Mapping international impact of Danish neuroscience from 2004 to 2015 using tailored scientometric methodology

This study was undertaken to assess the standing of Danish neuroscience and how this broad research field has evolved in Denmark in recent years as compared to a number of world-leading neuroscience countries. This article outlines the hitherto most comprehensive study of international research performance in the neurosciences using an elaborate search methodology to identify relevant research articles and robust advanced scientometric indicators to quantify performance. Despite the considerable interest in neuroscience research, only a few scholarly studies have examined aspects of research performance, and no study has carried out this comprehensively across carefully identified research specialties or comparatively at the level of countries (e.g., Sengupta, 1989; Schwechheimer & Winterhager, 2001; Glänzel et al., 2003; Sorensen & Weedon, 2011; Shahabuddin, 2013; Ohlendorf et al., 2015; Buchan et al., 2016; Leitner et al., 2016). Two recent commercial reports, though, have examined neuroscience research performance, one for a specific research initiative in the UK (Gunashekar et al., 2015), and the other, globally for selected countries (Elsevier, 2014); albeit, the latter study lacks a proper methodological outline, and it is overly reliant on Elsevier's standard analytical tools. Commissioned by the Lundbeck Foundation (www.lundbeckfonden.com), this study aimed at examining the international standing of Danish neuroscience research and how this field has developed over the past decade (2004–2015) compared to USA, Canada, the Netherlands, UK, Germany, Switzerland and Sweden. These countries were selected for comparison because they were the highest ranked of all countries in a previous unpublished performance analysis based on the Web of Science (WoS) neuroscience subject category. Contrary to previous analyses of neuroscience, and traditional scientometric performance analyses in general, this study was based on a specially constructed publication set which is assumed to broadly cover neuroscience and related research areas (i.e. here neuroscience encompassed basic and clinical research including neurology and psychiatry). The publication set was constructed by querying the PubMed database for neuroscience topics using a combination of Medical Subject Headings (MeSH), title and abstract words, and core journals. Subsequently identified publications from PubMed were matched with corresponding records in the WoS citation database, enabling advanced scientometric analyses. This approach has several advantages over traditional analyses; most importantly, we identified relevant articles according to their subject indexing at the individual publication level. Traditional performance analyses rely on arbitrary journal subject categories for delimiting the topic of interest, which often leads to the inclusion of irrelevant as well as exclusion of relevant papers published in multidisciplinary journals some of them with high impact. A comprehensive description of our methodological approach, including search formulations, is available in the Appendix S1. Briefly, we used MeSH subject headings in a recall-oriented search strategy to retrieve neuroscience publications from 2004 to 2015 in PubMed (n = 2 130 473). As citation data are only sporadically available in PubMed, matching to WoS was required. Using the approach outlined in Fig. 1, this was possible for 87% of records (n = 1 853 229). Restricting the set to the eight countries included in this study limited the number to 1 188 653 records. As indicated on the right-hand side of Fig. 1, the carefully developed search strategy was capable of identifying a substantial number of relevant neuroscience publications outside of the traditional WoS subject categories 'neuroscience', 'clinical neurology' and 'psychiatry'. The scientific impact of the analysed units (e.g. countries or regions or institutions) was established using the following standard indicators (Waltman et al., 2012): To be comparable across time, type and field, all citation indicators were normalised according to publication type, publication year and average citation rates in subject areas (Lundberg, 2007). We have not provided uncertainty estimates in relation to the indicators presented. The issue of quantifying uncertainty is a contested topic especially in relation to research evaluation and scientometric indicators (Schneider, 2013, 2015, 2016; Hicks et al., 2015). A current pragmatic compromise, although still disputed, is to use so-called 'stability intervals', which are bootstrapped confidence intervals (Colliander & Ahlgren, 2011; Schneider & van Leeuwen, 2014). We do present such intervals for the calculated citation indicators in section two of the Appendix S1. We strongly caution against interpretations of citation indicators that link them directly to research 'quality'. Research quality is a complex and multidimensional concept, which is very difficult, if not impossible, to measure by a single or even more indicators. However, citation impact is often used as a proxy for some vague notion of quality (Gläser & Laudel, 2007). While the number of citations to a single paper cannot be a measure of its quality per se, the citation impact for