Evaluating Approaches for Identifying Retracted Articles and Retraction Notices in Systematic Review Searching
Bibliographic record
Abstract
Caitlin J. Bakker,1,2 Erin E. Reardon,3 Nicole Theis-Mahon,4 Sara Schroter,5,6 Lex Bouter,7,8 Maurice P. Zeegers2 Objective Systematic reviews gather, appraise, and synthesize studies to inform research, practice, and policy. However, the inclusion of retracted articles, which often contain flaws and falsified or fabricated data, undermines the credibility of systematic reviews. Identifying retracted articles is challenging, as they are inconsistently flagged.1,2 Our study validated and compared approaches to identify retracted articles and their retraction notices. Design Our study, guided by an advisory panel of information specialists and researchers, evaluated approaches for identifying retracted publications from 8 health sciences databases (Cochrane Library, Embase.com, Ovid Embase, Ovid Medline, Ovid PsycINFO, PubMed, Scopus, and Web of Science). Using a reference set of 43,544 retracted publications and 27,755 associated retraction notices from Retraction Watch, we identified items found in each database. From August 10 to 14, 2024, we determined how many of the available items could be retrieved using each approach per database. Two search strategies, database indexing, and 2 citation managers were evaluated. The complete methodology, including search strategies, is available in our protocol.3 Recall (sensitivity) was calculated to evaluate identification effectiveness. Results Recall of retracted publications and notices varied across databases and retrieval approaches. Across databases, search strategy 2 consistently achieved the highest recall, with values ranging from 74.6% to 96.9%. Search strategy 1 also performed strongly, particularly in PubMed and Web of Science (both >93%). In contrast, indexing-based retrieval showed variable performance, with high recall in PubMed (94.6%) and Ovid Medline (94.5%) but much lower in Embase.com (40.6%) and PsycINFO (34%). Citation manager tools (EndNote and Zotero) yielded lower recall, with values rarely exceeding 64%. Recall was lowest for Ovid Embase and PsycINFO regardless of method, while PubMed and Web of Science showed the highest recall. https://assets.underline.io/markdown_image/1/image/b3957ae8a92f24f08736a44da92e7d56.png Conclusions There was substantial variability in the ability of databases and retrieval approaches to identify retracted publications and notices. No single approach captured all items, underscoring the need for multiple approaches in an iterative identification process. References 1. Bakker CJ, Reardon EE, Brown SJ, et al. Identification of retracted publications and completeness of retraction notices in public health. J Clin Epidemiol. 2024;173:111427. doi:10.1016/j.jclinepi.2024.111427 2. Boudry C, Howard K, Mouriaux F. Poor visibility of retracted articles: a problem that should no longer be ignored. BMJ. 2023;381:e072929. doi:10.1136/bmj-2022-072929 3. Bakker C, Reardon EE, Theis-Mahon NR, Schroter S, Bouter L, Zeegers M. Validation and comparison of methods to identify retracted publications during information retrieval. Open Science Framework. Cited June 7, 2025. https://osf.io/rwzym/ 1University of Regina, Regina, SK, Canada, caitlin.bakker@uregina.ca; 2Maastricht University, Maastricht, the Netherlands; 3Emory University, Atlanta, GA, US; 4University of Minnesota, Minneapolis, US; 5BMJ, London, UK; 6London School of Hygiene and Tropical Medicine, London, UK; 7Amsterdam University Medical Center, Amsterdam, the Netherlands; 8Vrije Universiteit Amsterdam, Amsterdam, the Netherlands. Conflict of Interest Disclosures Caitlin J. Bakker is cochair of the National Information Standards Organization Communication of Retractions, Removals and Expressions of Concern Standing Committee. No other disclosures were reported. Funding/Support This research is part of an ongoing PhD collaboration between BMJ and the team Meta-Research at Maastricht University (UM) on the responsible conduct of publishing scientific research. BMJ is published by BMJ Group, a wholly owned subsidiary of the British Medical Association. UM is a public legal entity in the Netherlands. This study is part of Caitlin J. Bakker’s self-funded BMJ/UM PhD. No exchange of funds has taken place for this research project. Role of the Funder/Sponsor The authors are wholly responsible for the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the abstract; and decision to submit the abstract for presentation. Additional Information All authors express their own opinions and not necessarily that of their employers.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.795 | 0.947 |
| Meta-epidemiology (narrow) | 0.006 | 0.008 |
| Meta-epidemiology (broad) | 0.015 | 0.023 |
| Bibliometrics | 0.111 | 0.083 |
| Science and technology studies | 0.010 | 0.008 |
| Scholarly communication | 0.022 | 0.030 |
| Open science | 0.017 | 0.030 |
| Research integrity | 0.009 | 0.007 |
| Insufficient payload (model declined to judge) | 0.021 | 0.005 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".