Publication misconduct and plagiarism retractions: a systematic, retrospective study.

Objectives: To investigate whether plagiarism is more prevalent in publications retracted from the medical literature when first authors are affiliated with lower-income countries versus higher-income countries. Secondary objectives included investigating other factors associated with plagiarism (e.g., national language of the first author's country affiliation, publication type, journal ranking).

Design: Systematic, controlled, retrospective, bibliometric study.

Data Source: Retracted publications dataset in MEDLINE (search filters: English, human, January 1966-February 2008).

Data Selection: Retracted misconduct publications were classified according to the first author's country affiliation, country income level, and country national language, publication type, and ranking of the publishing journal. Standardised definitions and data collection tools were used; data were analysed (odds ratio [OR], 95% confidence limits [CL], chi-squared tests) by an independent academic statistician.

Results: Of the 213 retracted misconduct publications, 41.8% (89/213) were retracted for plagiarism, 52.1% (111/213) for falsification/fabrication, 2.3% (5/213) for author disputes, 2.3% (5/213) for ethical issues, and 1.4% (3/213) for unknown reasons. The OR (95% CL) of plagiarism retractions (other misconduct retractions as reference) were higher (P < 0.001) for first authors affiliated with lower-income versus higher-income countries (15.4 [4.5, 52.9]) and with non-English versus English national language countries (3.2 [1.8, 5.7]), for non-original research versus original research publications (8.4 [3.3, 21.3]), for case reports and series versus other original research types (4.2 [1.4, 13.0]), and for publications in low-ranked versus high-ranked journals (4.9 [2.4, 9.9]). Up until 2012, there were significantly (P < 0.007) fewer 'serial offenders' (first authors with >1 retraction) with publications retracted for plagiarism (11.5%, 9/78) than other types of misconduct (28.9%, 24/83).

Conclusions: This is the first study to demonstrate that publications retracted for plagiarism are significantly associated with first authors affiliated with lower-income countries. These findings have implications for developing appropriate evidence-based strategies and allocation of resources to help mitigate plagiarism misconduct.