Six years of unacknowledged citation of a retracted meta-analysis on physician burnout and patient safety: a cross-sectional citation-context study

Background: Retracted articles continue to be cited, but it is rarely clear whether citations preserve the retracted numbers or treat the article as an authority. We also do not know how often authors acknowledge the retraction or whether editorial checks reach them.Methods: We identified all works citing a 2018 meta-analysis of physician burnout and patient safety after its retraction without replacement in May 2020 (689 works; primary frame: 562 works published from 2021). Citing text came from structurally parsed PubMed Central full text and Semantic Scholar citation sentences. Outcomes were author-visible acknowledgement of the retraction, verbatim reuse of the retracted odds ratios and citation intent. We also read the professional-society statements citing it, coded a random sample of 100 citing works, and analysed comparator retractions. We used Wilson 95% confidence intervals and the Cochran-Armitage trend test.Results: Citing text was available for 266 of 562 works. Six (2.3%; 95% CI 1.0-4.8) acknowledged the retraction in any author-visible form. One (0.4%) did so in body text, in a protocol to redo the retracted synthesis. Acknowledgement did not increase over 2021-2026 (p = 0.96). In comparator retractions, acknowledgement in citing sentences or authors' own words was similarly rare (median 0% per article). Reference-list flags were more common, but mostly consisted of publisher-renamed titles copied into reference lists. No citing passage from 2021 onwards reproduced the retracted odds ratios as digits (0/259), and citation intent was overwhelmingly background. A 2024 American Heart Association scientific statement restates the retracted pooled effect ("2-fold increased odds of unsafe care") as fact, with the retracted article as its only quantitative source. Two published peer-review records show an editorial instruction to flag retracted references that was not acted on. Keyword screening of raw PubMed Central XML misclassified all 264 citing full texts as acknowledging because the archive inserts retraction markup.Conclusions: About 98% of works citing this retracted meta-analysis over six years gave no sign of the retraction, similar to other retracted articles. The article persists as an authority rather than data. One guidance document repeats the retracted result as a paraphrased number, which numeric checks miss. Retraction checking should be part of submission systems.