A medical journal has retracted a 2016 paper over a series of errors, prompting it to lose faith in the paper overall. The authors have objected to the decision, arguing the errors weren’t their fault and could be revised with a correction — rather than retracting what they consider “an important contribution” to an ongoing debate in medicine.
The paper explored the so-called weekend effect—that patients admitted to the emergency department on the weekend are more likely to die than those admitted on a weekday. Whether the weekend effect is real is not clear. Some studies have supported the phenomenon in certain areas of medicine, but others (including the now-retracted paper) have failed to find an effect.
First author Mohammed A. Mohammed, based at the University of Bradford in the UK, told Retraction Watch that the errors were introduced by one of the hospitals that provided them the data:
We were surprised the editor took the view that the paper merited retraction. We analyzed the data the hospitals provided in good faith but one hospital got its data mixed up.
According to emails between the authors and journal, Paul Aylin, a professor of epidemiology and public health at Imperial College London, was the first to point out an error in the paper. After the authors corrected the initial error in August 2016, the Quarterly Journal of Medicine (QJM) said it planned to publish an erratum.First, however, the journal sent the revised paper to Aylin, who told us that the revised number of hospital admissions didn’t add up; he thought there might be a second data error.
After confirming a second error, the authors revised the analysis again and notified the journal. However, at this point, the journal began to doubt the reliability of the analysis. In November 2016, Seamas Donnelly, the editor-in-chief, asked the authors to retract the paper.
Mohammed responded several days later, asking Donnelly to reconsider:
Given … the potential importance of this paper to the scientific debate we would not wish to retract the paper and ask you to reconsider your initial response, or discuss with us what further reassurance you need as to its scientific integrity.
In his email to Donnelly, Mohammed added:
The existence and size of a weekend effect is hotly contested in the scientific community, and the policy and public domains. Withdrawal of the paper, because of the need to make these two corrections, means that an important contribution to the scientific debate will now not be available and indeed potentially discredited … Retraction may also bias the literature as our findings using a new method, contradict the established majority position. We feel that this would do a disservice to the public debate.
Aylin, who has published research about the “weekend effect” that produced opposite results to what’s reported in the now-retracted paper, told us he believes “the authors should have caught the errors,” casting doubt on the reliability of the work:
There were internal inconsistencies in their data that should also have prompted them to question their results.
Despite the authors’ objections, the journal decided to move forward with the retraction. Here’s the notice for “Adjusting for illness severity shows there is no difference in patient mortality at weekends or weekdays for emergency medical admissions,” first published in July 2016 and retracted last July:
After publication, errors in the data were identified, leading the authors to amend the data and analysis. Subsequently, further errors in the data were identified, leading the authors to amend the data and analysis once more. Although these were cases of honest error and the authors assert that the conclusions are unaffected, sufficient doubt on the part of the QJM remains as to the reliability of the data and its analysis to require the article to be retracted.
The paper has not yet been indexed by Clarivate Analytics’ Web of Science.
We contacted QJM about its decision to retract the paper, but the journal did not respond.
Finding a new home
This past September, the authors found a new home for their revised analysis in the Journal of Health Services Research & Policy. The paper includes a footnote, which explains:
…an earlier analysis of these data was reported in a publication in the Quarterly Journal of Medicine but was subsequently withdrawn.
Nick Black, the editor who handled the new submission, told us that the authors had been completely transparent with him about what happened with the retraction and the journal put the revised submission through its standard peer review process:
We were probably even tougher than normal during peer review, selecting particularly rigorous reviewers and as editors, we read it carefully and discussed it quite a bit.
Faced with a similar scenario, in which authors used data from a government database that contained errors, JAMA Internal Medicine retracted the initial analysis and replaced it with a corrected version, an approach JAMA has used to fix papers with honest, but pervasive, errors.
Mohammed told us:
We were grateful that the mistakes were discovered, but retraction seemed like an extreme position for the journal to take over an honest data error.
Update, 1000 UTC, 2/19/18: QJM‘s editor-in-chief Seamas Donnelly got back to us and explained that the journal chose to retract, not correct, the paper because:
…the two errors that were identified raised significant doubt about the integrity of the data presented in the paper.
Like Retraction Watch? You can make a tax-deductible contribution to support our growth, follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up for an email every time there’s a new post (look for the “Follow” icon at the bottom right of your page), or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].