Gender Differences in Authorship of Critical Care Literature

Abstract

Rationale: Gender gaps exist in academic leadership positions in critical care. Peer-reviewed publications are crucial to career advancement, and yet little is known regarding gender differences in authorship of critical care research.

Objectives: To evaluate gender differences in authorship of critical care literature.

Methods: We used a validated database of author gender to analyze authorship of critical care articles indexed in PubMed between 2008 and 2018 in 40 frequently cited journals. High-impact journals were defined as those in the top 5% of all journals. We used mixed-effects logistic regression to evaluate the association of senior author gender with first and middle author gender, as well as association of first author gender with journal impact factor.

Measurements and Main Results: Among 18,483 studies, 30.8% had female first authors, and 19.5% had female senior authors. Female authorship rose slightly over the last decade (average annual increases of 0.44% [P < 0.01] and 0.51% [P < 0.01] for female first and senior authors, respectively). When the senior author was female, the odds of female coauthorship rose substantially (first author adjusted odds ratio [aOR], 1.93; 95% confidence interval [CI], 1.71–2.17; middle author aOR, 1.48; 95% CI, 1.29–1.69). Female first authors had higher odds than men of publishing in lower-impact journals (aOR, 1.30; 95% CI, 1.16–1.45).

Conclusions: Women comprise less than one-third of first authors and one-fourth of senior authors of critical care research, with minimal increase over the past decade. When the senior author was female, the odds of female coauthorship rose substantially. However, female first authors tend to publish in lower-impact journals. These findings may help explain the underrepresentation of women in critical care academic leadership positions and identify targets for improvement.

At a Glance Commentary

Scientific Knowledge on the Subject

Gender gaps exist in academic leadership positions and scholarly activities in critical care. Publication of peer-reviewed research is crucial to career advancement, and yet little is known about gender differences in authorship of critical care research.

What This Study Adds to the Field

This study found that among more than 18,000 critical care clinical research and basic science studies published in 40 frequently cited journals between 2008 and 2018, women comprised 30.8% of first authors and 19.5% of senior authors, with minimal change over the last decade. When the senior author was female, the odds of female coauthorship rose substantially. However, compared with male first authors, female first authors tend to publish in lower-impact journals. These findings suggest factors that may contribute to the underrepresentation of women in academic leadership positions and scholarly activity in critical care, and they may help identify potential targets for improvement within the field.

Women comprise between 20% and 50% of the critical care medicine workforce, depending on the geographic region (1). However, with respect to their participation in the workforce, women are underrepresented in scholarly critical care activities, including speaking at and chairing critical care conferences (2–5), participating in guideline panels and editorial boards (4, 6–8), and serving in academic leadership positions (9). Several factors contribute to this gender gap. For example, unconscious bias, gender discrimination, and a lack of female mentors and role models are all associated with the underrepresentation of women in academic medicine (10). In addition, invitations to participate in scholarly activities such as conferences and guideline panels are often extended to scientists who publish within the field, highlighting the importance of academic publishing for career advancement (4).

It is unknown whether gender differences exist in the authorship of critical care research. Because publication of peer-reviewed research is crucial to career advancement (11), a better understanding of female authorship of critical care research may offer potential mechanisms for the observed gender gap in academic and scholarly critical care activities and provide insight into the degree to which this gap might be expected to close (12).

Therefore, we sought to describe gender differences in authorship of critical care research over time and to evaluate whether senior author gender is associated with the gender of first and middle authors. In addition, we evaluated whether female first authorship is associated with the odds of publication in lower-impact journals. We hypothesized that 1) female senior authors would be more likely than male senior authors to publish with women in the first and middle author positions, and 2) female first authorship would be associated with publication in lower-impact journals. Some of the results of these studies were previously reported in the form of an abstract (13).

Methods

Data Source

PubMed is an international online database of over 27 million citations of medical publications that is maintained by the National Center for Biotechnology Information through the U.S. National Library of Medicine (14). For this study, 40 frequently cited medical and scientific journals that publish clinical and basic science critical care research were identified by consensus opinion of all authors (Table E1 in the online supplement). Within these 40 journals, two authors and practicing intensivists with experience in both clinical and basic science/translational research (K.C.V. and A.J.R.) independently identified 25 critical care research studies (13 clinical and 12 basic science/translational studies) over the last two decades that were considered important to the field of critical care (Table E2).

