, extracorporeal carbon dioxide removal, and long-term outcomes of this approach.^{[23, 24]} However, the most recent study (EOLIA trial) in patients with severe ARDS reported no benefit in terms of 60-day mortality rates.^[25] The ventilatory strategies cluster, which contained the most articles, including comprehensive topics involving the open lung,^[26] PEEP level setting^[27] and fluid-management strategy,^[28] short-term prone positioning, and the burden of ARDS,^[29] has been inactive since 2010. The clusters for long-term outcome and acute lung injury have developed and remain active. In particular, the former was in an inherent relationship with the cluster of ventilatory strategies, and we found some overlapping topics in their core literature, such as PEEP^[30] and prone positioning.^[31] Treatment outcome has always been the focus of attention in research into ARDS and, perhaps for this reason, the long-term outcome cluster kept in close contact with the acute lung injury cluster and extracorporeal membrane oxygenation, serving as a bridge between the clusters. The core articles targeting prognosis of ARDS were contained in the aforementioned clusters. From a timeline view, “acute lung injury” was the most recent cluster, and is still active. The themes of core articles included ventilator-induced lung injury,^[32] investigation of ARDS sub-phenotypes,^[33] driving pressure,^[34] the effect of high-flow oxygen on acute hypoxemic respiratory failure,^[35] and the reassessment of ARDS burden.^[1] These topics appeared to be unfocused but involved basic data, issues of wide concern, and investigations in new directions, which were often co-cited in the papers to provide basic support and guidance and probably developed over time.

Figure 3.
Map of institutions active in publishing research on the treatment of acute respiratory distress syndrome between 2000 and 2019. Larger typeface indicates a larger number of articles published by researchers affiliated with the institution. Core institutions are marked with a purple ring and greater width of the ring indicates high centrality. The thickness of the curve in connecting lines represents collaborative intensity between institutions, with thicker lines denoting more intense collaborations.

A heavily cited keyword is considered an indicator of a topic of research interest or of an emerging trend. Extracorporeal Membrane Oxygenation and its abbreviation (ECMO) were first and second in the ranking of citation burst strength in recent years, indicating that this topic will continue to be studied. Meta-analysis can comprehensively assess the results of previous randomized controlled trials on ARDS to arrive at objective conclusions, and it is expected to be the popular research methodology from the strength value obtained. In the field of basic research, oxidative stress has become a popular topic of investigation in ARDS therapy. Table 5 shows the progress in ARDS research through the evolution of burst keywords over the past decade. Factors that can precipitate ARDS and appeared in the literature we analyzed were pneumonia, infection, and severe sepsis, in chronological order. A similar evolution was reflected in the therapeutic methods, from lung recruitment and protective ventilation to noninvasive ventilation to extracorporeal membrane oxygenation. However, ARDS phenotype was not detected by our analysis, perhaps because of the lack of enough supporting literature. This stratifies the heterogeneous syndrome into homogeneous sub-phenotypes with physiology, clinical data, outcomes, biomarkers, or a combination of these.^[36]

The results of the analysis may be applied in a comparatively narrow range. The researchers who conduct studies on ARDS may have grasped some of the contents, which may be more useful for policy makers and funders who seek to understand the current state and trends in the field. To perform the co-citation clustering analysis, we selected Web of Science as our final literature source, which may have excluded some studies. Valuable articles can also be found in evidence-based medicine databases such as the Cochrane Library and citation databases such as Scopus.

Figure 4:
Timeline view of the knowledge map from reference co-citation analysis of research into treatments for acute respiratory distress syndrome published between 2000 and 2019. Ten classified clusters, tagged with characteristic names, are represented by colored line segments and numbered from 0 to 9 (from the largest to smallest number of references). Tag names were assigned by a tag word extraction algorithm (log-likelihood ratio algorithm). The left endpoints of colored horizontal lines indicate the year when co-citation links in the fields appeared for the first time, and their length denotes the duration of publication of relevant research articles. Lines connecting the clusters reflect the co-citation relationship, and often-cited articles are highlighted by dots or rings.

We purposely excluded articles related to coronavirus disease 2019 (COVID-19); COVID-19 pandemic of 2020 will inevitably lead to an increase in ARDS treatment studies, but assessing their contribution to the field of ARDS treatment necessitates a hindsight view. The COVID-19 pandemic initiated multiple studies published in 2020 and 2021 that are related to ARDS and acute lung injury but whose objective was specifically the treatment of coronavirus infection. These investigations were conducted by researchers and clinicians who have not focused on ARDS treatments in the past and whose specific interest when conducting these studies lay in alleviating the symptoms of COVID-19. We therefore judged that including these studies in the body of literature analyzed in the present work would skew the findings away from the true numbers of research institutions and journals that maintain a long-term and consistent focus on ARDS treatment strategies. However, we speculate that the next few years would witness a trend in researching specific aspects of viral-induced ARDS. Research topics that are deemed exciting are often pursued intensively during a given period and result in multiple publications, but the findings may not be clinically valuable with the passage of time and further investigation, weakening the value of the research topic. In addition, the pursuit of these topics typically lasts only as long as research funders are interested in offering funds for such investigations. The present work provides a pre-COVID baseline that could be used as a comparison benchmark in future bibliometric analyses. Finally, we did not include conference abstracts in our search, which may have influenced the weighting of results.

On the basis of the bibliographic analysis of the literature published in the past 20 years, research into ARDS treatment has been continuous. Extracorporeal membrane oxygenation is currently a topic of active research, and it and oxidative stress are likely to remain the focus in the near future. Meta-analysis of original studies will still be a popular research method in this field.

This study was funded by the National Science and Technology Major Project (grant number 2018ZX10101004), The capital health research and development of special (grant number 2020-2-4094), Beijing Natural Science Foundation (grant number 7212130), and Peking University Third Hospital Foundation (grant number Y76478-01).