INTRODUCTION
Spanish scientific production in cardiovascular medicine has grown substantially in recent years. For the period 1994-1999, we identified 840 documents authored by Spanish researchers in the CD-ROM databases of the Institute for Scientific Information. Annual production rose by 80.9% between 1994 and 1999,1 whereas the number of documents recorded in the National Science Indicators and National Citation Reports for 1994-2002 rose to 2556, with a constant increase in papers per year, growing from 6616 in 1994 to 9143 in 2002.2 The Spanish journals that included greater numbers of articles in cardiovascular medicine were Revista Española de Cardiología (n=899) and Medicina Clínica (n=140)2. They were the two Spanish-language clinical journals that obtained the highest impact factors (IF) in the Journal Citation Reports of 2004 (IF=1.802 and IF=1.005, respectively) and first and third highest in 2005 (IF=1.769 and IF=1.074, respectively).3 They also occupied first and second place in the Potential Impact Factor of Spanish Medical Journals study for 2003 (IF=1.501 and IF=1.061, respectively).4
In cardiology and biomedical research in general, collaborative production of papers has increased progressively. Collaboration is necessary to advance knowledge because research problems require multidisciplinary approaches5 and because collaborative endeavors compensate for local weaknesses.6 In cardiology research, professionals collaborate fundamentally in cardiology, family medicine and internal medicine, but biologists, chemists, pharmacists, mathematicians, and radiologists are also involved.6,7
In this context, the Spanish government's national plan for scientific research, development and technological innovation (Plan Nacional de I+D+I) for 2004-20078 attempts to promote multidisciplinary research. In biomedicine, the call for thematic priorities of cooperative research (RETIC) by Madrid's Instituto de Salud Carlos III9 aims to contribute to constructing a scientific base for the programs and policies of the Spanish national health service in the areas prioritized in the aforementioned national plan through linking up multidisciplinary and multiinstitutional centers and biomedical research groups. It is hoped networks of this type will create more potent schemes of scientific cooperation that facilitate the achievement of objectives it would be difficult to undertake in more restricted contexts.10 At the time of writing, 3 networks exist in cardiovascular medicine9: "Risk factors, course and treatment of cardiovascular diseases and their molecular, and cellular mechanisms" (RECAVA) coordinated by Soler Soler at the Hospital Vall d'Hebron (Barcelona); "Genetic and environmental determinants of vascular dysfunction in hypertension, and ischemic heart disease" (HERACLES) coordinated by Marrugat de la Iglesia at the Institut Municipal d'Investigació Mèdica (IMIM) (Barcelona); and "Epidemiologic, pathophysiologic, clinical, and pathologic characteristics of sudden death in Spain" (EULALIA-MUSIC2) coordinated by Bayés de Luna at the Hospital de la Santa Creu i Sant Pau (Barcelona). Similarly, the Instituto de Salud Carlos III has been promoting Networking Biomedical Research Centers (CIBER)11 independent research organizations dedicated to research into a single specific disease that represents a special healthcare problem, and these constitute major centers for the transfer of research.
The study of scientific collaboration helps establish groups and work networks that can be analyzed and evaluated thru bibliometric techniques and represented in what some authors call "coauthorship networks"12,13 or "bibliometric maps."14
The objective of the present paper is to identify and represent graphically the networks of collaboration between authors and institutions that have published scientific papers in Revista Española de Cardiología between 2000 and 2005.
METHODS
Identification of Papers, Bibliographic Data and Normalization of Authors and Institutions
To undertake this study we identified research papers published in Revista Española de Cardiología during 2000-2005. For each paper selected we identified the name and surnames of the authors, as well as their institutional affiliation (institution, city, and country).
To normalize authorship we checked the signatures with which an individual appeared in two or more different forms, using coincidence in authors' places of work as the basic criterion of normalization. In the case of institutions, we have unified the different variants of hospital names to match the name recorded in the Spanish National Catalog of Hospitals 200515 and the National Registry of Universities, Centers, and Teaching.16 Similarly, given that institutional names in many bibliographic records included two or more institutions (eg, university hospitals and universities) and in order not to lose information, we have proceeded to distinguish between these signatures by, in these instances, recording as many signatures as individual macroinstitutions could be identified for each bibliographic record.
