A snapshot of pneumonia research activity and collaboration patterns (2001–2015): a global bibliometric analysis

For the performance of the study, we opted to identify documents about pneumonia by means of the MeSH thesaurus in the MEDLINE database because this is a detailed instrument for controlled terminology. The thesaurus employs both a human team of specialist indexers to analyze each article and assign medical subject headings to it, plus automated processes to improve indexing; the result is a highly consistent system of classification for research topics [14‐16]. The pneumonia descriptor was introduced in 1963 as a disease of the respiratory tract and the lung, and it was defined as “infection of the lung often accompanied by inflammation” [17]. Synonyms of this descriptor (and therefore also included in search results) are “Lung Inflammation” and “Pulmonary Inflammation”. Additional file 1: Table S1 shows the MeSH tree structure for “Pneumonia”.

The next step was to identify the documents assigned with the MEDLINE descriptor of “pneumonia” indexed in the WoS. This body of research constitutes the population of documents for the present study. Conceived by Eugene Garfield but now maintained by Clarivate Analytics, WoS is the top scientific citation search and analytical information platform worldwide, serving both as a multidisciplinary research tool supporting a variety of scientific tasks and as a dataset for large, data-intensive studies [18].

The use of the WoS databases enables the analysis of all institutional affiliations reported in the documents and the calculation of citation indicators. The WoS brings together the most visible literature at a global level. These qualities justify its choice as the database platform used in this study despite some limitations related to covering non-English biomedical journals [18].

Although initially no limitations were imposed on our search, to calculate the bibliometric indicators we considered only two types of documents, articles and reviews, as these are the primary references for researchers. The study period was limited to 2001–2015, as delays associated with assigning MeSH descriptors to documents mean that information on the most recent articles on pneumonia is not updated. The searches took place on the Clarivate Analytics WoS platform, which includes MEDLINE database, on March 20, 2018.

For each of the retrieved documents, data on the following bibliographic characteristics were extracted: year of publication, journal of publication and WoS subject category, document type, authorship, citations, institutional affiliation(s), and MeSH descriptors.

Data were then standardized: institutional affiliations corresponding to England, Northern Ireland, Scotland and Wales were grouped together under “United Kingdom,” while affiliations in Overseas France, British Overseas Territories, and island dependencies were also assigned to their ruling countries (for example, the documents signed by authors from French Polynesia, Guadeloupe, Martinique, New Caledonia, and Reunion were assigned to France), although regional designations correspond to geographical rather than political criteria. Scientific production from Taiwan, which in WoS is considered independently from the Democratic Republic of China (China) but whose status is disputed at an international level, was analyzed separately.

Countries responsible for publications were categorized according to their World Bank classification by income level: low-income (< USD 1025), lower-middle-income (USD 1026 to USD 4035), upper-middle-income (USD 4036 to USD 12,475), and high-income (≥ USD 12,476) countries. Each of the countries identified was assigned to a macro geographical (continental) region according to the groups established by the World Bank based on geopolitical and economic criteria and reflected in the World Bank Country and Lending Groups (see Additional file 1: Tables S2 and S3) [19].

Two kinds of indicators were obtained:

We calculated the following citation indicators by journal, country, and geographic region:

In order to assess the differences in the distributions of the publications according to the prestige of the journals, we performed a specific analysis of a sub-sample of publications in journals occupying the top 10% in the impact factor ranking in their respective subject categories in the Journal Citation Reports (2015 edition). We analyzed participation in these “prestigious journals” according to geographical location (regions and countries), collaboration level and number of citations.

We calculated the percentage of documents produced in international collaboration and the evolution by quinquennium in order to estimate the scope of cooperative practices at a global level, considering the whole population of documents analyzed (research field) by country and geographic region. To specifically analyze collaboration between countries, collaboration networks were generated for each of the three quinquenniums using Pajek software. To specifically analyze collaboration between countries, collaboration networks were generated for each of the three quinquenniums using Pajek software. The collaboration network is a graphic representation (graph), wherein the nodes represent authors’ countries (as determined from their institutional affiliations) and links between the nodes represent coauthorships between countries, that is, an international collaboration in published research. The more intense the collaboration, the thicker the links between the nodes. The spatial distribution of the nodes responds to the execution of the kamada-kawai algorithm in Pajek, which places the most prominent nodes (those with a greater number of documents and collaboration links) in the center of the map, and the nodes with a smaller number of publications and degree of collaboration towards the periphery.

