Similarity Analysis of Korean Medical Literature and Its Association with Efforts to Improve Research and Publication Ethics

ABSTRACT

In the present study, the frequency of research misconduct in Korean medical papers was analyzed using the similarity check software iThenticate®. All Korean papers written in English that were published in 2009 and 2014 in KoreaMed Synapse were identified. In total, 23,848 papers were extracted. 4,050 original articles of them were randomly selected for similarity analysis. The average Similarity Index of the 4,050 papers decreased over time, particularly in 2013: in 2009 and 2014, it was 10.15% and 5.62%, respectively. And 357 (8.8%) had a Similarity Index of ≥ 20%. Authors considered a Similarity Index of ≥ 20% as suspected research misconduct. It was found that iThenticate® cannot functionally process citations without double quotation marks. Papers with a Similarity Index of ≥ 20% were thus individually checked for detecting such text-matching errors to accurately identify papers with suspected research misconduct. After correcting text-matching errors, 142 (3.5% of the 4,050 papers) were suspected of research misconduct. The annual frequency of these papers decreased over time, particularly in 2013: in 2009 and 2014, it was 5.2% and 1.7%, respectively. The decrease was associated with the introduction of CrossCheck by KoreaMed and the frequent use of similarity check software. The majority (81%) had Similarity Indices between 20% and 40%. The fact suggested that low Similarity index does not necessarily mean low possibility of research misconduct. It should be noted that, although iThenticate® provides a fundamental basis for detecting research misconduct, the final judgment should be made by experts.