With this information we have constructed a Microsoft Access database.
Bibliometric Indicators of Collaboration and Networks of Coauthorship and of Institutional Collaboration
Throughout the present paper we use the term "coauthorship" to refer to joint authorship of a scientific paper by 2 individuals, and "institutional collaboration" to refer to joint authorship by different institutions. Furthermore, we distinguish between the following types of collaboration: intrainstitutional (between different units or departments of a single macroinstitution), national-interinstitutional (between institutions of the same or different autonomous regions), and international-interinstitucional, when a Spanish institution has collaborated with a non-Spanish institution. We use the term "cluster" to refer to constellations of nodes or vertices (authors or institutions) with close connecting links (relationships of coauthorship or institutional collaboration) but with sporadic connections moving away from the center.17 "Threshold" or "intensity of collaboration" refers to the figure used to form clusters of authors and institutions, referring to frequency of coauthorship between pairs of authors or of collaboration between institutions, and reflects the more or less well-established relationships between them when it comes to jointly publishing the results of their research. The threshold has been used in different bibliometric studies as a criterion to label identifiable clusters as "research groups."18,19
Collaboration between authors is portrayed by calculating the number of papers, signatures, collaborations, the index of signatures/paper (ISP), or collaboration index, which is the mean number of signatures/paper; and the index of authors/paper (mean number of authors/paper considering only the different authors). Similarly, we conducted a descriptive statistical analysis of the variables studied (mean and 95% confidence interval [CI]). To compare the means of the variables in a normal distribution we used ANOVA analysis of variance, determining the equality of variances with the Levene test. These comparisons were made using the Bonferroni test for equal variances and the Tamhane T2 test for unequal variances.
To construct coauthorship networks, we identified all combinations of pairs of authors for each paper. The number of coauthorships of each paper is related to the number of authors as it is equal to m!/(m-n)!n!, where "m" is the number of individual authors and "n" the number of elements in the groups constructed, in this case 2 as we identified pairs of authors who coauthored a paper. Once coauthorship was quantified, we established a threshold of 6 o more collaborations between pairs of authors to reduce the number of nodes and links that would prevent a clear view of the network and, thus, center analysis on the more intense coauthorship relationships. The same approach was applied to institutional authorship to construct the network of interinstitutional collaboration, although in this case we applied a threshold of 3 or more collaborations.
We obtained a series of measures to permit analysis of structure or social networks. If we take agents (authors or institutions) individually, we present 3 measures of centrality or cohesion that facilitate detailed analysis of the social network studied: degree, and indices of betweenness and closeness.20
Degree indicates the number of different agents, a specific author or institution is directly connected with, and is obtained by identifying and subsequently quantifying relationships of coauthorship and institutional collaboration. It's a measure that reflects the greater or lesser extent of collaboration mantained by authors and institutions. Betweenness determines the extent to which an agent is situated in the middle of or between other agents in the network, permitting us to make interconnections. Betweenness measures the prestige of authors and institutions and their capacity to access and control information flow. It is calculated as the sum of the shortest paths between the 2 agents that include between them the agent in question. Closeness enables us to determine the pace of interaction between an individual agent and the other agents in the network. It reflects the "proximity" of each author or institution to the rest. It is calculated as the inverse of the sum of distances from the agent in question to the rest of the agents they are connected with.20
To determine the cohesion of the clusters identified, we calculated the density of each one, a measure that determines the degree of connection between members according to the relationships between the number of existing links and the number of possible links. To do so, we applied the formula 2e/n(n-1), where "e" is the number of existing links and "n" the number of agents.20
To calculate indicators and construct networks we used TextToPajek software developed at the Universidad Complutense de Madrid21 and PAJEK analysis of networks software.22
RESULTS
Original Articles accounted for 455 (46.43%) of the 980 papers analyzed; Brief Reports accounted for 270 (27.55%) papers. Table 1 presents the distribution by year of publication and section.
