Gender trends in authorships and publication impact in Academic Radiology—a 10-year perspective

Objectives To analyze the development of publication numbers of female authors in high-, medium-, and low-impact radiological journals. Methods In this bibliometric analysis, gender of the first (FA) and senior author (SA) was assigned to all original research articles and reviews, published in 10 high-, medium-, and low-impact radiological journals in 2007/8 and 2017/18. The adjusted event rate (AER) and adjusted odds ratio (AOR) were calculated using mixed logistic and multinomial logistic regression models to assess and compare female publications according to impact factor, journal, author position, and combination. Results The proportion of female FA and female SA in N = 6979 (2007/2008) and N = 7383 (2017/2018) articles increased to 29.1% and 16.1% in 2017/2018, respectively. While most female authorships were continuously observed in medium-impact journals, the strongest increase occurred for both female FA (AOR 2.0; p < .0001) and SA (AOR 2.1; p < .0001) in low-impact journals. Female SA published significantly more often in a low- (AOR 1.5) or medium- (AOR 1.8) than in a high-ranking journal. Among the high-ranking journals, female FA published most frequently in European Radiology (32.4%; 95% CI [29.3–35.8]; p < .0001), female SA in Investigative Radiology (15.9%; 95% CI [13.7–18.4]; p < .0001). Male same-sex authorships decreased (AOR 0.9), but remained at least twice as common as all-female or mixed authorships. Conclusion The increase in female authorship is reflected in all impact areas. Female FA and SA increased most in low-ranking journals but are most common in medium-ranking journals. Female SA remain rare, especially in high impact journals. Key Points • Compared to the proportion of female radiologists worldwide, female senior authors are underrepresented in all impact areas, in particular in high-impact journals. • Among the included high-ranking radiological journals, female first authors and senior authors were strongest represented in European Radiology and Investigative Radiology, while across all impact areas they mostly published in medium-ranking journals. • Female author combinations were more frequent in low- and medium- than in high-ranking journals, whereas male author combinations remained more common than female senior author collaborations in all impact areas.


Introduction
Nowadays, female students make up for more than half of all medical students in European countries, as the UK, Sweden, or Germany [1], and in the USA [2]. The rate of female doctors in these countries ranges well within the gender balanced zone of 40-60% [1]. Concerning radiologists, 33.5% worldwide are female; in the age cohort of those under 35 years, the proportion of female radiologists ranges up to 48.5% [3].
However, although women are even more likely than men to begin academic careers in medicine after training, they lag behind their male counterparts in obtaining levels of rank such as faculty senior positions [4][5][6]. A major factor influencing an academic advancement is the h-index, which is dependent on scholarly productivity based on peer-reviewed scientific publications and the number of citations. Fewer high-ranked publications might lead to less citations and thus poorer chances of promotion to senior positions. Systematic analyses of female authorships were successful in revealing the unequal distribution of leading author positions in radiology, while at the same time demonstrating a positive trend in development [7,8].
The latest analyses however ended in 2016 and focused on high-ranking radiological journals such as European Radiology (Eur Radiol), Radiology, and the American Journal of Radiology (AJR) [7][8][9][10]. The aim of this study is to address this gap and to generate a comparison of the publication numbers of female authors from the last 10 years, between 2007/2008 and 2017/2018, in ten representative radiological journals of varying impact factors. Furthermore, the frequency of same-sex and mixed-sex first and senior author co-operations was examined to generate insights into gender-specific collaboration of junior and senior faculty members.