Based on an analysis of MeSH terms, we identified the main research focus of the studies in the area, generating density maps using the VOSviewer program with a spatial description of the main MeSH terms for each type of pneumonia [22]: (A) “Pneumonia, Aspiration” (B) “Pneumonia, Bacterial,” (C) “Pneumonia, Ventilator-Associated,” (D) “Pneumonia, Viral,” and (E) “Pneumonia, Pneumocystis”). The process of generating and interpreting the maps proceeded as follows:

All data used to perform the study, including the information downloaded from the database as well as that derived from the treatment of the bibliographic entries, are available in the Dataverse Project, an open access public repository [23] (https://​dataverse.​harvard.​edu/​, doi: https://​doi.​org/​10.​7910/​DVN/​02BUNE).

Due to the nature of the study and dataset, it was not necessary to obtain informed consent or approval from an institutional ethics committee.

The search yielded a total of 33,944 documents published between 2001 and 2015 and assigned with the descriptor “Pneumonia” in the MEDLINE database. Of these, 27,017 (79.59%) were indexed in the WoS Core Collection Databases; 20,918 (77.14%) of them were classified as articles and 1776 (6.57%) as reviews. Thus, the population of study documents was a dataset of 22,694 articles and reviews, which we used to calculate the indicators presented below. Letters (N = 2213; 8.19%), editorials (N = 1, 998; 7.39%), news (N = 58; 0.21%), proceedings (N = 17; 0.06%) and other document types (N = 31, 0.11%) were excluded from the analysis.

The number of publications rose from 981 in 2001 to 1977 in 2015.The evolution of scientific production by year was fitted to a linear growth model, showing an R² value of 0.956. Overall, the study period saw a two-fold increase in scientific production (Additional file 1: Figure S1).

The country with the greatest number of documents was the USA (38.49%), followed at some distance by the UK (7.18%), Japan (6.97%), Germany (6.80%) and France (6.73%). Table 1 shows the number of documents and the evolution of scientific production in the 15 most productive countries by quinquennium (see Additional file 1: Table S4 for results on the top 30 countries).

Although the USA ranks first in all periods, its relative contributions have declined, from 41.13% of all documents in 2001–2005 to 36.52% in 2011–2015. On the other hand, China’s emergence is highly notable, with a 1.13% share of total scientific production in the first period (rank = 22), compared to a 8.44% share in the third (rank = 2). South Korea has also seen considerable growth, contributing just 1.30% to total research production in 2001–2005 (rank = 19) but 3.21% in 2011–2015 (rank = 12). Likewise, Taiwan and Brazil have increased their production from 1.17 and 1.35%, respectively, to 3.02 and 3.19%.

Scientific production in different countries and geographic regions, and its evolution by quinquennium, is concentrated in North America and Europe & Central Asia; together these regions are responsible for 82.87% of the papers included in the population of documents. Research in the two regions has decreased the proportion of documents from 2001 to 2005 to 2011–2015 (− 5.46 and − 4.56%). Countries from East Asia & the Pacific and from Latin America & the Caribbean contributed with 20.90 and 4.84% of the documents, respectively. Growth was pronounced in these regions, at 13.18 and 2.52%. Table 2) (see Additional file 1: Figure S2 for a visual representation of density equalizing mapping projections).

Table 3 ranks the production of the top 15 countries, adjusted for demographic and economic indicators (see Additional file 1: Table S5 for results on the top 30 countries). When normalized by population, the most productive countries were Switzerland, the Netherlands, Iceland, and Denmark. Adjusted for the GDP index, the most productive LMICs were the Gambia, Malawi, Uganda, and Guinea Bissau. If we calculate the ratio of pneumonia publications to GNI per capita index, the USA, China, India, Malawi y Brazil were the most productive. Adjusting by R&D expenditure index, the USA ranked first, followed by Spain, the UK, China, and Italy. In relation to the researchers in R&D index, the USA also leads the ranking, followed by India, Uganda, and China. (see Additional file 1: Figure S3 and Figure S4 for a visual representation of density equalizing mapping projections of the number of documents and world development indicators, by GNI per capita index, GDP index and population index plus R&D expenditure index).