Of the papers analyzed, 95.1% (n=932) were the product of collaboration between 2 or more authors, whereas only 4.9% (n=48) were the work of a single author (Table 2). We identified 6108 signatures, giving a mean of 6.23 (3.1) signatures/paper. The ISP has held practically constant, with values near to 6 over the 6-year period. We found no statistically significant differences according to year of publication, but did find significant differences according to the section where papers were published (P<.02). Papers published in Original Articles attained higher ISPs during the study period (mean, 7.87 [2.88]), followed by Special Articles (mean 6.59 [5.02]) and Brief Reports (mean, 5.55 [1.2]). The sections with the lowest indices were Review Articles (mean, 2.72 [1.23]), Controversies (mean, 2.75 [2.87]), Update (mean, 2.95 [1.98]) and Images in Cardiology (mean, 3.3 [1.11]).
The number of different authors published during the period was 2927, with a mean of 2.99 authors/paper (Table 2). The overall mean of authors/paper for the entire period is lower than the annual mean because the number of different authors fell as the study period progressed.
Table 3 presents the 44 authors who published >11 papers and their patterns of collaboration, including the number of signatures and collaborators in papers in which they have intervened and the signatures/paper and authors/paper indices. Some authors stand out, even though they are not among the leaders in the productivity ranking, gathering a high number of signatures and a wide nucleus of collaborators, from which they derive high a ISP and high mean of authors/paper.
Applying a threshold of 6 or more coauthored papers has identified 25 clusters of authors with a high intensity of collaboration, consisting of 112 authors from 29 different institutions. Of the 73 most productive authors, 54 (73.97%) are participate in one or other of these clusters; this increases if we consider the 24 most productive authors (>14 papers), as 20 of them belong to one or other of the clusters (83.33%). Similarly, of the 112 authors in these clusters we identified 58 (51.78%) who, although not at the top of the productivity ranking (they stand in positions between 74 and 197), are characterized by the high intensity of their collaboration with other authors. The cluster with most authors was that led by M. Valdés Chavarri, composed of 14 authors; next are clusters around F.J. Chorro Gascó/J. Sanchís Forés, and M.P. Anguita Sánchez/D. Mesa Rubio, with 9 authors in each case (Figure 1). Among the clusters with 9 or more members, the highest cohesion index is held by F.J. Chorro Gascó/J. Sanchís Forés (0.64); among clusters of 5-7 members those centered on A. Medina Fernández Aceituno and J. Soler Soler stand out, with cohesion indices of 1 and 0.9, respectively (Table 4).
Figure 1. Clusters 1 to 3 (9 or more members) with an intensity of collaboration of 6 or more coauthors.
Table 5 presents the statistical values of centrality of authors calculated considering all coauthorships identified. Some authors who do not occupy leading positions in the productivity ranking exercise an outstanding role as "intermediaries" with other authors or of "proximity" to these in the coauthorship network.
In 504 papers (51.43%) we found observed some type of (intrainstitutional or interinstitutional) collaboration, in contrast with 476 papers (48.57%) in which there is no collaboration. Considering the 473 papers in collaboration in which Spanish institutions intervene (Table 6), we find intrainstitutional collaboration comes first (40.87%), closely followed by collaboration between institutions within the same autonomous region (39.7%). Some distance behind that we find collaboration between institutions from different autonomous regions (15.41%) and international collaboration (4.02%). It is important to note a tendency towards increased collaboration when papers are analyzed by year of publication, as the number of papers authored in collaboration has grown from 43.02% in 2000 to 56.69% in 2005, when the highest collaboration index of the whole study period appears (Table 6).
Table 7 presents the 42 most productive institutions (>9 papers) and their patterns of interinstitutional collaboration. The productivity ranking is headed by Complejo Universitario La Paz, (Madrid), with 50 papers; Hospital General Universitario Gregorio Marañón (Madrid), 45 papers; Hospitals Vall d'Hebron (Barcelona) and Hospital Universitario Vi rgen de la Arrixaca (Murcia), 40 papers; Clínica Universitaria de San Carlos (Madrid), 39; Hospital Clínic i Provincial de Barcelona, 34; and Hospital Ramón y Cajal (Madrid), 33. Another 12 institutions published >20 papers and 23 institutions, 10-19 papers.