Two authors (K.C.V. and A.J.R.) then identified relevant search terms referenced in the titles or abstracts of the 25 selected critical care articles. A total of 26 search terms were selected and agreed on by coauthors; these terms were then used to extract all original critical care research articles from PubMed between November 2008 and October 2018 (Table E3). This reverse engineering approach was developed by coauthors in an effort to maximize academic rigor and replicability of our study selection methods. For these analyses, a year was defined as November to October. From PubMed, we retrieved each study’s PubMed identifier, study type (e.g., clinical trial or observational study), article title, date of publication, and author list. We included original research articles, systematic reviews, clinical practice guidelines, and consensus reports. We excluded from analyses all editorials, commentaries, letters, news, and retractions (15).

Author Identification and Journal Characteristics

Authors were categorized as first, middle, or senior authors on the basis of author sequence. The first author, senior author, and up to 15 middle authors were identified. Because co–first and co–senior authorships were not consistently identified across different journals, author sequence was used to determine first and senior authorship (15). Author gender was determined by matching authors’ first names to entries in a previously validated Genderize database containing 216,286 names across 79 countries and 89 languages (15, 16). The gender of authors whose names were not included in the Genderize database were identified by manual Internet searches of authors’ professional websites, including photographs of authors and/or references to authors by male or female pronouns. Manuscripts with missing gender for first or senior authors were excluded. Authors with the highest number of critical care publications (regardless of authorship position) during the study period were included in analyses of the “100 most prolific authors.”

Characteristics of Publications

Studies were classified by the impact factor of the journal (i.e., high impact or lower impact), the geographic region of the corresponding author, and whether the study was a clinical trial based on MeSH (Medical Subject Headings) annotations (17). We defined high-impact journals a priori as those in the top 5% of all scientific journals indexed in 2017, which corresponded to an impact factor of 6 or higher as ranked by the 2017 InCites Journal Citation Reports (18, 19). Journals without an impact factor in 2017 were considered lower-impact journals. We identified the region of the corresponding author using the country code top-level domain assigned to corresponding authors’ e-mail addresses (20). E-mail addresses of corresponding author e-mails were extracted from PubMed or obtained via manual review of included studies. Countries of corresponding authors were then grouped into geographic regions (i.e., Asia, Australia/New Zealand, Europe, North America, or Other). Countries and/or regions grouped in the Other category were those with <5% of studies attributable to corresponding authors from those locations. See the online supplement for further details about methods used to determine the region of the corresponding author.

In addition, we chose a priori to examine female first and senior authorship in the three highest-impact journals based on impact factor (New England Journal of Medicine, The Lancet, and JAMA). Studies were also classified by publication in either clinical or basic science journals. As determined by consensus of coauthors, clinical journals were defined a priori as those that publish observational clinical studies and/or randomized controlled trials, with or without basic science or translational studies. In contrast, basic science journals were defined as those focused on basic science or translational studies.

Statistical Methods

We first calculated the percentages of women in first and senior author positions across included studies. To evaluate temporal trends in female first and senior authorship, we performed simple linear regression analyses of the proportion of first and senior authors during each year of the study. We confirmed the validity of the linearity assumption by comparing the fitted line from our model with a locally weighted scatterplot smoothing plot. We used descriptive statistics as appropriate to examine the proportion of female middle authors in studies with more than two authors.

We then performed several distinct multivariable analyses. In the first and second analyses, the primary exposure was senior author gender, and the primary outcomes were first author and middle author gender, respectively. Middle author gender was defined as the presence or absence of at least one female middle author. Covariates in the first model included journal domain (i.e., clinical or basic science), type of study (i.e., clinical trial or other), year, and region of corresponding author. The second model included these same covariates and an additional covariate of the number of middle authors. In the third analysis, the primary exposure was first author gender, and the primary outcome was journal impact factor (i.e., high vs. lower impact). Covariates in this model included senior author gender, journal domain, type of study, year, and region of the corresponding author.

Finally, we performed an exploratory analysis to evaluate factors associated with senior author gender. Covariates in this model included journal domain, type of study, year, and region of the corresponding author. For all analyses, we performed mixed effects logistic regression to test for adjusted differences in outcomes. Because publications are not independent events, we included a random effect of senior author to account for clustering of publications. All statistical tests were two sided, and a P value <0.05 was considered statistically significant. All analyses were performed using Stata version 15 software (StataCorp). The study was considered exempt from review by the Stanford University Institutional Review Board.

Results

We extracted 18,483 critical care research articles published between November 2008 and October 2018. Using the Genderize database, we identified the gender of 75.4% of 12,687 unique first authors across 14,190 publications and 74.2% of 9,792 unique senior authors across 14,406 publications. After manually identifying missing author gender for the remaining authors, we identified the gender of 96.5% of first authors and 94.0% of senior authors. At the publication level, we identified gender of the first author for 17,033 studies (92.2%) and gender of the senior author for 16,902 studies (91.5%) (Figure 1).

