Detecting anomalous referencing patterns in PubMed papers suggestive of author-centric reference list manipulation

Area of Science:

• Bibliometrics
• Scholarly Publishing
• Research Integrity

Background:

• Citations are key metrics for academic influence.
• Reference list manipulation (RLM) can artificially inflate perceived research impact.
• The prevalence and extent of RLM among authors remain unknown.

Purpose of the Study:

• To quantify the degree to which authors engage in reference list manipulation.
• To identify potential indicators and statistical anomalies associated with RLM.
• To estimate the proportion of highly published authors potentially involved in RLM.

Main Methods:

• Analyzed non-self-citations (NSC) from a subset of active, highly published authors (n=20,803) in PubMed.
• Applied anomaly detection methods to identify deviations from expected citation patterns (Zipf's law).
• Correlated citation anomalies with the Non-Self-Citation Gini Index and self-citation rates.

Main Results:

• Deviations from expected citation frequencies approximate a normal distribution.
• Authors engaged in RLM are likely to exhibit statistical anomalies.
• Between 81-231 authors (0.4%-1.1%) showed significant anomalies suggestive of chronic RLM.
• Approximately 16% of authors may have engaged in RLM to some extent.
• High self-citation rates (≥18%) correlate with Non-Self-Citation Gini Index distortions.

Conclusions:

• Statistical anomaly detection can identify potential cases of reference list manipulation.
• A significant proportion of authors may engage in RLM, impacting citation metrics.
• High self-citation may indicate a predisposition towards coercive citation practices.