To establish institutional collaboration networks we applied a threshold of 3 or more collaborations, representing the differing intensity of collaboration through different thicknesses in the links. Figure 2 shows the principal nucleos of the institutional network. In terms of statistics of centrality of institutional collaboration (Table 8), the principal agents in the network are, first, Hospital Universitario La Fe (Valencia) and, second, Hospital de la Santa Creu i Sant Pau (Barcelona), for the 3 indicators. Other institutions that occupy an outstanding place in terms of the number of different institutions with which they collaborated (degree) are Institut Municipal d'Investigació Mèdica (Barcelona), Hospital Clínic i Provincial de Barcelona and Hospital General Universitario Gregorio Marañón (Madrid); all 3 placed among the first 10 in the betweenness ranking. Following them, we find Complejo Asistencial de Salamanca, Complejo Universitario La Paz (Madrid), Hospital Universitario Virgen de la Arrixaca (Murcia) and Complejo Hospitalario Universitario Juan Canalejo (A Coruña); all these also play an outstanding role in the rest of the indicators.
Figure 2. Principal nucleus of network of institutional collaborations with intensity of collaboration of 3 or more collaboration relationships.
DISCUSSION
The method applied has enabled us to identify the most productive authors and institutions, and the composition of 25 clusters of authors who collaborate intensively in cardiovascular medicine in Spain, all of whom published in Revista Española de Cardiología during 2000-2005, characterizing their scientific activities through bibliometrics indicators. However, we should make some observations on the method applied.
1. The scientific production analyzed has been drawn exclusively from Revista Española de Cardiología. An exhaustive, complete review of cardiovascular medicine would need to take into account all the Spanish journals, as well as the scientific production of Spanish authors in journals published abroad. However, the method applied has the advantage that, as we are dealing with a study conducted on one of the clinical journals edited in Spain with higher impact factor in the scientific community, the results obtained represent a panorama of the nucleus of greater repercussion and quality of research in Spain.23
2. The problems of normalization. The importance of normalizing names of authors signing scientific papers is fundamental in coauthorship analysis studies to avoid errors caused by failing to recognize different forms of the name and surnames of a single author, or considering scientific production of 2 or more different authors with the same names as belonging to one individual. We have tried to avoid these errors thru careful manual supervision of the bibliographic references managed, as described in detail above (Methods).
3. The study of the relationships between scientific papers. Analysis of references and bibliographic citations between scientific papers enjoys a long tradition in bibliometric studies.24 However, analysis of coauthorship to construct author networks is a more recent object of study12 and uniform criteria to identify communities or research groups within previously constructed networks do not exist. This represents an obstacle to the interpretation of results, especially when comparing earlier studies conducted with different methods.18,19
Independently of the country or discipline being studied, one of the principal phenomena that can be seen in diachronic studies of scientific collaboration is the trend towards a growth in cooperation between researchers, as measured by the ISP.25 In biomedicine, a recent study of the number of signatures/paper has highlighted the fact that the mean has increased from 4.5 in 1980 to 6.9 in 2000.26 In Spanish medical journals, the ISP has risen from 3.47 at the beginning of the 1980s to 4.59 at the beginning of the 1990s.27 In cardiovascular medicine, in 1990-1993, the ISP of Spanish scientific production included in the Science Citation Index was 5.78,19 a trend to growth that has been maintained, with the ISP currently at 6.23. However, we should bear in mind that in some article types the number of signatures permitted has been limited, which means the ISP tends to be lower.
The tendency to increase the number of signatures/paper may be due to different causes, among which we should point out the complex, interdisciplinary nature of current medical practice28,29 and, as a consequence, the need to collaborate with external teams.30 The ISP obtained is greater than that found in journals such as Revista Española de Anestesiología y Reanimación (ISP=3.1),31 Atención Primaria (ISP=3.77),32 Farmacia Clínica (ISP=3.86),33 and in fields such as drug dependency (ISP=4.1)34 and HIV/Aids (ISP=4.72).35 However, it is very similar to that of Actas Urológicas Españolas (ISP=6.1).36 The number of signatures should be viewed in the light of the abuse of signatures in scientific papers in the form of hyper-authorship that does not correspond to the real contribution of the authors to papers but to a variety of reasons, among which we would highlight those related with the increase in scientific prestige and obtaining sources of finance.37-39 To avoid abuses in the coauthorship of scientific publications various approaches and initiatives have arisen, among which the recommendations contained in the International Committee of Editors of Medical Journals' proposal should be noted. Their view on the signatures of papers40 has been criticized as excessively rigid and impossible to fulfill41 in favor of others which propose describing precisely the contribution of the signatories42 or designating one as responsible for the publication.43 It is beyond the scope of this paper to identify the influence these factors may have had on the ISP of Revista Española de Cardiología
In parallel with the increased ISP in scientific papers, we note increased productivity of authors. In fact, 32.2% of authors responsible for Spanish scientific production in cardiovascular medicine included in the SCI in 1990-1993 published >1 paper,19 and this rose to 35.76% in the present study.