Figure 1.

Inclusion and exclusion criteria for selection of critical care research articles published between 2008 and 2018.

Trends over Time

Across all publications during the study period, 30.8% and 19.5% had female first and senior authors, respectively. Between 2008 and 2018, there was a small but significant increase in the proportion of female first and senior authors of critical care publications (Figure 2). The proportion of female first authors rose from 27.5% in 2008 to 32.6% in 2018, with an average annual increase of 0.44% (95% confidence interval [CI], 0.23–0.65%; P < 0.01). The proportion of female senior authors increased from 18.9% in 2008 to 21.4% in 2018, with an average annual increase of 0.51% (95% CI, 0.22–0.80%; P < 0.01). Female authors accounted for only 11 of the top 100 most prolific authors of critical care research during the study period (Figure 3).

Figure 2.

Trends in female first and senior authorship of critical care research between 2008 and 2018.

Figure 3.

Top 100 most prolific authors of critical care research between 2008 and 2018. “Most prolific authors” is defined as authors with the highest number of critical care research publications (regardless of authorship position) during the study period.

Manuscript Characteristics and Authorship Position by Gender

Women were first and senior authors in 30.4% and 20.1% of high-impact journals, respectively (Table 1). Nearly one-third (30.2%) of all publications did not include any female middle authors. Women comprised 29.9% of first authors of studies published in clinical journals, compared with 34.6% of basic science journals. Europe and Australia/New Zealand had the highest percentages of female first and senior authors, respectively, whereas Asia had the lowest percentage of female first and senior authors (Table 1). Among studies published in New England Journal of Medicine, 20.2% had female first authors and 32.4% had female senior authors. In The Lancet, 21.3% and 25.8% of first and senior authors were women, respectively, compared with 27.1% and 19.5% in JAMA.

Table 1.

Proportions of Female First and Senior Authors of Critical Care Research Articles Published between November 2008 and October 2018

	Percentage of Female First Authors (95% CI)	Percentage of Female Senior Authors (95% CI)
Overall	31 (30–32)	20 (19–20)
Impact factor
Top 5%	30 (29–32)	20 (19–21)
Lower 95%	31 (30–32)	19 (18–20)
Domain
Clinical	30 (29–31)	20 (19–20)
Basic science	35 (33–36)	20 (18–21)
Region
North America (United States/Canada)	32 (30–33)	21 (20–22)
Asia	25 (23–27)	17 (15–19)
Australia/New Zealand	30 (26–33)	22 (19–25)
Europe	32 (30–33)	18 (17–19)
Other	35 (32–39)	22 (19–25)
Clinical trial	30 (28–32)	21 (19–23)
Highest-impact journals
New England Journal of Medicine	20 (15–26)	32 (25–40)
The Lancet	21 (14–28)	26 (18–34)
Journal of the American Medical Association	27 (21–33)	20 (14–25)

Definition of abbreviation: CI = confidence interval.

Among the 17,251 publications with more than two authors, 25.5% of middle authors were women. Female and male senior authors included a similar mean number of middle authors in each publication (mean ± SD, female senior authors, 5.6 ± 4.1; male senior authors, 5.7 ± 3.9; P = 0.30). In unadjusted analyses, which included 14,935 studies that had more than two authors and nonmissing gender for both first and senior authors, female first and senior authors published with a significantly higher proportion of female middle authors than male first and senior authors, respectively (first author, 30.4% vs. 23.8% [P < 0.01]; senior author, 30.6% vs. 24.8% [P < 0.01]).

Multivariable Analyses

The odds of female first authorship nearly doubled when the senior author was female rather than male (adjusted odds ratio [aOR], 1.93; 95% CI, 1.71–2.17) (Table 2). Similarly, publications with a female senior author had a 48% increase in odds of including any female middle authors (aOR, 1.48; 95% CI, 1.29–1.69) (Table 2). Female first authors had higher odds than male first authors of publishing in lower-impact journals (aOR, 1.30; 95% CI, 1.16–1.45) (Table E4). In exploratory analyses, female senior authorship was not significantly associated with journal domain, type of study, year, or region of the corresponding author (Table E5).

Table 2.

