Abstract
Experience With 12 Years of Plagiarism and Duplication Screening
Markus K. Heinemann,1 Andreas Boening,2 Kazunori Okabe,3 Jessica Bogensberger4
Objective
In 2013, a cardiothoracic surgical journal introduced plagiarism detection (Crossref, powered by iThenticate). After a phase of 2 years during which all submissions were screened, since January 2015 only returned manuscripts after revision were checked by default. Primary submissions were then screened at the discretion of the Editor in Chief (EiC) only. The aim of this study was to report the 12-year experience.
Design
All manuscripts submitted between January 1, 2013, and December 31, 2024, and branded by the plagiarism search engine with a similarity index above 33% were analyzed regarding the originator of the screening (EiC or revision default), reasons (actual plagiarism or self-plagiarism, which is a form of duplication), similarity percentage, and final decision. Descriptive statistics were applied. All updates of the search tool were implemented.
Results
During the investigation period, 4083 manuscripts were submitted. A total of 99 (2.4%) showed a similarity index greater than 33%. The mean percentage was 53.3% (minimum, 34%; maximum, 99%; median, 50%; SD, ±14.8). Plagiarism and duplication search was prompted by the EiC in 48 papers and by automated default in 51 papers. Self-plagiarized or duplicated content was by the same authors in 71 of 99 papers (71.7%), by different authors in 18 (18.2%), and both in 10 (10.1%). Fifty-five manuscripts (55.6%) remained unpublished: 17 rejected without review, 34 rejected, and 4 withdrawn. Forty-four of 99 (44.4%) were accepted after revisions. There were 4 attempts at duplicate publication, all rejected. Preprints were discovered in 3 cases, with 1 accepted after revision: the preprint had been uploaded by a different journal for review and was rejected there, but then not deleted. The 44 manuscripts finally accepted contained previously published sequences by the same authors in 36, by different authors in 4, and both in 4. In 25 cases, the EiC rendered the similarity uncritical (eg, materials and methods section). The most frequent necessary correction was adequate citation. No adequate citation was given for self-plagiarized/duplicated content in 23 cases and for content by different authors in 4.
Conclusions
Although the overall rate of plagiarism and duplication can be considered low, scientific misconduct was detected and its publication prevented in a few cases. In many instances, duplicated passages contained similar material and methods previously published by the same group. Deficiencies in adequate citation are a problem and necessitate continued education of authors. Selective plagiarism and duplication screening remains an important tool, although chosen thresholds of similarity remain arbitrary. Further improvements of sensitivity are expected, especially with the implementation of machine learning advances. Editorial decisions, however, cannot be made on automated findings alone, but must be put into personal perspective.
1German Society for Thoracic and Cardiovascular Surgery (DGTHG), Germany, heinemann@uni-mainz.de; 2Universitaetsklinik Giessen, Germany; 3Bell Land General Hospital, Osaka, Japan; 4Thieme Publishers, Stuttgart, Germany.
Conflict of Interest Disclosures
Jessica Bogensberger is an employee of Thieme Publishers, the publishing house of the journal investigated. All other authors have no conflict of interest to declare.