The identification of the most productive authors and of clusters of authors who collaborate closely can be of use in decision-making both in public research organizations and by coordinators of networks when they have reliable information about consolidated research groups and their members. We have detected 10 clusters in which one of the institutions involved is included in RETIC. Similarly, 2 of the coordinators of these networks (Soler Soler and Marrugat de la Iglesia) are among the most productive authors, although a third (Bayés de Luna) has not been identified in the present study because, with few exceptions, the majority of his papers appear in journals published abroad. The existence of numerous clusters indicates that, despite policies of promoting interinstitutional, multidisciplinary research, there remain numerous research groups with no cohesion among themselves, with a limited number of members, with a mean of 4.48 authors for the 25 clusters formed with a threshold of collaboration of 6 or more papers. However, if we consider a threshold of 4 or more coauthored papers, the number of clusters would rise to 46, made up of 278 authors with a mean of 6.04 authors per cluster, and the largest cluster consisting of 41 authors. For a threshold of 3 or more coauthored papers, the number of clusters is 58, with a mean of 8.36 authors per cluster, linking 187 authors directly or thru intermediaries. If we increase the threshold for institutional collaboration, only a very small number appears linked by stable collaborative relations. So, for a threshold of 3 or more papers in collaboration, the principal nucleus is formed of 34 institutions, with 7 subgroups made up of 18 institutions isolated from the principal nucleus. For a threshold of 4 or more papers in collaboration the principal nucleus is formed by only 10 institutions, with 6 subgroups made up of 20 institutions isolated from the principal nucleus. Together with the limited number of institutions that maintain stable, consolidated, collaborative interinstitutional relations, another of the most significant aspects is the marked nature of cooperation between institutions and within autonomous regions as together these add up to 80.57% of collaborations.
Measures of centrality permit us to identify authors and institutions that occupy outstanding positions in the networks of coauthorship and institutional collaboration, on the basis of indicators such as number of authors and institutions which they have collaborated with, implying a greater capacity to access and exchange information.
If we only considered papers in the "Original articles" section when forming networks, we identify the same clusters of authors and a similar network of institutional collaborations. From this, we deduce that this section constitutes the nucleus of the journal where the research conducted by the regular collaborators of the journal are published.
CONCLUSIONS
The most relevant conclusions of the present study are:
1. Some 95.1% of papers in Revista Española de Cardiología are the product of collaboration between 2 or more authors, with indices of signatures/paper and authors/paper similar or above those of other journals and disciplines in biomedicine.
2. We identified various clusters of authors who constitute established groups that can be considered the cutting edge of research in cardiology published in Spanish.
3. Despite the high number of coauthored papers (95.1%) and policies that promote interinstitutional, multidisciplinary research, numerous isolated research groups, with a limited number of members, remain.
4. Even though 51.43% of papers are the product of collaboration between institutions, this mainly occurs within institutions and autonomous regions, although in 2005 we did note a clear increase in collaboration between autonomous regions.
In addition to the topics dealt with in the present paper, we could go more deeply into the scientific activity of the groups identified thru bibliometric indicators of productivity, repercussion and impact.14,18 Given the dynamic nature of science and of research groups, it would also be interesting to study the evolution of these groups over time and to broaden the scope of our analysis of Spanish cardiovascular research activity to other journals, both national and international.
Correspondence: Dr. J.C. Valderrama Zurián.
Instituto de Historia de la Ciencia y Documentación López Piñero. Facultad de Medicina.
Avda. Blasco Ibáñez, 15. 46010 Valencia. España.
E-mail: Juan.Valderrama@uv.es
Received January 4, 2006.
Accepted for publication September 28, 2006.