Bibliometrics

Bibliometrics is a method used to analyse scholarly publications and identifies a range of characteristics, including the general themes contained within the published work and the existence or non-existence of collaborative networks of scholars, as well as a wide range of other measures (Ball 2017).

While bibliometrics has been used by information and library scientists for several years, it is only within the past decade that it has been used extensively by nurse researchers (Smith and Hazelton 2011). There is a wide range of measures both at an aggregate level of analysis (across the population of interest, in this case the fellows), and an individual level (the authors or specific fellows), that can be used to conduct bibliometric analysis. However, only aggregate metrics relevant to this study will be covered in detail in this article.

In the past, rudimentary analysis of individual scholarship has been conducted to quantify the output of nurse academics in the UK, Canada and Australia (Hack et al 2010, Thompson and Watson 2010, Hunt et al 2011). All of these previous studies have examined metrics relating to individuals, but this study encompasses general individual and population measures.

Data collection and analysis

The RCN maintains a roll of honour of fellows and honorary fellows. For the purpose of this study, the analysis was limited to individuals who have been awarded full fellowship of the college, with honorary fellows not included (RCN 2020c).

There are three potential aggregate sources that could be used to extract bibliometric data relating to the RCN fellows and the citation of their work – Scopus, Web of Science or Google Scholar. Of the three, Scopus has the most comprehensive coverage of nursing-related content, and although it requires individual scholars to ensure their entries are up to date, the authors selected it as the database of choice for extracting the necessary information.

There are several challenges associated with indexing databases such as Scopus, including variations in the spelling of author names or institutional attribution (Jacso 2008, Hack et al 2010). Additionally, on occasion, work published by an author with a name similar to the author of interest can be misattributed, so careful data cleansing is necessary. While Scopus has the most comprehensive coverage of nursing content, it does not index all of the nursing material produced. Some journals and grey literature are not indexed in the Scopus database; therefore, the analysis presented in this article was an under-representation of the scientific contribution of fellows. Nevertheless, the study provides a valuable starting point.

The study focused on five variables of interest for each fellow – the total number of indexed publications, the number of citations received, the number of citing articles, the h-index (highly cited index) and the number of co-authors. Using the summary analysis function of Scopus, these five variables of interest were recorded in a tabular format for each fellow to calculate the basic aggregate statistics of median, average (mean) and maximum value for each variable.

The h-index was developed by Hirsch (2005) as a means of quantifying the aggregate effect of an individual’s scholarship and is increasingly used in the academic sector to assess the output of researchers (Thompson and Watson 2010). The h-index is calculated based on both the number of articles published and the frequency of the citation, and is insensitive to the potentially distorting influence of a small number of highly cited works (Watson et al 2016).

Several authors have offered suggestions on the importance of either the number of citations that an article receives, or the h-index of the author as a means of identifying the most prolific and impactful scholars (Hack et al 2010, Thompson and Watson 2010). These authors have noted that the cut points (the numerical values that represent boundaries between categories) that determine what constitutes a ‘good’ number of citations or h-index may vary between disciplines, roles within the same discipline or even the subtopics being studied. One way to objectively assess such data is to use Jenks’ natural breaks analysis, which identifies natural breaks in the data to ascertain if there are naturally occurring cut points (Jenks 1967, Moffitt 2019).

In addition to the descriptive statistics, multivariate cluster analysis was conducted to determine whether the fellows comprised a connected scholarship community (co-author analysis of authorship). To achieve this, the software package VOSviewer, which was developed at the University of Leiden in the Netherlands by van Eck and Waltman (2010), was used. This software examines the bibliographic data relating to each article and examines the co-authors of individual fellows. A visualisation of the co-authors’ network is then displayed in the form of a cluster diagram, with individuals who regularly work with each other allocated to a colour-coded cluster. However, the physical location of an author is calculated based on the relationships between all authors and co-authors and consequently individuals can be ‘stacked’ on top of one another, resulting in only the uppermost author being visible on a printed image. In large datasets such as the one used in this study, the consequence is that when a two-dimensional cluster diagram is displayed by the software, some authors’ names are obscured by their peers. While it is not ideal that the visual representation of the co-authors’ network obscures the work of some fellows at the expense of others, this is an inherent limitation of the software, because it seeks to present an overview of the fellows’ collective work. When using the VOSviewer software, it is possible to zoom into and rotate the image to view a particular individual, but this will then obscure others.

The VOSviewer software also enables the frequency of co-authorship to be set. Callon et al (1983) stated that a minimum of three articles is sufficient to identify regular co-authorship. Furthermore, by examining the co-author’s country of affiliation, the extent of any network of collaborators beyond the UK can also be displayed.

To ascertain the general themes associated with the work of the fellows, a co-occurrence analysis of author keywords was conducted. A co-occurrence analysis examines the keywords that the author has allocated to an article. The VOSviewer software examines the most frequently occurring keywords and clusters them together based on their co-occurrence within articles. As a result, clusters of commonly occurring keywords are generated, then mapped as an image. Closely related keywords are located near to each other and non-related words are located further apart. These keyword clusters can contain a mixture of research methods, the populations being studied as well as the subject matter under investigation. To summarise the content of the keyword clusters, each of them was reviewed independently by the authors of this article and a brief title and succinct description was generated for each cluster. Having completed this work independently, the authors compared their results and where there were differences in interpretation, these were discussed until agreement was reached.

