Review characteristics
Of the 626 systematic reviews included in our study, 365 (58%) listed the first author as the corresponding author; therefore, we assessed 887 total author positions (365 “First and also the corresponding authors”, 261 “First authors”, and 261 “Corresponding authors”) (Table 1). Authors could appear multiple times in our sample if they published more than one review; accordingly, we identified 751 unique authors in the 887 total author positions assessed. Most authors appeared only a single time (Table 1). The proportion of reviews with the same first and corresponding author was greater for Cochrane than non-Cochrane reviews (respectively 75% and 41%), which led the Cochrane reviews to have a larger sample of “First & corresponding authors” and non-Cochrane reviews to have a larger sample of “First authors” and “Corresponding authors” (Table 1). We also found 11 authors (three women and eight men) who contributed to both Cochrane and non-Cochrane reviews.
Author gender
The majority of genders was ascertained through a strong association of the name with a particular gender (309/751, 41%) or an internet search (175/751, 23%) (Table 1). We were unable to ascertain the gender of 104/751 (14%) authors: 4% (11/301) of Cochrane authors’ genders could not be determined, but 21% (93/439) of non-Cochrane authors’ genders were unknown. Overall, women comprised 276/751 (37%) unique authors and 322/887 (36%) of all author positions analyzed, as authors could appear multiple times. Cochrane reviews had a higher proportion of unique authors who were women (142/301, 47%) compared with non-Cochrane reviews (131/439, 30%) (Table 1).
Among all “First and also the corresponding authors” (n = 365), 158 (43%) were women; the proportions of women authors were comparable between Cochrane (106/237, 45%) and non-Cochrane (52/128, 41%) reviews (Table 2). Among all “First authors” (n = 261) and “Corresponding authors” (n = 261), 98 (37%) and 66 (25%) respectively were women (Table 2). This apparent discrepancy is likely due to the distribution of author types between sources as Cochrane authors comprise the majority of “First and also the corresponding authors” and have a higher proportion of women, whereas non-Cochrane authors have lower representation of women but comprised the majority of “First authors” and “Corresponding authors”. For these two author positions (i.e., first or corresponding author), Cochrane reviews had a greater proportion of women – respectively 40/76 (53%) and 35/76 (46%) – than non-Cochrane reviews – respectively 58/185 (31%) and 31/185 (17%) (Table 2). The odds of having a woman in the positions of “First and also the corresponding author”, “First author”, and “Corresponding author” were 1.14 (95% CI: 0.72 to 1.82), 1.50 (95% CI: 0.81 to 2.77), and 3.67 (95% CI: 1.90 to 7.07) for Cochrane reviews as compared to non-Cochrane reviews, respectively.
The gender of “First author” and “Corresponding author” positions in non-Cochrane reviews had a substantial amount of uncertainty – respectively 53/185 (29%) and 37/185 (20%) were unknown – as compared with authors of Cochrane reviews for which only 2/76 (3%) and 5/76 (7%) of authors’ genders were unknown (Table 2).
In our sensitivity analyses, in which we assumed the extreme situations of women comprising all or none of the missing genders, we found that representation of women among Cochrane authors did not substantially change (Table 2). Among non-Cochrane authors, under a best-case assumption that all missing author genders were women, the representation of women met or surpassed among Cochrane authors for two of the categories (Table 2). Among corresponding authors for non-Cochrane reviews, however, under the extreme assumption of all missing author genders being women, only 68/185 (37%) would be women (Table 2). The odds ratios revealed similar patterns under these sensitivity analyses with only the corresponding author position being held by women significantly more often for Cochrane reviews than non-Cochrane reviews under both best-case (OR = 1.91, 95% CI: 1.07 to 3.40) and worst-case (OR = 4.24, 95% CI: 2.24 to 8.00) assumptions.
Author location
We were able to identify the geographic location for all authors based on their institutional affiliation. Most authors for the reviews included in our sample had institutions in either European (247/751; 33%), Western Pacific (243/751; 32%), or North American (211/751; 28%) regions of the world. As compared to non-Cochrane reviews, Cochrane reviews had a greater proportion of authors from Europe – respectively 125/301 (41%) and 116/439 (26%) – and a lower proportion from the Western Pacific region – respectively 59/301 (20%) and 181/439 (41%) (Table 1 and Figure). Other parts of the world were uncommon contributors of authors to reviews in eyes and vision, with Africa being the least frequently appearing region with only three authors hailing from African countries (Table 1 and Figure).
The specific countries which contributed the most authors to reviews in our sample were China (172/751, 23%), the United States of America (USA) (170/751, 23%), and the United Kingdom (UK) (124/751, 17%). Between these three countries, Cochrane reviews had more authors from the USA (78/301, 26%) and UK (95/301, 31%) than China (19/301, 6%), whereas non-Cochrane reviews had the opposite with 151/439 (34%) of authors being located in China and 90/439 (20%) and 25/439 (6%) respectively located in the USA and the UK.
To explore the possibility that the differences in gender distribution observed between Cochrane and non-Cochrane reviews was influenced by the location of contributing authors, we examined the gender among the 458 unique authors from Europe and North America. We found the gender representation was equal among the Cochrane authors (49% women, 50% men, 1% unknown), but there was a disparity in the representation of women among non-Cochrane authors (39% women, 60% men, 1% unknown).