Do bibliometric findings differ between Medline, Google Scholar and Web of Science? Bibliometry of publications after oral presentation to the 2013 and 2014 French Society of Arthroscopy (SFA) Congresses

International audienceIntroductionBibliometrics consists in quantitative and qualitative analysis of an individual's or group's communication (volume, visibility), and impacts research funding. There are a number of bibliometric data sources, functioning in different ways and liable to give rise to differing statistics. This point has not been investigated in relation to publication following presentation to a French congress. We therefore conducted a study comparing the main bibliometric instruments, aiming to assess: (1) publication rates following oral presentation to the 2013 and 2014 French Society of Arthroscopy (SFA) Congresses according to the database used, and (2) citation rates for these publications according to database.HypothesisPublication and citation rates differ according to database.Material and method All 199 Abstracts of oral presentations to the 2013 and 2014 SFA Congresses were included. Based on author names and key-words, manual search was conducted in the Medline, Web of Science and Google Scholar databases. Publication characteristics (citation rate) were studied using the 3 databases and the French SIGAPS (Système d’Interrogation, de Gestion et d’Analyse des Publications Scientifiques: Scientific Publication Search, Management and Analysis System) website.ResultsPublication rates according to Medline and Google Scholar were the same (48.2%: 96 articles for 199 presentations), but significantly lower on Web of Science (44.7%: 89/199; p = 0.002). Citation rates differed significantly (p < 0.001) between sources, with Google Scholar listing a mean 1.5-3.4-fold more citations per article than the other 2 databases. Citation rates between the 3 databases correlated strongly (r = 0.93).DiscussionThe example presented in this study illustrates the differences in bibliometrics found between different databases. There was a 4% difference (7/199 articles) in publication rates following oral presentation to an SFA Congress, and even greater differences in citation rates per article, with 1.5-3.4-fold more citations according to Google Scholar. Bibliometric studies need to acknowledge the database(s) being used, which should be as many as possible to enhance exhaustiveness.Level of evidenceIV; descriptive epidemiologic study