OP0339 ALTMETRIC - DOES SOCIAL MEDIA IMPACT CITATION RATE IN RHEUMATOLOGY JOURNALS?

Abstract

Background:The coloured altmetrics donut has become a standard feature of online publications. The colours depict different online sources by which an article was mentioned, while the number in the donut, the Altmetric Score (AS), reflects the summarised attention an article has received. The Dimensions database joins citations from any kind of scientific or mainstream publication. Studies analysing the link between the AS and the citation rate of an article suggest that this connection is strongly dependent on the field of research, the type of article and the type of analysis used.Objectives:To analyse the connection between AS and citation rate in articles published in rheumatology journals.Methods:We retrieved data on article usage, AS and citations of articles published in ARD and RMD Open between January 2015 and November 2019. For time-dependent analyses on the influence of AS on citations, articles published in 2019 were excluded. Forward-stepwise regression models were used to explore factors influencing total citation rates. We performed subanalyses, dividing articles in categories of correspondence, original research and editorials/viewpoints. We dichotomised articles by reaching the top 25% in terms of citation count within the first, second, third and fourth year after publication according their category. We explored the risk of reaching these top 25% in dependency of AS using logistic regression (log transformed AS) and receiver operating curve analyses (ROC, reported cut-offs were identified coinciding with 80% specificity).Results:We used 1597 articles published in ARD and 409 articles of RMD Open with complete data on AS and article usage within the mentioned timeframe. AS are higher in more recently published articles (p=0.04, ß: 1.3 per year), but the number of Dimensions citations is lower in more recently published articles (ß: -8.5 per year, p≤0.001). Twitter shows by far the highest activity among the AS subcategories (highly correlating with AS r=0.8, p≤0.001). The total number of twitter mentions increased by 2.8/year from 2015 to 2019, indicating that more recently published articles were more often picked up on twitter. Changes in R2in the regression model indicated that besides time since publication and AS, also the type of article influences citation count. For original research and editorials, AS may significantly add to the variability of the citation count, which was not the case for correspondences. The influence of AS on citation count of editorials added 16% to the 12% variability explained by publication time. Both factors showed similar ß-coefficients (months: ß: 0.76; ß: AS: 0.83). This effect was smaller in original articles (month: ß=0.74; AS: ß=0.11, Total R2: 23.7%). AS significantly coincides with reaching the top 25% of citation counts according to time since publication. For the first year those articles with AS >15 showed a positive Likelihood Ratio (+LR; 95%CI) of 1.6 (1.4-1.9) to reach the top 25%, the second year AS>15: +LR: 1.9 (1.6-2.2), the third year AS>13: +LR 2.3 (1.9-2.7) and in the fourth year AS>12 +LR: 2.1 (1.7-2.7). This effect was again different between publication categories, with no effect of AS in correspondence articles. Figure 1 highlights that AS influences citations of editorials to a larger extent than of original articles, except within the first year of publication.Conclusion:We could show, that Altmetric scores are higher in more recent publications and seem to have an influence on the number of citations, particularly early citations in the first two years after publication. The early implementation of published articles in social media could help to increase overall citation.Acknowledgments:We thank Lindsey Fountain from BMJ for providing citation and usage data.Disclosure of Interests:Paul Studenic Grant/research support from: Abbvie, Caroline Ospelt Consultant of: Consultancy fees from Gilead Sciences.