Gender Influences on Editorial Decisions at Epidemiology.

Abstract

Gender inequities have been documented across many fields of science, demonstrated by underrepresentation of women scientists in senior positions, citations, and funded grants.1–5 One metric that has been used to measure these inequities is authorship of peer-reviewed journal publications. The proportion of publications authored and reviewed by women scientists do not always reflect the proportion of women scientists, potentially reflecting a gender inequity that has been observed across many scientific disciplines1,6–10 and has likely grown as a result of the COVID-19 pandemic.11 Although there is evidence that access to authorship and rankings in ordered lists of authors may be unfair, there is a lack of research on gender bias in the peer-review process. Such biases may partly explain the ultimate inequity in published articles.7,12,13 In the field of epidemiology, editors and other investigators have considered gender makeup of faculty members and authorship of published articles,14,15 but not, to our knowledge, potential bias in the publishing process. The editors of Epidemiology undertook a self-study to assess whether there are differences in editorial decisions by author gender in this general interest epidemiology journal. For every original research article submitted to Epidemiology between 1 January 2015 and 31 December 2020, we extracted from the editorial database author names, institutions, and order, as well as editorial decision and time to decision. We imputed gender for 91% of authors based on first name using the Genderize.io database, which has been used in similar analyses.1,16 The probability of the assigned gender for a given name ranged from 50% to 99% for our dataset (Figure 1). For the names that were not identified by the Genderize.io algorithm, we attempted to infer gender from internet research to find gender-specific pronouns or picture confirmation using author name and identifiers such as country, credentials, and institution. We identified gender for an additional 6% of authors using this method. We calculated proportions of submissions at each stage of decision: rejected without review, rejected with review, and accepted. We also calculated the time to first decision for submissions categorized by whether the manuscript received review (ultimately accepted or rejected) or not (rejected without review). In addition, we examined requests for appeals from any author by systematic search and review of emails sent to the Editor-in-Chief that contained the keywords “reconsider” or “appeal” and pertaining to an editorial decision. We label feminine imputed or inferred names as female and masculine imputed or inferred names as male without connotation of sex assigned at birth. One of us with no other editorial role (Kiang) performed imputations, inferences, and analyses. The Editor-in-Chief (Lash) coded the review of emails to protect confidentiality of correspondence; Kiang also analyzed these data.FIGURE 1.: Histogram of the probability of a given name corresponding to the imputed gender for genders imputed by use of Genderize.io database, among authors analyzed for submissions to Epidemiology 2015–2020.Between 2015 and 2020, Epidemiology received 3,346 submissions of original research articles; 3,198 (96%) had two or more authors and 148 (4%) a single author. Among submissions by multiple authors, women were equally likely to be first authors (48% women, 49% men, 3% unidentified; Table 1) but less likely to be last authors (37% women, 61% men, 2% unidentified). The most common pair of first and last authors, respectively, were both men (33%), followed by a woman and a man (27%), both women (21%), and a man and a woman (15%). Women were also less likely to be single authors (20% compared with 79% men, 1% unidentified). Across all submissions, 12% were accepted, 11% rejected with review, 76% rejected without review, and 1% did not yet have a decision with final status at the time of analysis (Table 2). These proportions of decisions were similar in all combinations of author pairs (Figure 2) and among submissions by a single author. Although all of the small number (N = 29, <1%) of submissions by a female single author were rejected without review, this proportion is not dissimilar to the other submission categories. We also compared editorial decisions by gender of the corresponding author (Table 1) and found similar trends. Table 1. - Imputed Genders of Authors of Submissions to Epidemiology From 2015 to 2020 Total Female Male Unknown Single authors, n (%) 148 29 (19) 117 (79) 2 (1) Multiple authors, n 3,198 First author, n (%) 1,528 (48) 1,573 (49) 97 (3) Last author, n (%) 1,185 (37) 1,963 (61) 50 (2) Corresponding author, n (%) 3,346 1,446 (43) 1,830 (55) 70 (2) Table 