an aggregate set of papers linked to a unit, for example, is a different matter. We only consider the impact of sets of publications, and the recommended interpretation of an impact level higher than the expected is that a unit's set of publications is highly visible among peers. Hence, citation impact is an important performance measure as it examines the use of the research literature and therefore is indicative of research fronts and base knowledge of fields of research. Indeed, the concept of 'impact' in scientometrics was constructed to signify the difference from 'quality' (Martin & Irvine, 1983). Publication output generally increased over the period examined (Fig 2A and B) for each of the countries in the study. Noticeably, relative growth rates were in most cases inversely related to the country size (Fig. 2C). Interestingly, Denmark was the smallest country examined, but had the highest relative growth in publication output. Three of the four smallest countries, Denmark, Netherlands and Switzerland, clearly stand out with the highest relative growth rates, whereas the growth rate of Sweden is more moderate and comparable to considerably larger countries such as Canada or Germany. Output per capita (represented as publications per 100 000 inhabitants) for the first and last periods shows some interesting trends (Fig. 2B). First, the countries with the largest populations have the lowest per capita output. Secondly, although all countries showed an increase in per capita output over the period analysed, as above, Denmark, Netherlands and Switzerland clearly showed the greatest increase. The size effect of countries is also visible when examining the proportion of collaborative publications (Fig. 3A and B). The smaller countries stand out (Fig. 3A), and 82% of the Danish neuroscience publications in 2013-15 are a result of collaboration between at least two institutions. However, the pattern changes when examining international collaboration (Fig. 3B); here, Switzerland stands out with a consistent 10-per cent point gap over the period relative to the other countries (Denmark highest among them). The proportion of publications with international collaborations has grown at similar rates for all countries, but in the latest periods, the proportion for Switzerland is around 70% whereas the proportion for Denmark is 60%. This observation is important because internationally co-authored publications on average have higher citation rates compared to national collaborative or non-collaborative publications. Note that the USA is an exception to this (Katz & Hicks, 1997). When we examined the journal publication profiles using MNJS for the eight countries, two groups emerged (Fig. 3C). A group comprising Switzerland, USA, UK and the Netherlands had profiles of generally publishing in journals with higher visibility (i.e. journals attracting more citation activity, higher impact journals). Denmark belongs to the second group comprising Canada, Germany and Sweden. The Danish journal publication profile slightly increased until 2010-12 but seems to have stagnated since. Interestingly, the separation between the two groups has remained distinct through the whole period examined. This observation is also important as the journal publication profiles are very good predictors of citation performance at the country level (Bonitz, 1997). Scrutinising the journal publication profiles according to collaboration types (Fig. 3D, E and F) reveals that internationally collaborative papers were generally published in journals with higher impact (Fig. 3F). Publications with no inter-institutional collaboration were generally published in journals with lower impact (Fig. 3D). For all countries, the set of publications with no inter-institutional collaboration constitutes the smallest set, for example since 2009-11, less than 20% of the Danish neuroscience publications. Nevertheless, the general pattern of two distinct groups is less clear when publications are examined according to collaboration type. The journal publication profile for Danish publications with no institutional collaboration changed considerably with a marked rise moving the Danish profile into the top group from 2008-10 to 2010-12, thereafter the Danish profile levelled out and was surpassed by Sweden, Canada and Germany, to become the lowest ranked of all countries in 2013-15. The drop after 2010-12 is remarkable. The development in the journal publication profile for the Danish national collaborations (Fig. 3E) is more stable. This set of publications constitutes slightly more than 20% of the Danish neuroscience publications during the period examined. Interestingly, the Danish journal publication profile for international collaborative publications (Fig. 3F) was below all other countries except Sweden until 2009-11. However, from 2009 to 2011, there has been a marked increase solidly placing the Danish profile among the other countries. This set of internationally collaborative publications constituted 50% of the Danish neuroscience publications in 2004-06 and 62% in 2013-15. Consequently, in the period examined, we see a development in Danish neuroscience research towards more international collaboration; however, similar trends are also visibly for the other countries examined. What sets Denmark apart is the marked