Materials and methods
An institutional review board exemption was obtained for this retrospective study. Ten radiological journals of high, medium, and low impact according to the Clarivate analytics web of science's journal citation reports for "Radiology, Nuclear Medicine, and Medical Imaging" were chosen. Care was taken to include only representative, general radiological journals and no subspeciality journals, e.g., for cardiovascular imaging, and to include journals from the European, American, and Asian continent to take potentially varying developments between journals from different regions into account. Also, as far as feasible journals were only included, if their impact factor did not change between the defined low impact area (impact factor range: 1.9 to 1.5), medium impact area (range: 3.2 to 2.1), or high impact area (range: 7.6 to 3.9) over the course of the study period (2007/2008 to 2017/2018). We finally included the following journals in our analysis: high impact journals-Radiology, Investigative Radiology (Invest Radiol), Eur Radiol, and Magnetic Resonance in Medicine (MRM); medium impact journals-AJR, European Journal of Radiology (EJR), and Clinical Radiology (Clin Radiol); lower impact journals-Fortschritte auf dem Gebiet der Röntgenstrahlen und bildgebenden Verfahren (Rofo), Acta radiologica (Acta Radiol), and Japanese Journal of Radiology (Jpn J Radiol). Of these, only MRM changed from medium to high impact between 2007 and 2008 and Clin Radiol from low to medium impact from 2007/2008 to 2017/2018. Both were assigned to their impact area based on the impact factor in 2017/2018. All publications of 2007, 2008, 2017, and 2018 were listed and classified as original research articles (ORA), reviews, and others (e.g., editorials, pictorial essays, letters to the editor). Only ORAs and reviews were included into further analysis, as authorships of letters to the editor, comments, etc. were considered to be less relevant for a scientific career. The full names of the first authors (FA) and senior authors (SA) were obtained from all included ORAs and reviews. All first names were manually assigned to their supposed gender by the German authors IW, LW, and TN. Asian names were re-evaluated by native speakers (Japan, Korean, Vietnamese, Chinese). If available, ORCID IDs were used to verify the identification. In the case of genderambiguous first names (about 40%), several steps to ascertain high assignment were taken. First, further information concerning the author was searched by common and scientific search engines as "google" and "research gate." If only author initials were provided, other publications of the same research group as well as pictures, gender-specific pronouns, and CVs on the webpages of the authors' institutes were analyzed. Lastly, if the sex was still uncertain, it was classified as unknown. An overview of the number of all included publications, of the number of article types, and of female FA and SA depending on article type and impact factor is given in the flowchart (Fig. 1).
To assess an under-or overrepresentation of female authors in the included international journals, female author proportions were compared to the number of female radiologists, as given by a study of Cater et al [3] from 2018 who made inquiries about the members of radiological societies from North, Central, and South America, Europe, Asia, Africa, and Oceania. Mean rate of female radiologists worldwide was thus given with 33.5%, with a range of 27.2% (in the USA) to 85% (in Thailand) and proportions that ranged between 28.8 and 68.9% for European radiological societies [3]. As the first authors are more likely to be junior faculty members, female FA were additionally put in relation to the mean number of radiologists of 35 years or younger (48.5%) [3].

Statistical analysis
Logistic mixed regression models were used to analyze any difference in the percentages of female and male authors. The outcome variable was indicating whether the FA or SA of the publication was female. FA variable and SA variable were the same for ORAs and reviews. If not indicated, otherwise all results apply thus for authors of ORAs and reviews taken together. Since for each publication two observations were given (i.e., one information for the first author and one for the senior author), a random effect was included in the model indicating the publication identification. As fixed effects, year, journal, impact factor, and the variable indicating whether the observations belonged to the first or senior author were considered in the model as well as interactions between the fixed effects variables. Within this model, an observation was defined missing if the gender for the first and senior authorship was unknown. A publication could be included in the analysis if at least for one authorship (either first or senior, or both) gender was given. A multinomial logistic regression model was used to evaluate any differences between percentages of the four possible author combinations (i.e., male-both, female-both, male FA-female SA, female FA-male SA). Thus, the categorical outcome variable consisted of these four groups. The year and journal as well as the corresponding interaction term were included as independent variables in the model. Adjusted event rates (AER) and adjusted odds ratio (AOR) along with 95% confidence intervals are reported throughout the manuscript, i.e., odds ratios of a comparison are adjusted for all other independent variables in a model (model-based estimates). The odds ratio was used to describe differences between time points or between the publication shares of two groups, e.g., female first versus male first authors. All analyses and thus p-values are considered descriptive due to the explorative study design (i.e., no adjustment for multiple testing was conducted). Statistic software R (Version 3.5.1) was used. Statistical analyzes were conducted by AO.  The number of the first and senior authors (male and female) which were successfully assigned to one gender is given. Abbreviations: FA first author, SA senior author (29.5%) and 1395 (16.9%) were female FA and SA, respectively, in ORA. Publication shares of female authors were thus lowest for SA in original research papers.