The citation analysis by geographical regions reflects the balance in the absolute number of citations received by researchers in North America and Europe, with the rest of the regions trailing considerably. In contrast, North America presents a somewhat higher citation rate (CR) than Europe (35.76 versus 29.20); among the other regions, Africa showed the best performance on this indicator (CR 31.41), with the rest presenting values of 20.07 to 24.00. In consonance with these data, at a country level the HICs (which are mostly in Europe and North America) showed higher CRs than countries in the rest of the income categories. By individual country, articles with author affiliations from the USA were the most cited (N = 316,942), followed by articles from the UK (N = 62,612), France (N = 48,019), Spain (N = 43,459) and Germany (N = 43,434). Regarding the country-specific CR, Vietnam dominated (CR 50.79), followed by the Switzerland (CR 42.94), South Africa (CR 42.85), New Zealand (CR 40.49), Saudi Arabia (CR 38.62) and the UK (CR 38.44). The USA and the UK were the top-ranked countries with an h-Index of 197 (USA) and 106 (UK), followed by France (96), Spain (94) and Germany (94) (Table 4) (see Additional file 1: Table S6 for the 30 most productive countries).

Overall, 18.80% of the articles published in the study period were written in international collaboration, although the rates increased from 14.35% in the 2001–2005 quinquennium to 21.64% in 2011–2015. Among the top 15 most productive countries, international collaboration was much more intense in the European countries, Brazil, Canada, and Australia (34 to 62%) compared to the USA (26.33%) and the most productive countries of East Asia & Pacific (China, South Korea, and Taiwan: 16 to 28%) (Table 5). The very high levels of international collaboration are even more pronounced in some Latin American, South Asia and particularly African countries. Indeed, the analysis of collaboration by geographical regions shows that globally, sub-Saharan Africa collaborated on 46.08% of the papers produced. The results for Latin America and the Caribbean (22.66%) are heavily weighted by research from Brazil, but the rates of international collaboration were 63.01% in Colombia, 60.94% in Argentina, and 52.21% in Mexico, while in East Asia & Pacific and South Asia (and looking beyond the most productive countries like China), countries like Bangladesh showed levels of international collaboration of 73.61%; Thailand, 60.29%; and Pakistan, 58.82%.

Figure 1 shows the collaboration networks between different countries by quinquennium. The most prominent countries in all time periods, occupying central positions in the networks with multiple cooperative links, are the USA, Canada, the UK, Germany, France, and the Netherlands. The presence of South American and African countries is scarce in all periods. Only South Africa has a notable presence in the third quinquennium (Fig. 1a). A few other countries also “emerge” with a high degree of collaborative links in the second period, like Spain, Greece, Italy, Australia, China, and Japan, although the latter two countries are not fully integrated in global networks, showing collaborative ties only with the USA (Fig. 1b). Finally, other European countries, while present throughout all three periods, stand out to a greater degree in the third period. This is the case of Sweden, Switzerland, Belgium, and Austria. At the same time, China and Japan seem more implicated in the network in this third period, while India and South Korea also gain relevance (Fig. 1c).

The documents we analyzed were published in 2115 scientific journals. Twelve journals accounted for 16.63% of the pneumonia literature Table 6 . shows a list of the 15 top journals with the highest number of papers published from 2001 to 2015, as well as their impact factors for the year 2015, subject category according to the Journal Citation Reports classification, and CR (Additional file 1: Table S7 for results on the top 30 journals). The journals publishing the most articles on pneumonia were PLOS ONE (N = 494), Clinical Infectious Diseases (N = 412), and Chest (N = 397), whereas the journals with the most citations were Clinical Infectious Diseases, (N = 26,351), American Journal of Respiratory and Critical Care (N = 22,647), and Chest (N = 22,212); all of these were also among the 15 most productive journals. The journals with the highest CRs were the New England Journal of Medicine (75 documents, CR 278.13), The Lancet (54 documents, CR 210.17) and JAMA (49 documents, CR = 199.71) (see Additional file 1: Table S8 for results on the top 30 journals with highest absolute and relative citations).

The comparative analysis of the scientific production and CRs of different journals is noteworthy in that some journals (such as the American Journal of Respiratory and Critical Care, Critical Care Medicine, and Intensive Care Medicine) present a very high CR in relation to their total scientific production (Additional file 1: Figure S5 for the top 15 journals producing the most research on pneumonia, plus citation rates).

With regard to the subject categories to which the journals are assigned, the most prominent are “Infectious Diseases” (17.57% of the documents), “Respiratory System” (15.77%), “Immunology” (14.08%), “Microbiology” (11.85%), and “Critical Care Medicine” (9.26%) Table 7. Many of the most productive journals in pneumonia also fall into these subject categories. Moreover, over the course of the three study periods, nearly all of the subject categories saw a moderate decrease in their relative contribution, as research articles became more dispersed and made headway into different disciplines producing less research on pneumonia Table 7.