Multivariable Analysis Evaluating the Association of Senior Author Gender with First and Middle Author Gender across Critical Care Research Articles Published between 2008 and 2018

	Odds of Female First Authorship (95% CI)	Odds of Female Middle Authorship* (95% CI)
Senior author gender†
M	Reference	Reference
F	1.93 (1.71–2.17)	1.48 (1.29–1.69)
Year	1.03 (1.01–1.04)	1.02 (1.00–1.04)
Number of middle authors	—	1.47 (1.44–1.51)
Domain
Clinical	Reference	Reference
Basic science	1.29 (1.15–1.44)	1.10 (0.97–1.24)
Location of corresponding author
North America (United States/Canada)	Reference	Reference
Asia	0.68 (0.59–0.77)	0.28 (0.24–0.32)
Europe	1.09 (0.98–1.22)	0.90 (0.80–1.02)
Australia/New Zealand	0.91 (0.70–1.18)	1.01 (0.76–1.33)
Other	1.21 (0.96–1.51)	0.87 (0.67–1.12)
Clinical trial
No	Reference	Reference
Yes	0.98 (0.84–1.14)	1.00 (0.84–1.19)

Definition of abbreviation: CI = confidence interval.

Statistically significant comparisons are shown in bold.

^*In 14,935 publications with more than two authors and nonmissing gender of both first and senior authors.

^†Primary exposure.

Discussion

In this study, we found that women comprised less than one-third of first authors and less than one-fourth of senior authors of critical care research published between 2008 and 2018. Furthermore, the increase in the proportion of women in first or senior author positions over the last decade has been minimal. Although female first authors more often publish in lower-impact journals than male first authors do, female senior authors have substantially higher odds of collaborating with other women. These findings suggest factors that may contribute to the underrepresentation of women in academic leadership positions and scholarly activity in critical care, and they may help identify targets for improvement within the field.

Our findings are consistent with several recent studies confirming an existing gender gap in academic publishing (12, 21, 22). For example, of the 27.3 million researchers who authored 5.5 million research papers indexed in the Web of Science between 2008 and 2012, over 70% were men (21, 23). A 2016 study found that the number of female first authors of original research in six high-impact journals increased from 27% in 1994 to 37% in 2014. However, this study also reported that since 2009, female first authorship in those same journals had either plateaued or declined, suggesting that underrepresentation of research by women in high-impact journals remains an important concern (12).

Multiple factors could contribute to these observations. First, most journals do not conduct double-blind review of submitted manuscripts, which has been shown to favor increased representation of female authors by overcoming unconscious biases and gender stereotypes (12, 24, 25). A minority of journal editors, members of editorial boards, and participants in the development of clinical guidelines are women, which may adversely affect women’s success in publishing their research in a particular journal (4, 5, 7, 12).

The “confidence gap” between men and women observed in other industries may also contribute to a reluctance of female scientists to submit their manuscripts to high-impact journals. A widely cited Hewlett-Packard survey found that the company’s female employees applied for promotion only when they believed they met 100% of the qualifications required for a particular position, whereas male employees applied when they believed they could meet 60% of the job requirements (26). Similarly, a Science audit of the gender of its published authors noted that the journal received one-third fewer manuscripts from female authors than would be expected on the basis of the number of women in their respective fields (27). A parallel signal has been observed in NIH grant applications: Women tend to have lower application rates than men, but similar success rates (12, 28).

Moreover, a recent study found that clinical articles in which the first and last authors were both women were significantly less likely to use positive terms to describe their research findings than articles in which the first and/or last author was a man (29). These gender differences in the positive framing of research findings was most pronounced in the highest-impact journals and suggest one possible mechanism for our findings that female first authors were more likely to publish in lower-impact journals. Publishing first-author work in lower-impact journals could lead to less opportunity for career advancement, including NIH grants at the crucial assistant professor–to–associate professor stage, and thus could contribute to a vicious cycle in which underrepresentation of women in higher ranks of academia leads to poor participation as peer reviewers and authors, leading to fewer publications, less funding and awards, and ultimately less advancement (30, 31).

In addition, the persistent gap in female authorship of critical care research has been slow to improve. This finding mirrors the slow rate of increase in the proportion of female intensivists over the last several decades—a number that has lagged behind the growing number of female medical graduates worldwide (32). For example, in the United States, pulmonary/critical care fellowships have had among the slowest rates of increase in the proportion of female fellows since 1991, with women representing only 32.6% of all pulmonary/critical care fellows in 2016 (33). Similarly, a 1995 survey of graduates from intensive care medicine training programs in Australia and New Zealand identified 13 of 120 respondents (11%) as women (34); a follow-up study in 2015 found that women comprised 18.7% of the total fellowship of the College of Intensive Care Medicine of Australia and New Zealand (32).

