Scientific integrity and fraud in radiology research

Abstract

PURPOSE: To investigate the view of radiologists on the integrity of their own and their colleagues' scientific work. MATERIALS AND METHODS: Corresponding authors of articles that were published in 12 general radiology journals in 2021 were invited to participate in a survey on scientific integrity. RESULTS: A total of 219 (6.2 %) of 3,511 invited corresponding authors participated. Thirteen (5.9 %) respondents reported having committed scientific fraud, and 60 (27.4 %) witnessed or suspect scientific fraud among their departmental members in the past 5 years. Misleading reporting (32.2 %), duplicate/redundant publication (26.3 %), plagiarism (15.3 %), and data manipulation/falsification (13.6 %) were the most commonly reported types of scientific fraud. Publication bias exists according to 184 (84.5 %) respondents, and 89 (40.6 %) respondents had honorary authors on their publications in the past 5 years. General confidence in the integrity of scientific publications ranged between 2 and 10 (median: 8) on a 0-10 point scale. Common topics of interest and concern among respondents were authorship criteria and assignments, perverse incentives (including the influence of money, funding, and academic promotions on the practice of research), and poorly performed research without intentional fraud. CONCLUSION: Radiology researchers reported that scientific fraud and other undesirable practices such as publication bias and honorary authorship are relatively common. Their general confidence in the scientific integrity of published work was relatively high, but far from perfect. These data may trigger stakeholders in the radiology community to place scientific integrity higher on the agenda, and to initiate cultural and policy reforms to remove perverse research incentives.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 17 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.