How should medical journals deal with errors?

Errors in the articles published in medical journals are common and longstanding. As far back as 1950, librarians were debating how to correct errors in medical journals. ¹ We have seen considerable progress in this area in recent decades and most journals now have policies to correct errors, as well as for dealing with fraud and scientific misconduct, in research articles. However, errors in other types of articles – such as editorials, commentaries and reviews – often remain uncorrected, even after these have been bought to the attention of the journals' editors or the articles' authors.

That errors occur in published articles is not surprising. Errors occur even in research articles that undergo extensive pre-publication peer review. Articles that undergo less rigorous peer review – such as commissioned editorials – will therefore also contain errors. What is needed from journals are rigorous and consistent policies for dealing with such errors when these are detected. This is not just an academic point. Statements and statistics in leading medical journals are likely to be referenced in future articles, and could also influence clinical practice and health policy. We therefore need to ensure that these facts are as accurate as possible and corrected quickly if inaccurate.

In Box 1, I've listed some statements in journal articles that are either wrong or that are highly unlikely to be true but which remain uncorrected. The journal used to select these examples, the BMJ, is probably no worse in this respect than other medical journals; and may in fact be better than many other journals in detecting and correcting errors because of its policy encouraging the submission of rapid responses to the articles it publishes. It is though a journal that prides itself in leading the way in developing publication ethics and policies, and for publishing research on the adequacy of the peer-review process. If we see errors in the BMJ, they may be even more common in smaller journals that have more limited resources for reviewing and checking articles.

Box 1

Selection of errors from the BMJ

The reported incidence [of polycystic ovary syndrome] varies between 3% and 15% of women of reproductive age. ⁹

An annual incidence of 15% would imply that after 10 years, a woman of reproductive age would have around an 80% risk of developing polycystic ovary syndrome. Even the 3% incidence figure would result in around a 26% risk of developing the syndrome after 10 years. What the author is probably referring to is ‘prevalence’, which is an entirely different concept to incidence.

Study finds hepatitis C infections more prevalent than previously thought. ¹⁰

This news item reported a small pilot study that had a high degree of selection bias in the people who were tested for hepatitis C, thus leading to a much higher rate of hepatitis C infection than reported in the larger and more representative surveys carried out by the Health Protection Agency.

In any one year one in four people in the United Kingdom have their thyroid function checked. ¹¹

Two references were given to support this statement. One contained no data on the frequency of thyroid function testing; the other reported that around 10 million thyroid function tests are carried out annually but did not report the source of this statistic. Even if we accept this statistic as accurate, this equates to about one test per six people in the UK, not one per four, assuming that the tests are independent (i.e. each test is carried out on only one patient). Some people will have multiple tests and hence the actual proportion of the population undergoing thyroid function testing annually will be less than one in six, and very different to the authors' quoted figure of one in four.

In 2004 a 1000 point plan called the quality and outcomes framework (QOF) was introduced. ¹²

The QOF contained 1050 points when it was introduced in 2004; the number of points was not reduced to 1000 until 2006.

Medical journals have made great strides in recent decades in improving the accuracy of their articles, but the focus has been very much on improving the quality of the research articles they publish. This quality improvement work has included, for example, the development of pre-publications checklists and statistical guidance to improve the accuracy of reporting of the methods and results sections in papers; ² and guidance on disclosing conflicts of interest, particularly those of a financial nature. ³ Groups such as the Committee on Publication Ethics (COPE) have also developed guidance for journals, but these have been focused on dealing with fraud and scientific misconduct, and policies for dealing with the retraction of papers. ⁴ There is less guidance from these bodies on how to deal with unintentional but more common errors.

Can medical journals learn from other fields of publishing in improving the accuracy of their articles and dealing with errors? There are lessons that can be learned from newspapers and guidelines for the publication of government statistics. Some newspapers have appointed Readers' Editors to help correct significant errors quickly. This includes The Guardian newspaper in the UK, for example, whose readers' editor states that his role is to ‘collect, consider, investigate, respond to, and where appropriate come to a conclusion about readers’ comments, concerns, and complaints in a prompt and timely manner, from a position of independence within the paper'. ⁵ As part of this process, The Guardian has a section on its website where corrections and clarifications are published so that these are clearly visible to its readers. The publication of government statistics in the UK is regulated by a code of practice for official statistics that includes a commitment to correct any errors discovered in statistical reports and notify relevant people promptly. ⁶

With increasing recognition of the limitations of pre-publication peer review, there have been calls to rely more on post-publication review of articles. ⁷ However, such post-publication review can only be effective if journals have policies in place to ensure that errors in articles are identified and corrections published. Although some journals do claim to have policies that encourage authors to respond to significant criticisms of their articles, such policies are currently applied haphazardly, and journals do not appear to have any sanctions that they can use to make authors respond more appropriately to potential errors and misleading statements. The establishment of the role of a readers' editor, the creation of a clarification and corrections page on their website, ⁸ and the publication of a commitment to correct errors quickly are all initiatives that could be adopted rapidly by the larger medical journals. These journals play a key role in the dissemination of medical knowledge, the development of clinical practice, and the shaping public health and health service policies. As part of this role, journals should commit fully to post-publication review of their articles and not leave potentially inaccurate or misleading statements in these articles uncorrected.

DECLARATIONS

Competing interests

None declared

Funding

AM's academic post is funded by the Higher Education Funding Council for England; no separate funding was received for this article

Ethical approval

Not applicable

Guarantor

AM

Contributorship

AM is the sole contributor

Acknowledgements

The Department of Primary Care & Public Health at Imperial College London is grateful for support from the NIHR Collaboration for Leadership in Applied Health Research & Care (CLAHRC) scheme, the NIHR Biomedical Research Centre scheme, and the Imperial Centre for Patient Safety and Service Quality