Research with Integrity – Destigmatising the correction of scientific mistakes

Over the last decade awareness of problems with reproducibility, questionable research practices and fraudulent data in scientific papers has grown.

This has been in part due to the efforts of a growing community of integrity sleuths, pioneered by Elisabeth Bik, and platforms such as PubPeer that enable review and discussion of papers post-publication, as well as the Retraction Watch database and blog that report on retracted papers. Identification of problems in papers post-publication has also been aided by forensic software tools that detect text plagiarism, such as iThenticate, or image anomalies and duplications, for example Imagetwin, and most recently by the Collection of Open Science Integrity Guides. This has led to universities and research institutions conducting research misconduct investigations and to journals introducing procedures to correct or retract papers.

And rightly so, deliberate research misconduct should be taken seriously and robust policies on investigating breaches of research integrity should be adhered to. However, it has been estimated that nearly 40% of retractions are due to errors rather than fraudulent behaviour. So, despite pre-publication checks by authors, integrity specialists, reviewers and journal editors, mistakes, unfortunately, do make it into published papers.

Encourage correction

If we want to correct the scientific record, we need to recognise that mistakes happen and treat them differently from deliberate breaches of research integrity (see box below for the Concordat to Support Research Integrity’s definition of research misconduct). At the CRUK Scotland Institute we have been encouraging our researchers to admit mistakes in their research, and supporting them, where possible, to resolve them.

As well as assessing identified or anonymous concerns about problems in papers, we encourage our researchers to be open and pro-active about self-reporting as well as raising concerns about other researchers. We believe that correcting mistakes in scientific papers should be a normal part of the research process and that this should be done without blame. Too often the focus has been on identifying a researcher at fault, rather than identifying why and how mistakes have occurred and how they can be resolved.

We have developed an objective and transparent evidence-based process to assess all potential issues with our publications, and where mistakes are identified we endeavour to resolve them and do this without retribution. To reinforce this, support for researchers is signposted in multiple places, including during the Institute’s mandatory research integrity training programme, our manuscript pre-submission research integrity review, and in our inhouse research integrity toolkit.

We assess the validity and extent of the problem in published papers, and determine who, when and where the problematic data were generated. In 2016, we introduced a data management policy that mandates archiving a copy of published data in our in-house data repository, which enables the original source data to be assessed. If review of the published figures and the raw data indicates that a mistake was made, we notify journal editors and work with them to correct the paper with replacement figures created from the original raw data. For historical papers, where data are not available, this is not always possible and may lead to an editorial expression of concern or a retraction.

Our aim is to contribute to the integrity of the scientific record and to demonstrate the benefits of correcting mistakes to our researchers. Thereby creating a research culture where this is normal practice. A vital part of this process involves interacting with our researchers in a non-accusatory and non-judgemental manner. By being objective but supportive we have developed a framework that enables open discussion of mistakes and a process to resolve them. To ensure that lessons are learned we try to understand the cause of mistakes so that we can revise our research integrity policies, research integrity training and research processes and introduce preventative measures.

Taking responsibility

Of course, if our assessment of the error in a published paper and review of the raw data indicate that the problem is the result of potential research misconduct, then our policy on Misconduct in Research is implemented.

Our approach resonates with the work the UK Research Integrity Office is doing to support institutions in normalising reporting of concerns about research integrity and the Concordat to Support Research Integrity’s recognition that “errors committed in good faith and honest mistakes are seen as productive elements of research” and that employers “create the conditions necessary for honest errors to be openly admitted without undue detriment”.

So, what would you do if you found or were alerted to an error in one of your published papers? This question was posed to scientists in an anonymous poll by The Times Higher Education. Whilst the majority of respondents said they would report a mistake to the journal, 5% said they would do nothing, even if the error was serious and their paper was in a high-impact journal. The percentage of people who would contact a journal increased slightly for serious errors in run-of the-mill journals and rose to just over 20% for minor errors that didn’t affect the paper’s conclusions.

Owning up to mistakes can be hard as you might be embarrassed, fear that your reputation will be damaged or that mistakes will be viewed as research misconduct. I strongly urge you to reconsider. Take responsibility and be part of the community who want to reduce errors in research and enhance data reproducibility and trustworthiness. Hiding or ignoring mistakes is a questionable research practice that could lead to future work based on incorrect data, ultimately wasting time and money and potentially result in research misconduct.

By sharing experiences and being open you can show others the merit of addressing mistakes and correcting papers, as well as improving the reliability of research data.