JOURNAL CLUB: Plagiarism in Manuscripts Submitted to the AJR: Development of an Optimal Screening Algorithm and Management Pathways.

OBJECTIVE: The objective of this study was to investigate the incidence of plagiarism in a sample of manuscripts submitted to the AJR using CrossCheck, develop an algorithm to identify significant plagiarism, and formulate management pathways.

MATERIALS AND METHODS: A sample of 110 of 1610 (6.8%) manuscripts submitted to AJR in 2014 in the categories of Original Research or Review were analyzed using CrossCheck and manual assessment. The overall similarity index (OSI), highest similarity score from a single source, whether duplication was from single or multiple origins, journal section, and presence or absence of referencing the source were recorded. The criteria outlined by the International Committee of Medical Journal Editors were the reference standard for identifying manuscripts containing plagiarism. Statistical analysis was used to develop a screening algorithm to maximize sensitivity and specificity for the detection of plagiarism. Criteria for defining the severity of plagiarism and management pathways based on the severity of the plagiarism were determined.

RESULTS: Twelve manuscripts (10.9%) contained plagiarism. Nine had an OSI excluding quotations and references of less than 20%. In seven, the highest similarity score from a single source was less than 10%. The highest similarity score from a single source was the work of the same author or authors in nine. Common sections for duplication were the Materials and Methods, Discussion, and abstract. Referencing the original source was lacking in 11. Plagiarism was undetected at submission in five of these 12 articles; two had been accepted for publication. The most effective screening algorithm was to average the OSI including quotations and references and the highest similarity score from a single source and to submit manuscripts with an average value of more than 12% for further review.

CONCLUSION: The current methods for detecting plagiarism are suboptimal. A new screening algorithm is proposed.