More broadly, gender inequality exists in leadership positions across specialties within academic medicine. According to the Association of American Medical Colleges, women comprise only 24% of all division chiefs, 24% of vice chairs, 15% of department chairs, and 16% of medical school deans (35). A 2014 analysis of more than 90,000 physician faculty at U.S. medical schools found that women were significantly less likely than men to achieve full professor status, even after adjusting for experience, productivity, specialty, and age (36).

In 2018, the American Thoracic Society published a workshop report identifying several factors that contribute to this gender gap within the fields of pulmonary, critical care, and sleep medicine: 1) gender climate (e.g., implicit and perceived gender bias), 2) disproportionate burden of family responsibilities among women faculty, 3) the lack of women in leadership positions, 4) poor retention of women, and 5) the lack of gender equality in compensation (37). Several strategies have been implemented to address gender inequality in academic medicine. For example, the University of California, Davis School of Medicine established the Women in Medicine and Health Science program to help women overcome barriers to career advancement. This program includes explicit advocacy for women’s advancement in leadership positions and the collection of employee data to inform institutional decisions and actions. In the first 10 years of the program, the percentage of female faculty doubled from 18% to 36%, with a concomitant increase in the number of female full professors and department chairs (38). The NIH Office of Research on Women’s Health has also implemented several successful initiatives to promote the careers of women in biomedical sciences. These include programming to support researchers returning to the workforce after a qualifying hiatus, as well as the NIH Working Group on Women in Biomedical Careers that assists with mentoring and networking opportunities (39). Finally, leading journals have made commitments to address gender imbalances in academic publishing, such as setting diversity targets for commissioned content, peer reviewers, and editorial roles (21) and providing training to editors on diversity initiatives and unconscious gender bias (21, 40).

In particular, mentorship is associated with increased career satisfaction, faculty retention, research productivity (including publication and grant success), and career advancement (41). Women are less likely than their male colleagues to have a mentor across varying levels of training and are often less satisfied with their mentorship experiences (42–44). Our study adds to the literature by demonstrating that female senior authors are significantly more likely than male senior authors to publish with female first and middle authors, suggesting that women may mentor and collaborate with other women more often than men do. These findings highlight the potential value of efforts to increase the pool of women in senior academic positions available to serve as mentors for both junior and midcareer female faculty, who are particularly vulnerable to attrition from academic medicine (42, 44).

The importance of women advancing to senior author positions also raises the potential benefits of sponsorship by current academic leadership. Sponsors differ from mentors in that sponsors have both the position and the power to advocate publicly for the advancement of nascent talent within their organization (45). Sponsorship programs are increasingly being used in the corporate world to help raise women’s visibility, enhance their credibility, and advance them into upper levels of leadership (45, 46). In this way, sponsorship is one tool that may also help address the gender gap that exists in leadership positions and scholarly activities within critical care. For example, between 2010 and 2016, women comprised only 5–31% of speakers overall at five major critical care conferences worldwide (5). Women have also been notably absent from the development of high-impact critical care consensus statements and clinical practice guidelines in recent years (4, 47, 48). Our findings may galvanize both men and women in positions of leadership to be more proactive in mentoring and sponsoring female scientists to help increase the number of female senior investigators in the field.

Strengths of our study include the large sample size and the identification of the gender of 97% and 94% of first and senior authors, respectively. Our study also identified potentially important geographic differences in the authorship of critical care research, which should be explored in future research. This study also has limitations. Although we included 40 frequently cited journals, a substantial body of literature exists beyond these journals. It is possible that the 26 search terms used to identify critical care research did not identify all relevant studies published between 2008 and 2018. There is also a risk of publication type, name, or gender misclassification, as well as a risk of bias resulting from the reverse engineering approach used to identify studies.

Importantly, the actual difference in rates of female first authorship in high- versus low-impact journals was small before adjustment and represents only one possible factor contributing to the gender gap in academic critical care. In addition, residual confounding likely exists despite our adjustment for potential confounders. For example, we were unable to account for institutional or cultural differences that may influence the ability of both women and men in academic critical care to obtain grant funding and/or publish research. Finally, the true denominator of women pursuing careers in academic critical care worldwide is unknown.

Conclusions

Women comprise less than one-third of first authors and less than one-fourth of senior authors of critical care research, with minimal increase over the past decade. Although female first authors tend to publish in lower-impact journals more often than male first authors do, female senior authors have substantially higher odds of collaborating with other women. These findings improve our understanding of the gender gap in authorship of critical care literature and help identify potential targets for improvement within academic critical care medicine.