2. - Imputed Genders of Authors by Editorial Decision on Submissions to Epidemiology From 2015 to 2020 Submission Category Total Accepted Rejected With Review Rejected Without Review Missing Status All, n (%) 3,346 401 (12) 367 (11) 2,529 (76) 49 (1) Female corresponding author, n (%) 1,446 201 (14) 163 (11) 1,059 (73) 23 (2) Male corresponding author, n (%) 1,830 198 (11) 202 (11) 1,404 (77) 26 (1) Unknown corresponding author, n (%) 70 2 (3) 2 (3) 66 (94) 0 (0) All by single author, n (%) 148 10 (7) 16 (11) 120 (81) 2 (1) Female single author, n (%) 29 0 (0) 0 (0) 29 (100) 0 (0) Male single author, n (%) 117 10 (9) 16 (14) 89 (76) 2 (2) Unknown single author, n (%) 2 0 (0) 0 (0) 2 (100) 0 (0) All by multiple authors, n (%) 3,198 391 (12) 351 (11) 2,409 (75) 47 (1) Female first author, female last author, n (%) 670 96 (14) 74 (11) 488 (73) 12 (2) Female first author, male last author, n (%) 848 117 (14) 96 (11) 621 (73) 14 (2) Male first author, female last author, n (%) 486 59 (12) 44 (9) 375 (77) 8 (2) Male first author, male last author, n (%) 1,061 118 (11) 133 (13) 797 (75) 13 (1) Unknown pair, n (%) 133 1 (1) 4 (3) 128 (96) 0 (0) FIGURE 2.: Editorial decisions for submissions to Epidemiology 2015–2020, restricted to papers with multiple authors, by categories of genders of author pairs.We considered time to decision for submissions, categorized by submissions that received a review (either accepted or rejected) or did not receive a review (rejected). The median time to first decision for those sent for review was 34 days with an interquartile range of 28 to 46 days, and 2 days (2–4) for those rejected without review (Table 3). Again, the median time to decision was similar across author pairs, for those with and without review. Additionally, we repeated these analyses using only gender imputed by Genderize.io algorithm and the findings were consistent (results provided as an eAppendix; https://links.lww.com/EDE/B894). Table 3. - Median (Interquartile Range) Days to Editorial Decision by Imputed Gender of Authors of Submissions to Epidemiology From 2015 to 2020 Category Review No Review All 35 (28–46) 2 (2–4) Female corresponding author 35 (28–46) 2 (2–4) Male corresponding author 35 (27–47) 2 (2–3) Unknown corresponding author 28 (24–30) 2 (1–3) Female single author – 2 (1–3) Male single author 34 (12–41) 2 (2–4) Unknown single author – 5 (3–6) Female first author, female last author 35 (27–43) 2 (2–4) Female first author, male last author 36 (29–49) 2 (2–4) Male first author, female last author 40 (30–48) 2 (2–4) Male last author, male last author 35 (27–44) 2 (2–3) Unknown pair 31 (28–34) 2 (1–3) We received 49 email appeals for reconsideration of decision, 36 from male correspondents (73% of appeals) and 13 from female correspondents (27%). There was a difference in the rate of success; the decision was reversed for 13 of the 36 male authors (36%) and 9 of the 13 female authors (69%) (difference in proportions = 33%, 95% confidence interval: 4, 63). For papers published in Epidemiology during the term of the current Editor-in-Chief (2015–2020), author gender and combinations of author gender were not associated with the decisions to accept, reject, and reject without review. Likewise, the time to first decision does not depend on author gender or the combinations of author gender. The distribution of combinations of authorship does demonstrate female authors least often in the last, presumably senior, authorship position. This result is in agreement with previous literature that has shown that, in the field of epidemiology, women are more likely to be at earlier career stages and first authors, and less likely to be last authors.13 Female authors are less likely to submit work as single authors. Women represented about half of all designated authors, yet only about one-quarter of requests for reconsideration. However, requests for reconsideration were more often successful when received from women than men. Given the gender distributions of authors, these results are consistent with women being less likely to seek appeals of rejections. It has also been noted in previous studies that women are less likely to speak highly of their own work.16 We are overall heartened at the lack of any notable difference in Epidemiology’s editorial decisions or time to decision in relation to imputed or inferred gender. Although the decisions studied here are perhaps the most important decisions made by the editors, we acknowledge there are other journal activities not yet assessed for gender bias (e.g., invitations for review, participation on the editorial board, selection of the Rothman Paper Prize). Overall, these analyses were somewhat time consuming, but far from difficult. We encourage other editors in the community of scientific publications to undertake similar self-studies.