Fake publications in biomedical science: red-flagging method indicates mass production

Area of Science:

• Bibliometrics and scientific publishing integrity.
• Biomedical research and academic fraud detection.

Background:

• Academic publishing integrity is threatened by "paper mills" producing fake publications using AI.
• The scale of fake publications in biomedicine is largely unknown, necessitating reliable detection methods.

Purpose of the Study:

• To develop and validate an easy-to-apply rule for red-flagging potentially fake biomedical publications.
• To estimate the current scale and growth of fake publications in the biomedical literature.

Main Methods:

• A 9-step bibliometric analysis of 17,120 PubMed® publications.
• Development and validation of simple classification rules, including a multifactorial tallying rule.
• Application of the best-performing rule to random samples from 2020 and 2023, followed by Bayes' rule for estimation.

Main Results:

• The multifactorial tallying rule achieved 94% sensitivity and an 11.5% false-alarm rate.
• Red-flagged articles increased, reaching an estimated 16.3% in 2023.
• An estimated 5.8% of biomedical literature consists of actual fake publications, with over 100,000 articles annually.

Conclusions:

• Fast-and-frugal classification rules can effectively red-flag potentially fake scientific publications for scrutiny.
• The scale of fake publishing in biomedicine is substantial, significantly exceeding retractions and posing risks to public trust and health.
• Developing effective fake detection tools is crucial for maintaining the integrity of the scientific record and preventing further damage.