The analysis of the 4100 documents published in the top 10% of prestigious journals shows a higher participation from the USA (27.66%, compared to 38.49% in the overall body of documents) and from some other European countries like the UK or Spain. In contrast, the weight of Asian countries, particularly Japan and China, is much lower (Table 8). Overall, international collaboration in these journals (N = 1065, 25.98%) was sensibly higher than in the overall body of documents (18.8%), and the greater degree of collaboration was much more pronounced for countries like Brazil, Japan, China, and even European countries like Italy and Germany (Table 8).

The high degree of collaboration was also confirmed between regions in the publications appearing in these journals (Table 9). With regard to the degree of citation, we observed notable increases in the citation rate of the USA and the European countries; these were even more significant for countries in the Middle East & North Africa, and for sub-Saharan Africa when they participated in these journals (Table 9).

With regard to types of pneumonia studied, the MeSH terms to appear most frequently were “Pneumonia, Bacterial” (19.99%), followed by “Pneumonia, Pneumococcal” (7.02%), and “Pneumonia, Ventilator-Associated” (6.79%). Table 10 shows the number of documents assigned to each term describing the different types of pneumonia (Additional file 1: Table S10 for the 30 top general MeSH).

Table 11 ranks the top 15 countries in crude numbers of retrieved articles, stratified by types of pneumonia (Additional file 1: Table S11 for information on the 30 most productive countries). For “Pneumonia, Aspiration”, the main countries were the USA, Japan, and Germany; for “Pneumonia, Bacterial”, the USA, France, and Spain; for “Pneumonia, Pneumocystis”, the USA, France, and the UK; for “Pneumonia, Ventilator-Associated”, the USA, France, and Spain; and for “Pneumonia, Viral”, the USA, China, and Japan.

Table 12 shows the relationship between MeSH terms referring to age groups with those corresponding to different types of pneumonia. The closest associations for “Aged, 80 and over” and “Aged” were with “Pneumonia, Aspiration” (22.58 and 40.56%, respectively), while “Pneumonia, Viral” was the most frequent topic for studies in pre-adults (“Infant”, “Child”, “Child, preschool” and “Adolescent”). The one exception to this was “Infant, newborn”, where the highest proportion of articles was about “Pneumonia, Pneumocystis.” In “Adult” and “Middle aged” people, studies most frequently focused on “Pneumonia, Bacterial” and “Pneumonia, Ventilator-Associated.”

Figure 2 shows the subject area maps with the main MeSH terms in the documents on (a) “Pneumonia, Aspiration”; (b) “Pneumonia, Bacterial”; (c) “Pneumonia, Ventilator-Associated”; (d) “Pneumonia, Viral”; and (e) “Pneumonia, Pneumocystis.” The principal MeSH term related to “Pneumonia, Aspiration” is “Deglutition Disorder”, but research is linked to a broad array of topics, including epidemiological aspects (“Incidence”, “Risk Factor”, “Retrospective Studies”), treatment approaches in intensive care, and surgical techniques procedures facilitating breathing, swallowing, and feeding (Fig. 2a).

The two main MeSH terms that appear most frequently with “Pneumonia, Bacterial” are “Community-acquired Infections” and “Anti-bacterial Agents”, reflecting the central focus that research has taken to identify risk factors and test different therapeutic approaches. MeSH terms related to specific bacteria and infections, such as Streptococcus, Chlamydia, Acinetobacter, and Haemophilus influenzae, are also prominent (Fig. 2b).

For its part, research on “Pneumonia, Ventilator-associated” seems more disperse, although three areas of interest can clearly be differentiated: (a) epidemiological studies, clinical protocols, and treatment in intensive care units (the term “Intensive Care Unit” is the most prominent in this area); (b) treatment outcomes (“Treatment outcome” and “Anti-Bacterial Agents”); and (c) cross infections (“Cross infection”) (Fig. 2c).

Research on “Pneumonia, Viral” also shows a disperse nature, with different areas of interest. Epidemiological aspects are covered under terms such as “Community-acquired Infections” and “Hospitalization”, while at a researcher level, interests reside in the virus “Influenza, Human” and “Orthomyxoviridae Infections” (Fig. 2d). With regard to “Pneumonia, Pneumocystis”, one prominent subject focus is on “AIDS-Related Opportunistic Infections” and another is on “Pneumocystis jirovecii” (Fig. 2e).