After parsing the data to identify commonly occurring terms and the relationships between them, a visualisation of the relationship between the variable of interest is constructed and displayed by the VOSviewer software. The more related the concepts, authors, countries or keywords are to one another, the closer they are located on the visualisation. The more prolific a topic, author or country is, the larger the circle drawn on the mapped image. Connections marked by lines between the circles on the mapped image display the relationships between the variables being studied; the thicker these lines, the stronger the connection between the two variables.

Ethical approval

The study analysed data obtained from a publicly available database and used a secondary analysis of literature, therefore no ethical approval was required. However, for the purposes of courtesy, RCN fellows were informed in advance that the analysis was being undertaken.

Multivariate analysis

Figure 1 provides a mapping of the connections between fellows and their co-authors. Many of the co-authors are not fellows and, in some cases, come from disciplines other than nursing. As can be seen in Figure 1, this is a highly complex and connected map. Several fellows are prolific authors and have many co-authors. In this two-dimensional image, some fellows are obscured by having the data of other authors placed on top of their own data.

Figure 1.

Mapping of the connections between fellows and their co-authors

It was identified that the most prolific fellows are David Thompson and Roger Watson, but the larger circles show that there are a significant number of fellows who – over a period of just under 20 years on average – have contributed to nursing scholarship on multiple occasions.

To gain further insight and to visualise an individual fellow’s one-step connections (those who the fellow is directly linked to through publishing activity), it is possible to use the VOSviewer software to identify a fellow’s network of co-authors by clicking on their name, which will then detail their one-step network. For example, Figure 2 shows the one-step co-author network of David Thompson’s Scopus indexed work. Authors who are part of the same cluster have the same colour and form a network that routinely works on related issues.

Figure 2.

One-step co-author network of David Thompson’s Scopus indexed work

Co-authorship analysis of collaborations by country yielded a complex and diverse network of 72 country or territory connections. As can be seen from the co-authorship connections by country or territory shown in Figure 3, the most dominant countries or territories (those with the largest-sized circles) are the UK, Australia, the US, Hong Kong and Canada. There are also significant connections to, and among, various European countries. Additionally, there are connections to Middle Eastern and Asian countries, Commonwealth member states, and even the Pacific Islands territory of Wallis and Futuna.

Figure 3.

Co-authorship connections by country or territory

Figure 4 shows a co-occurrence analysis of the connectivity between 893 author-defined keywords that met the minimum frequency requirement of at least three occurrences. These are grouped into 24 keyword clusters comprising 10-67 terms per cluster. Due to the density and overlapping nature of the clusters, which limited the visibility of underlying terms, a detailed co-occurrence analysis of the clusters was conducted and is summarised in Table 3. Table 3 provides a brief title and succinct description of each of the keyword clusters from the co-occurrence analysis, which were derived from analysis of the terms included in each cluster.

Figure 4.

Co-occurrence analysis of the connectivity between 893 author-defined keywords

Limitations

There are four limitations that need to be considered in relation to this study.

First, while Scopus has the most comprehensive coverage of material and offers integrated citation data on indexed articles, it is incomplete because not all peer-reviewed journals are included and on occasion there is a delay in the publication of an article and its subsequent inclusion in the index. As a result, the findings of this study underestimate the contributions made by fellows to the scholarship literature. This has implications for underestimating the extent of the co-authorship networks established and maintained by fellows. It is likely that the fellows have a wider and more extensive national and international network of collaborators than the analysis presented in this article demonstrates.

Second, on occasion Scopus misattributes the work of one individual to another, resulting in either an overestimation or underestimation of a fellow’s contribution. By carefully cleansing the data, the authors were able to ensure that articles not authored by a fellow were removed. As a result, the authors do not believe that errors in attribution have resulted in any overestimations of fellows’ profiles. By completing the two-stage search for authors, the authors of this article believe they have minimised underestimation of fellows’ profiles, but they cannot provide a definitive statement on this issue. Given that one of the aims of the Nursing Now (2020) campaign is to ensure that nurses and midwives have an increasingly prominent voice in health policymaking, the RCN may wish to emphasise this issue to the fellows and their membership to ensure that nursing in general, and individual nurses in particular, are fully discoverable through bibliographic searches.

Third, although modern technology can capture data concerning mentions of an individual’s work on social media platforms such as Twitter and LinkedIn, this is not a factor that the software used in this study was able to accommodate.

Fourth and finally, this study explored data from indexed sources and as a result has not captured the entirety of contributions made by the fellows, particularly those who work in clinical or policy areas. Based on these limitations, the authors conclude that the findings of this article do not fully document the scholarly impact of fellows, but nevertheless offer an important insight into their interests and connections.