The study included the six general biomedical journals with the highest impact factors in 2008 according to the Institute for Scientific Information’s Journal Citation Report
: Annals of Internal Medicine
, Nature Medicine
, New England Journal of Medicine
, and PLoS Medicine
In 2008 these journals followed ICMJE guidelines for authorship (see web extra table 1).
As in the 1996 study, we included three types of articles from these journals: original research reports (research); reviews, including meta-analyses (reviews); and editorials, commentaries, and other opinion articles (editorials). Before we established a sampling frame, two of us (JSW and AF) evaluated each article type by journal based on criteria established a priori. These authors reviewed three journals each, randomly checked each other’s coding, discussed any uncertainties or discrepancies, and resolved these differences by consensus. After identifying all of these article types published in 2008 in the six journals, we used a random number generator to select a stratified random sample of articles from each journal reflecting the proportion of the three types of articles published in each journal and the proportion of the total number of articles published by all six journals.
The corresponding author of each selected article was identified. If an article designated more than one corresponding author, one of these authors was randomly selected. Articles (primarily editorials) written by journal editors and staff writers and high government officials (without publicly available personal email addresses) were not eligible for inclusion and, if selected, were replaced with a randomly selected article of the same type from the same journal. If a corresponding author appeared in more than one selected article, only one of these was randomly chosen and the other article replaced with a random selection (see web extra figure on bmj.com).
The corresponding authors of sampled articles were sent an email message in April 2009 inviting them to complete an online questionnaire and explaining that participation was voluntary and that their identities and responses would be kept confidential. Up to three reminder emails were sent to non-responders at about one week intervals. No incentive or compensation was offered for participation.
We developed an online questionnaire based on the mail questionnaire used in the 1996 survey.14
The questionnaire consisted of 30 questions about the corresponding author, the contributions and functions of all authors, and writing and editing assistance and other contributions from unnamed authors and other individuals (see web extra questionnaire on bmj.com). The questionnaire was pretested among a sample of 15 authors and editors who were not included in the final study.
Using the same base definition as that used in the 1996 survey14
(based on the ICMJE authorship criteria11
), we defined an article as having an honorary author if the corresponding author reported any of the following:
- 1) An author did not meet these three criteria:
- (a) “conceiving and designing the work,” “analysing and interpreting the data,” or “collecting data or other material” (we included data collection as an acceptable criterion for authorship because the ICMJE had added this to their guidelines before we conducted this study)
- (b) “writing the manuscript or part of the manuscript” or “revising the manuscript to make important changes in content”
- (c) “approving the final version of the manuscript”
- 2) An author did not “feel comfortable explaining the major conclusions” of the article
- 3) An author performed “only one function, and nothing else” from a list of 17 activities: supervising the work of any of the coauthors; recruiting coauthors; recruiting study subjects; analysing or interpreting data; conducting literature search; analysing or interpreting literature; reviewing the manuscript; communicating with journal editor(s); signing the statement of copyright transfer to the journal; conceiving and designing the work; collecting data and other material; obtaining funding or material support; performing statistical analysis; writing the manuscript or part of the manuscript; approving the manuscript before submission to a journal; revising the manuscript or making important changes in content; reviewing page proofs or the journal’s edited version of the manuscript (see questions 8, 9, and revised question 11 in the study questionnaire on bmj.com).
Using the same definition as that used in the 1996 survey,14
we defined an article as having a ghost author if the corresponding author reported any of the following:
- 1) An individual who was not listed as an author made contributions that merited authorship
- 2) An unnamed individual participated in writing the article.
We also examined the acknowledgment and methods sections of articles meeting these criteria for indications of writing and editing assistance (see questions 13, 17, and 19 in the study questionnaire on bmj.com).
We determined the prevalences of articles with honorary authorship or ghost authorship, or both, in 2008 in these six high impact journals and compared these prevalences to those reported in the six journals in the 1996 study.14
We also compared prevalences of inappropriate authorship between 2008 and 1996 by article type (research, review, editorial). Because the earlier study was limited to US based corresponding authors and included only three of the six journals in the current study, we also conducted a secondary analysis comparing the prevalence of honorary authors and ghost authors in the three journals included in both the 1996 study and in the present study (Annals of Internal Medicine
, and New England Journal of Medicine
) using a subsample of the 2008 sample that only included US-based corresponding authors.
Four journals in this study (Annals of Internal Medicine, JAMA, Lancet, and PLoS Medicine), publicly require (that is, in their instructions for authors) authors to report their individual contributions and publish these contributions in the article. Two journals (New England Journal of Medicine and Nature Medicine) do not publicly require reporting of individual author contributions and do not routinely publish these contributions in the article. In a subanalysis we compared the prevalence of honorary and guest authors in the journals that required reporting of author contributions versus those without such requirements.
Each article served as the unit of analysis for determining the prevalence of honorary or ghost authors. We hypothesised a 10% reduction in the prevalence of articles with honorary or ghost authors from the 29% prevalence reported in the previous study of six journals in 1996.14
We estimated that 149 articles would be required to detect this difference with β of 0.20 and a two tailed α of 0.05. Assuming a 67% response rate (based on recent JAMA
author surveys), we estimated that we needed to sample at least 223 corresponding authors to ensure an adequate sample of 149 completed surveys. Using the same parameters, we estimated that 587 articles would be needed for the subanalyses comparing the four journals with and the two journals without requirements for reporting author contributions. Again assuming a 67% response rate, we estimated that we needed to sample at least 877 authors to achieve the minimum of 587. Corresponding author response rates were calculated according to American Association of Public Opinion Research (AAPOR) standard definitions.17
We used χ2
tests to compare proportions between journals and groups of journals, and to compare prevalences between the two studies (1996 and 2008). Prevalence estimates and adjusted Wald 95% confidence intervals were also calculated. Statistical analyses were computed with SPSS 14.0.18