Female authorships stratified by the journal
In the 10-year comparison, the publication shares of female lead authors (FA and SA together) increased significantly in Invest Radiol (12.5 (Fig. 3a).   (Fig. 3b, Table 3). The lowest proportion of female FA in all journals from the high or medium impact area and also the lowest increase over 10 years (16.2 to 19.7%, AOR 1.3 [95% CI 1.0-1.7]; p = .09) was documented for MRM. Across all impact areas, Jpn J Radiol showed the lowest proportion of female FA associated with one of the lowest increases of female FA in 10 years (10.8 to 13.8%, AOR 1.3 [95% CI 0.7-2.4]; p = .38) (Fig. 3b, Table 3).
With regard to the development over time, the shares of same-sex male combinations taken together for all impact areas decreased slightly between 2007/2008 and 2017/2018 (AOR 0.91 [95% CI 0.9-0.9]; p < .0001) and the shares of female same-sex combinations increased slightly (AOR 1.04

Discussion
This is the first study which investigated and compared female author shares between radiological journals of low, medium, and high impact factor. The main findings of this study were that (a) apart from a positive trend in the publication shares of female authors in all impact areas between 2007/2008 and 2017/2018-which however was significant only in Eur Radiol, EJR, Invest Radiol, and Radiology-(b) women were differently represented depending on the impact areas. Lowranked journals showed the highest increase of female authors, while female author proportions were consistently highest in medium impact journals. Especially, female SA remained underrepresented in all impact areas, notably in high-ranked journals. (c) Same-sex male authorships in 2017/2018 were still four times more common than samesex female or mixed-sex female SA authorships and twice as common as mixed-sex male SA authorships. The proportion of female FA (29.1%) and SA (16.1%) taken together for all impact areas in 2017/2018 as ascribed in this study is lower than in previous studies. For example, in an analysis from 2011 to 2015 of the medium-to high-ranked journals Radiology, Academic Radiology, and AJR, Campbell et al [7] attributed 30% of FA and 25% of SA to women. A study of the four medium-to high-ranking journals Radiology, EJR, Journal of Computer Assisted Tomography, and AJR described a share of 31.5% FA and 22.1% SA in 2014 [10]. For European Radiology, 35.0% FA and 18.0% SA were described for 2016 [9]. These differences may be due to the fact that our study examined a higher range of representative journals and thus a higher number of articles from all impact areas. In addition, our analysis with data from the years 2017/ 2018 is more up to date compared to previous studies that examined publication dates until 2016 [9].
To judge about female representation based on the number of described female authorships, the number of female radiologists as provided in the literature was considered. The rate of 17% female SA demonstrated an apparent underrepresentation compared to even the country with the lowest female radiologist rates (USA 2018: 27.2%) [3]. Female FA would be underrepresented compared to the mean rate of female radiologists worldwide (33.5%) [3]. If comparing female FA to the mean rate of female radiologists aged 35 years or younger (48.5%), as FA are mostly from junior faculty ranks, this gap becomes even more evident [3].
With regard to the lower shares of female SA in high-ranked journals compared to medium-or low-ranked journals, a gap in financial means can be discussed as an influencing factor. In most cases, senior authors, as principal investigators, set the course for the success of a study by providing not only their scientific experience but also financial means. Female applicants for third-party funding, however, are on average less likely to receive funding and mostly raise smaller amounts of money than their male counterparts [11]. Concerning governmental grants, the acceptance rates of National Institutes of Health (NIH) grants are comparable between both sexes, but the proportion of female grants awarded still remains noticeably lower [12]. The striking differences in the publication shares of some journals of similar impact factor in this study are comparable to the results of previous studies, e.g., while high rates were described for female FA and SA in European Radiology, low rates were found for Investigative Radiology, both of them being high-ranked journals [9]. This may in part be explained by different scientific priorities of the journals [9]. For example, the proportion of female authorships is significantly higher in subspecialties for women's health, such as breast imaging, pediatric radiology, and gynecological imaging, than for interventional subjects [7,9].
