Handing the Microphone to Women: Changes in Gender Representation in Editorial Contributions Across Medical and Health Journals 2008-2018

The editorial materials in top medical and public health journals are opportunities for experts to offer thoughts that might influence the trajectory of the field. To date, while some studies have examined gender bias in the publication of editorial materials in medical journals, none have studied public health journals. In this perspective, we studied the gender ratio of the editorial materials published in the top health and medical sciences journals between 2008 and early 2018 to test whether gender bias exists. We studied a total of 59 top journals in health and medical sciences. Overall, while there is a trend of increasing proportion of female first authors, there is still a greater proportion of male than female first authors. The average male-to-female first author ratio during the study period across all journals was 2.08. Ensuring equal access and exposure through journal editorials is a critical step, albeit only one step of a longer journey, towards gender balance in health and medical sciences research. Editors of top journals have a key role to play in pushing the fields towards more balanced gender equality, and we strongly urge editors to rethink the strategies for inviting authors for editorial materials.


Supplementary file 1 Methods
We considered two sets of top journals in health and medical sciences. The first set comes from Google Scholar's two sets of rankings of top 20 health and medical sciences journals in April 2018, of which 32 unique journal names were present and considered for the analysis (Table S1). The second set comes from four sets of rankings of the Web of Science's InCites Journal Report Citation website, including 1) medicine (general and internal), 2) public, environmental, and occupational health, 3) healthcare science and services, and 4) health policy and services. These four sets of rankings included 47 unique journal names, all of which were included in the analysis.
For all 79 journals, we downloaded the metadata on all editorial materials published between 2008 to April 2018 from Thomson Reuters' Web of Science database. Journals not listed in Web of Science, as well as journals with less than 100 editorial materials during the study period were excluded, the latter due to small numbers of editorials published each year, making the estimation of the ratios unreliable. Editorial materials are listed under different sections in each journal, such as "Comment" in the Lancet, "Commentary" in the New England Journal of Medicine, and "Opinion" and "Viewpoint" in JAMA. To ensure consistency, we searched for articles classified under "editorial materials" under Web of Science, which editorial materials to include editorials, interviews, commentary, and discussions between individual, post-paper discussions, round table symposia, and clinical conferences, 35 but not letters or correspondences from the readers to the journal editor concerning previously published articles. To ensure that the metadata includes primarily editorial materials and not clinical case reports (papers that describe clinical cases that may be of interest to the clinical community), 36 we first downloaded the metadata of case reports from the Web of Science and removed listed articles from our data. 52 of the 79 journals matched our inclusion criteria and were included in this analysis, listed in Table S1.
For each journal, we first assigned a country of origin based on information provided on the journals' or publishers' websites. For many, the websites did not specify the country of origin, so we used the country in which the publisher is based instead (see Table S1). We determined whether the journal is a medical, public health, life sciences, or healthcare/other journal based on the journal's own description or Wikipedia, which specifies whether the journal is a "medical," "scientific," "public health," or "healthcare" journal. We then collected the list of all editors from the journals' respective websites. We extracted the names of editors in chief and all editors who likely have decisive functions regarding manuscript acceptance. This includes senior (executive) editors, acting chief editor, executive editors, deputy editors, associate editors, regional editors, research editors, (senior) managing editors, scientific editors, and section-specific editors (such as healthcare policy and law editors, news editors, perspective editors, letters editors, editorials editors). We excluded editorial staff members, such as editorial assistants, technical editors, images editors, digital content editors, assistant web editors, multimedia editors, and statistical editors, as well as honorary editors such as editors emeritus, honorary deputy editor, senior guest editor, and founding editors. Editorial materials with any of the journal editors listed as the first author were excluded (see list of journal editors in Supplementary file 1).
We extracted the first names of the first authors from all papers included in the dataset. We estimated authors' gender using the gender package in R, which relies on U.S. Census or Social Security datasets. To quantify uncertainty in our automated gender classification, we selected a random sample of 100 names from the articles, assigned gender for each based on Google search results, and assessed the performance of the gender classifier on this sample. We found 86% to be accurately classified, and 14% to remain as unclassified. This suggests that, when a gender is assigned to a name by the classifier, we can be highly confident about the result. Among papers with names that were not assigned a gender, we randomly selected 10% of papers and manually searched for the gender of the author based on the full name and the institution to which the author belongs to through Google.

Limitations
First, this study is descriptive in nature, and we cannot determine the factors that lead to the findings of higher male-to-female first authors. Second, we were unable to find the exact processes in which journal editors determine the content of the editorials: who decides the topics of the editorials, whom to invite, and which editorials to publish. The instructions listed on most journals' websites state that most editorials are invitation-only or not, but do not discuss which route is more common. Third, we only looked at the years 2008-2018 and not earlier years because the metadata from Web of Science did not include full first-names in prior years. We were only able to obtain the names of journal editors in April 2018 and not prior years, therefore limiting our study to recent years reduces the possibility that previous editors authored some of the editorials. Fifth, we only analyzed the gender of the first author, who we assumed is the main author of the article, instead of the senior author or all co-authors.