Additional support for RCR: A validated article-level measure of scientific influence

In their comment, Janssens et al. [1] offer a critique of the Relative Citation Ratio (RCR), objecting to the construction of both the numerator and denominator of the metric. While we strongly agree that any measure used to assess the productivity of research programs should be thoughtfully designed and carefully validated, we believe that the specific concerns outlined in their correspondence are unfounded. Our original article acknowledged that RCR or, for that matter, any bibliometric measure has limited power to quantify the influence of any very recently published paper, because citation rates are inherently noisy when the absolute number of citations is small [2]. For this reason, in our iCite tool, we have not reported RCRs for papers published in the calendar year previous to the current year [3]. However, while agreeing with our initial assertion that RCR cannot be used to conclusively evaluate recent papers, Janssens et al. also suggest that the failure to report RCRs for new publications might unfairly penalize some researchers. While it is widely understood that it takes time to accurately assess the influence that new papers have on their field, we have attempted to accommodate this concern by updating iCite so that RCRs are now reported for all papers in the database that have at least 5 citations and by adding a visual indicator to flag values for papers published in the last 18 months, which should be considered provisional [3]. This modified practice will be maintained going forward. Regarding article citation rates of older articles, we have added data on the stability of RCR values to the “Statistics” page of the iCite website [4, 5]. We believe that these new data, which demonstrate that the vast majority of influential papers retain their influence over the period of an investigator’s career, should reassure users that RCR does not unfairly disadvantage older papers. Our analysis of the year-by-year changes in RCR values of National Institutes of Health (NIH)-funded articles published in 1991 reinforces this point (Fig 1). From 1992–2014, both on the individual level and in aggregate, RCR values are remarkably stable. For cases in which RCRs change significantly, the values typically increase. That said, we strongly believe that the potential for RCR to decrease over time is necessary and important; as knowledge advances and old models are replaced, publications rooted in those outdated models naturally become less influential. The RCR denominator (the expected citation rate [ECR]) is calculated by aggregating the article citation rates of peer papers that have the same field citation rate (FCR) and are published in the same year as the article in the numerator. FCR, as noted by Janssens et al., is defined as the collective 2-year journal citation rate for all papers in the co-citation network of the article being evaluated. This calculation is conceptually distinct from that used in determining journal impact factors; rather than relying on journal of publication to define an