Comparison of citation and usage indicators: the case of oncology journals Christian Schloegl ��� Juan Gorraiz Received: 11 May 2009 / Published online: 13 February 2010 �� Akademiai �� Kiado, �� Budapest, Hungary 2010 Abstract It is the objective of this article to examine in which aspects journal usage data differ from citation data. This comparison is conducted both at journal level and on a paper by paper basis. At journal level, we define a so-called usage impact factor and a usage half- life in analogy to the corresponding Thomson���s citation indicators. The usage data were provided from Science Direct, subject category ������oncology������. Citation indicators were obtained from JCR, article citations were retrieved from SCI and Scopus. Our study shows that downloads and citations have different obsolescence patterns. While the average cited half-life was 5.6 years, we computed a mean usage half-life of 1.7 years for the year 2006. We identified a strong correlation between the citation frequencies and the number of downloads for our journal sample. The relationship was lower when performing the analysis on a paper by paper basis because of existing variances in the citation-download- ratio among articles. Also the correlation between the usage impact factor and Thomson���s journal impact factor was ������only������ moderate because of different obsolescence patterns between downloads and citations. Keywords Journal metrics Journal impact factor Usage impact factor Cited half-life Usage half-life Introduction After the advent of electronic journals, it has become much easier to collect journal usage data. Contrary to the hard copy era, it is possible to view scholarly communication through C. Schloegl (&) Institute of Information Science and Information Systems, University of Graz, Universitaetsstr. 15/F3, 8010 Graz, Austria e-mail: christian.schloegl@uni-graz.at J. Gorraiz Library and Archive Services, Bibliometrics Department, University of Vienna, Boltzmanngasse 5, 1090 Vienna, Austria e-mail: juan.gorraiz@univie.ac.at 123 Scientometrics (2010) 82:567���580 DOI 10.1007/s11192-010-0172-1

the eyes of the reader now (Rowlands and Nicholas 2007, p. 223). Besides easy and cheap data recording, usage metrics have further advantages compared to citation metrics. They are available much more timely and reflect use on a much broader scope (Duy and Vaughan 2006, p. 513). While citations can be considered an indirect measure of use but a direct measure of usefulness, reads could be seen the other way round. Accordingly, usage metrics can be regarded complementary to citation metrics (e.g. Armbruster 2007 Bollen et al. 2005). Since e-journal usage data have been available for several years only, there is a need for further research. Kurtz et al. (2005) even expect similarities and differences of reads and citations to become a central issue of bibliometric research. Several usage indicators have been suggested in recent years. For instance, Darmoni et al. (2002) introduced a so-called reading factor. Most suggestions base on the classical citation indicators from the Journal Citation Reports (JCR), using download data (usually full-text article requests) instead of citations. The corresponding usage metrics are usage impact factor (UIF) (Rowlands and Nicholas 2007 Bollen and Van De Sompel 2008), usage immediacy index (Rowlands and Nicholas 2007) or download immediacy index (Wan et al. 2008), and usage half-life (Rowlands and Nicholas 2007). According to Bollen and Van de Sompel (2008) comparisons between citation and usage data can be made at a local and at a global level. Most of the existing studies used local usage data. Darmoni et al. (2002) accessed data from the Rouen University Hospital digital library and compared download and JIF rankings. Duy and Vaughan (2006), who analysed journal titles from the subject areas of chemistry and biochemistry at Concordia university library, found a significant correlation between electronic journal usage and local journal citation data but no relationship with the JIF. Also McDonald (2007) detected a medium relationship between local article downloads and locally collected citations for more than 1,500 journals (mainly from sciences) owned by the California Institute of Technology. Bollen and Van de Sompel (2008) collected usage data from the entire California State University system and aggregated them for more than 6,000 journals. Their analysis shows a modest negative correlation between the UIF and the JIF. The analysis at discipline level revealed that cor- relations differ for particular disciplines. A weak positive relation could be observed espe- cially for disciplines with a higher proportion of graduate students and graduate faculty. Global usage studies were mainly performed using subject repositories and open access journals. Chu and Krichel (2007) investigated the top 200 documents at RePEc, a digital library in economics. They related the number of downloads to the number of citations from SSCI and Google Scholar and found a moderate correlation between them in both cases. Brody et al. (2006) used download and citation data from arXiv.org and identified a signif- icant and ������sizeable������ correlation between citations and downloads, which varied between 0.33 and 0.46 depending on the considered field. Kurtz et al. (2005) used the NASA Astro- physics Data System for their research. Besides examining the relationship between citation and readership numbers, the authors also compared the obsolescence function as measured by readership with the obsolescence function as measured by citations. Another study investi- gating the relationship between downloads and citations and possible differences in obso- lescence patterns for articles published in a single journal was conducted by Moed (2005). In particular, Moed tried to determine the effect citations have on downloads and the effect of initial downloads made during the first three months of an article���s life-span on citations later- on. Wan et al. (2008) concentrated on the download immediacy index and explored its relationships with several citation indicators. They used the Chinese Academic Journal Full- Text Database in order to retrieve the download data. The citation indicators were extracted form the Chinese Academic Journals Comprehensive Citation Report which is an equivalent to JCR. 568 C. Schloegl, J. Gorraiz 123