changes in journal publication profiles for the national publications with no inter-institutional collaboration and the set of publications with international collaboration. The former profile indicates a change in publication behaviour towards lower impact journals, whereas as the latter indicates a change towards higher impact journals. These contradictory developments combined with the minor but continuous drop in the journal publication profile for the national collaborations seem to level out the overall national MNJS indicator (Fig. 3C) after 2012-14. What should be considered here is the potential influence of these changes in journal publication profiles upon the citation impact of articles. On average, international collaborative publications have higher visibility and thus citation impact. Nevertheless, the other two collaboration types still constitute around 40% of the publications and the developments in their impact relative to the other countries is still important for the overall Danish citation impact. Note that publications with no inter-institutional collaboration generally have slightly higher impact scores compared with publications with national collaboration. The MNJS indicator is a good predictor for citation impact. On an aggregate level, publishing in journals with higher impact will mostly likely also lead to higher citation impact for these publications. It is therefore concerning that the MNJS indicator for the set of publications with no collaboration (Fig. 3D) drops so markedly and that this seems to be unique for the Danish case. The patterns seen in the collaboration and journal publication profiles influence the citation impact performance and developments both for the mean normalised citation scores (MNCS) and the proportion of highly cited publications (PPtop10%; Fig. 4). Two groups of countries are again clearly visible, essentially showing two performance levels where Switzerland, UK, USA and Netherlands stand out. Interestingly, while Switzerland and USA have consistently been the highest performing countries, the UK and to a lesser extent the Netherlands have experienced a continuous increase in impact over the period. In the second group, showing more fluctuation over the period probably due to the smaller publication set, Denmark had the highest performance peaking around 2010-12. It should be emphasised that all countries performed above the database average citation impact (MNCS = 1) and the statistically expected performance levels (PPtop10% = proportion of 10%). Among the countries in the lower performance group, the set of Swedish neuroscience publications is distinctive. Sweden's performance levels are the lowest among all countries examined (see Glänzel et al., 2003). Examining citation impact according to collaboration confirms a marked drop in the impact of Danish publications with no inter-institutional collaboration as predicted by the MNJS indicator (see Fig. S5 in the Appendix S1). Note the similar marked drop occurred in the PPtop10% indicator for Switzerland. It is unclear, however, why these drops occurred and why they are so marked, but it testifies that on an aggregate level, publication in lower impact journals mostly likely influence the actual citation impact of the publications. These findings are documented in the Appendix S1. Interestingly, when we did a traditional scientometric analysis, in which neuroscience is delimited according to WoS journal subject categories, similar overall results were observed with two distinct groups of countries at different performance levels (see Appendix S1). However, the rankings between the countries and historical developments were clearly different. In the highest performing group, the UK consistently had the highest impact whereas Switzerland showed a marked increase over the years reaching the UK level in the last period examined. In the lower performing group, Denmark and Sweden were clearly and consistently the lowest performing of the four countries. Notice, however, that a traditional analysis leaves out relevant neuroscience research published in broad medical and multidisciplinary journals. The latter is important because such journals usually have a markedly higher visibility and are likely to influence the overall citation performance which indeed seems to be the case for Denmark having lower performance levels in the WoS subject category analysis. A final main finding at the country level is the general trend in the proportion of so-called 'industry publications' (publications affiliated solely to industry-linked authors and public–private collaborative publications; Fig. 5). Denmark and Switzerland stand out with clearly the highest proportion of industry publications; this is likely to be a consequence of the relatively high concentration of pharmaceutical companies in the two countries. However, the trend is declining for all countries, especially from 2010 to 2012 onwards, but most markedly for Denmark where the continuous drop began in 2009-11. The financial crisis in 2008-09 could very likely be a partial explanation for these trends. To examine research performance at the institutional level in Denmark (Fig. 6), four major Danish universities and the affiliated university hospitals were focused upon [Copenhagen, Aarhus, University of Southern Denmark (Odense) and Aalborg]. The difference in publication output between the four universities was stable over the period analysed. Copenhagen had the largest output, almost twice as many publications in neuroscience and related areas as Aarhus, ranked second. Aarhus in turn had approximately twice as many publications as the University of Southern Denmark ranked third, while the gap between Aalborg and the University of Southern Denmark was somewhat smaller. It is noteworthy that three pharmaceutical companies (H. Lundbeck A/S, Neurosearch A/S and Novo Nordisk A/S) are among the 10 most productive units in Danish neuroscience research in the period examined. Compared to the country analyses, the analyses at the institutional level arise from relatively small publication sets which means that fluctuations become more likely. Examination of the journal publication profiles clearly shows these fluctuations, although the profiles for the two largest units, Copenhagen and Aarhus, are more stable during the entire period. Until the end of the period, these two units also have the highest journal publication profiles according to the MNJS indicator. When it comes to citation impact, we found that the largest units had the most stable developments with Copenhagen and Aarhus as the highest performing institutions although differences between the units were small. However, from 2009 to 2011, we saw a sudden steep rise in impact for the set of publications linked to H. Lundbeck A/S (pharmaceutical company). Some study designs, such as clinical trials, tend to have higher citation rates (Andersen & Schneider, 2011). In the case of Lundbeck A/S, 20% of the publications from the periods 2009-11 to 2012-14 were clinical trials. These trials had a MNCS impact of 2.08 compared to 1.37 for all other study types, which is a substantial difference. In addition, the Danish Technical University seemed to increase their general impact level in that period settling on a level comparable to Copenhagen and Aarhus. Note that although the Danish Technical University does not have any formal affiliation with university hospitals, it is included in the analyses due to the university's publication volume in the dataset. The funding pattern of Danish neuroscience research reveals four main clusters: (i) a cluster of national and international pharmaceutical companies; (ii) a cluster evolving around the National Institutes of Health (NIH) in the US; (iii) a cluster of European funding sources both EU and national funding agencies; and (iv) a cluster of Danish funding sources both public and private (including the Lundbeck Foundation; Fig. 7). The most prolific Danish funding sources are the Danish Research Council (also known as the Danish Council for Independent Research) and the Lundbeck Foundation. Moreover, there is a strong relation between the Danish Research Council, the Lundbeck Foundation and the Danish National Research Foundation, that is the latter funds Centers of Excellence, which means that these funding sources very often are acknowledged together in the same publications. In addition, NIH funding seems to be the main hub linking public national and international funding activities with funding from pharmaceutical companies. The latter finding supports the general observation that most Western countries have their largest share of co-authored papers with the USA roughly in the range between 15 and 20% of their annual output from public biomedical research (Sørensen & Wiborg Schneider, 2017). The analysis shows that Danish neuroscience research is doing well in terms of both publication output and impact judged from acknowledged impact measures, being placed in the top of a group of countries also comprising Germany, Sweden and Canada. However, there is room for improvement as the USA, Switzerland, UK and the Netherlands cluster in a group well ahead of the aforementioned countries. For all the countries, scientific output is steadily increasing over the years with Denmark having the highest relative growth of publication output. The proportion of publications with international collaboration grows for all countries. In the latest 2-year period, the proportion for Denmark is 60%, while Danish publications originating from a single institution are becoming fewer and with declining impact. Since the financial crisis, there has been a general decline in the number of publications coming out of public-industry collaborations on biomedical research, including neuroscience research. The development for Denmark is however more pronounced as the decline seems to have set in before the crisis and is markedly intensified in the period from 2009 until now. The most prolific funding sources of Danish neuroscience research measured through acknowledgements in publications have been the Danish Research Council, the Lundbeck Foundation and the Danish National Research Foundation, and it is noteworthy that there are mutually strong links between these funding institutions. The work is commissioned and sponsored by the Lundbeck Foundation. Jens Peter Andersen involved in study planning, data collection and analyses. Kim Krogsgaard and Anne-Marie Engel involved in study planning and writing. Jesper W. Schneider involved in study planning, data analyses and writing. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.