Gender disparity in editorial board composition could be another influencing factor. According to a recently published study on this subject, 81% of large radiological society board members were male, 19% female [13]. While high-ranking (radiological) journals mostly follow a double-blinded review process, 52.1% of the journals in the category "radiology, nuclear medicine, and medical imaging" still used single-blinded review (author identities known by the editor and reviewer) in 2020 [14], which improves acceptance rates for established authors [15]. As those are more likely senior faculty members and thus male, gender disparities in acceptance rates of such journals might exist. Initial submission rates, which we do not know, should also be kept in mind as a probable reason for different gender proportions of female authors between individual journals and different impact areas. It might be that women prefer to submit to medium instead of high impact journals, e.g., due to a lack of confidence, which however would not be logical from an academic career perspective.
Both, the demonstrated general publication gap and the reduced proportions of female senior authors in high-ranked journals are likely to have consequences on female careers. Less authorships in high-ranked journals contribute to lower citation rates of female authors [10] and thus a lower h-index and academic progress. This relation is supported by a study, which showed that after multivariable adjustment of publication counts and NIH funding rates, full professorship among male and female radiologists was not significantly different [11]. As a narrowing of the citation gap for articles of male and female authors in major radiological journals between 1984 and 2014 has been described, further studies should evaluate whether this important development continues [10].
With regards to the higher publication share of male same-sex compared to female same-sex author co-operations and the higher proportions of male SA combinations compared to female SA combinations, one might conclude that junior scientists of both sexes prefer cooperating with a senior male scientist. However, other studies have described a preference of same-sex publishing [9] and collaboration choices might not always be voluntary considering the low proportion of female senior faculty members and the overall low number of female senior authors.
While some journals such as European Radiology [16] already strive to evaluate and support female scientists, studies such as this are needed to continuously evaluate the progress of female authors' representation in Science. Not only is gender parity in Science desirable to support female careers in academic Fig. 4 Publication co-operations between female and male leading authors in 2017/2018. The highest publication shares (adjusted event rate %) were found for same-sex male first authors and male senior authors. The lowest publication shares were found for same-sex female authors medicine but also because a higher proportion of female authors shifts the focus from the historical orientation of medical research towards the male gender to a better balanced one [7]. Equal representation of women in science on the long-term thus will allow us to also better care for our patients.
This study has some limitations. First, in some cases, it was not possible to assign the author's gender. Statistics however were exclusively based on the number of total observations, defined as all FA and SA for which gender was successfully assigned. Secondly, the total observations of author names which were successfully assigned to one gender were lower in 2007/ 2008 than in 2017/2018. This was because in 2007/2008 more first names were only provided as initials and ORCID ID did not exist yet. However, this should have affected male and female authors equally. Also, publication numbers differed between journals (e.g., n = 2122 publications in European Radiol in 2017/2018, n = 222 in Rofo) which partially resulted in low numbers for the subgroup analysis of individual journals. As each impact area was represented by three journals, low numbers of individual journals however were averaged out in the analysis about the impact areas. Lastly, two of the included journals changed the impact area (MRM between 2007 and 2008 from medium to high; Clin Radiol from low to medium between 2007/ 2008 and 2017/2018). However, no changes occurred from 2017 to 2018 which would concern the most recent results on author proportions.
In conclusion, the positive trend in the development of female authorships in radiology over the past 10 years is reflected in all impact areas. Female FA and SA increased most in low-ranked journals but are most common in medium-ranked journals. Female SA remain underrepresented in all impact areas, especially in high impact journals. In every impact area, same-sex male authorships remain at least twice as common as all other author collaborations.