High phrasal similarity, usually represented by a high percentage of the matching text or a high percentage of the similarity index, has become one of the most perplexing issues for authors, reviewers, and editors working in the medical field worldwide. Authors and editors are afraid of being suspected of plagiarism without any respect to their intentions. As some editors [1] have pointed out that increasing plagiarism obliges research institutes and journals to take measures against text-theft through the internet, and dedicated web-based applications have been widely used to estimate the similarity index. The Korean Journal of Anesthesiology (KJA) has introduced a new process where all of the submitted papers must undergo an estimation of the similarity index since April 2012. The similarity index is a percent score estimating how many phrases of a manuscript are identical to those in a previously published work. Some medical journal editors have closed their submission gate to a manuscript having a high similarity index, by equipping with a filtration step prechecked by a computer-driven system. Still, some of our in-house editors have claimed that they could hardly find any value in papers having a high similarity index. However, others have commented that a high similarity index would be inevitable according to the research discipline or the nature of research itself. Although the percentage of the similarity index has become a standard measure against plagiarism among authors and editors working in the medical writing field, its characteristics, particularly those affecting the value of a paper, have not been investigated. To identify the correlation between the similarity index and the value of a paper, we employed ‘probability to be accepted,’ ‘academics cited-by frequency,’ and ‘appearance in social media.’ More specifically, the research aimed to answer the following questions:

The goal of our research was to eradicate a previous insight that a submitted paper having a high similarity index might hardly have any value. We did not mention about plagiarism or insincere writing throughout the manuscript. Instead, we depict how the similarity index appears, how it affects the decision of a peer review, and how editors are confident about the insights independent of the similarity index.

The authors were permitted to handle the datasets extracted from the online KJA submission system be used in this research at the 62nd Meeting of The Research/Publication Ethics Committee of the KJA Editors (26 January, 2016).

In our research, in a Korean medical journal, some degree of the similarity index was inevitable, and the permissible range of the similarity index varied depending on the type of the submitted paper. Once published, the similarity index was not related to the value of a paper in the academic or the social aspect.

Our dataset was contaminated by well-understood flaws as the estimated similarity index value was exposed during peer review and final decision. A high similarity index might incline the reviewer and the editor towards relinquishing publication of the paper. Furthermore, we omitted some bibliometric measurements in the analysis, which have been known to be potential candidates that influence the cited-by frequencies of scientific articles [2]. Readers must be aware of the possible expansion of the role of the similarity index in explaining the value of a submitted paper. It should be noted that our finding indicated a correlation rather than causality. Despite the inherent weakness in our dataset and the analysis we have chosen, the result would be worth examining in detail because the dataset registered the full submitted papers since 2012. This is the first published report showing the relationship between the similarity index and the value of a scientific paper.

The highest cutoff value for acceptance of clinical investigation rose to 38% compared to 20% during the years before 2014, which also supported our editorial policy that peer review and a decision should be made independent of the similarity index, as mentioned above.

We adopted the cited-by measures as the academic value of a published paper, and then we obtained them from 3 sources. The cited-by measures showed consistent behavior in the principal component analysis, which concluded that the measures from 3 different sources were needless in a research of this nature. The cited-by measures should be interpreted with caution. It must grow with age of the published paper. In the Table 3, the publication year must indicate the inverse age of the paper, which dominates the negative influence on the cited-by measures and the social measure in the first component. Papers published in the year 2015 were enrolled in the analysis, which had a low chance to be cited or to appear in social media likewise. The similarity index showed up twice in the analysis of published papers, being the first in an inverse direction to the publication period and in the same direction to the value of a paper. This suggests that “old papers had a low similarity index and had a high value,” in plain English. Second, the direction to the publication period was the same, which may be interpreted as a “time-dependent increase in the similarity index of a published paper.” The latter might reflect a gradual change in our editor's standpoint along with the editorial policy that has educated peer reviewers and editors not to reject a paper simply because of a high similarity index. Taken together, we assume that the similarity index varied according to the publication period rather than according to the value of a paper. This is why the first component was named the ‘age effect,’ and the third component was named the ‘education effect.’ Independent of these 2 effects, three social media measures accounted for 33% of total variances.

The Altmetric.com measures account for a wide spectrum of values following the social appearance; composed of social media shares, blog posts, Wikipedia citations, and news stories.¹⁾ The KJA has a long history but it is relatively new as an international journal; therefore, appearances in social media are quite rare. Only 179 of the 420 papers appeared in social media as compared with 254 of the 420 papers cited at least once. This discrepancy was also evident on examining the differences in loadings of the first component.

As the academic citation and social appearance shared a unit measurement, count, they were meant to be interpreted by comparison of the absolute values of loadings. The age of an article contributed more to the growth of the academic value than to the growth of the social value (0.5 vs. 0.1). Low contribution of the similarity index, regarding the social value, limited aggressive interpretation. For these reasons, our research did not treat the temporal schema of Tweets (a message sent using Twitter), which can predict high academic citation in an article that appeared on Tweets within the first 3 days of publication [3].

The discovery was made after a cross-sectional investigation of a single medical journal. The finding suggested that the similarity index no longer had an influence on academic citation or social media appearance according to the paper after publication, while the similarity index affected the acceptance/rejection of a submitted paper. Phrases written by authors belonging to a discipline must share a similarity, especially in the case of a non-English native author [4]. Proofreading and intervention to reduce the similarity for finalizing a draft by the editors might play a role in achieving uniform quality of the publication.

Editors are asked to provide a free and transparent exchange of scientific findings, essentially when these findings are related to the article-selection process. With our novel approach to feature the similarity index, medical journal editors will improve their instinct and expand their knowledge about how the similarity index works in decision making for publication and how it propagates beyond its publication. Based on our finding, while the similarity index influences the destiny of a submitted paper in a complicated way depending upon the article type, once the paper is published, the similarity index lacks influence over the academic citation and appearance in social media. Although readers should be aware that our finding was based on a single international journal published by a medical society in Korea, it necessarily concurs with a gleam in an average editor's eye.