Objectives The ‘gender gap’ in academic medicine remains significant and predominantly favours males. This study investigates gender disparities in research performance in an Academic Health Science Centre, while considering factors such as mentoring and scientific collaboration.
Materials and methods Professorial registry-based electronic survey (n=215) using bibliometric data, a mentoring perception survey and social network analysis. Survey outcomes were aggregated with measures of research performance (publications, citations and h-index) and measures of scientific collaboration (authorship position, centrality and social capital). Univariate and multivariate regression models were constructed to evaluate inter-relationships and identify gender differences.
Results One hundred and four professors responded (48% response rate). Males had a significantly higher number of previous publications than females (mean 131.07 (111.13) vs 79.60 (66.52), p=0.049). The distribution of mentoring survey scores between males and females was similar for the quality and frequency of shared core, mentor-specific and mentee-specific skills. In multivariate analysis including gender as a variable, the quality of managing the relationship, frequency of providing corrective feedback and frequency of building trust had a statistically significant positive influence on number of publications (all p<0.05).
Conclusions This is the first study in healthcare research to investigate the relationship between mentoring perception, scientific collaboration and research performance in the context of gender. It presents a series of initiatives that proved effective in marginalising the gender gap. These include the Athena Scientific Women's Academic Network charter, new recruitment and advertisement strategies, setting up a ‘Research and Family Life’ forum, establishing mentoring circles for women and projecting female role models.
- EDUCATION & TRAINING (see Medical Education & Training)
- MEDICAL EDUCATION & TRAINING
Statistics from Altmetric.com
The traditional discrepancies between women and men in clinical medicine have dropped dramatically over recent years. The data speak for themselves. In the USA, the percentage of female medical students has seen an unparalleled increase from 5% to 50% between 1960 and 2006.1 A similar—though less pronounced—pattern has been observed in the UK, where the number of female medical students rose from 25% in the 1960s to 61% in the mid-2000s.2 Also, looking at this trend after graduation from medical school, there is no difference between men and women in terms of career progression in the National Health Service. However, a particular caveat applies to this last statement, as this is only valid when men and women work full-time—and this is rarely the case for women.3
Despite those impressive improvements, this does not reflect the situation in academic medicine. As evident from an editorial in The New England Journal of Medicine, in 2005 only 11% of full professors were women and only 15% of departmental chairs in the USA were women.1 This represents a massive disparity in women's representation in academic medicine as compared with men in the USA.4 From a UK perspective, the situation appears no better. The more senior the academic position, the smaller the percentage of female representation (42.3% of lecturers, 30.1% of senior lecturers/readers and 15.1% of professors).5 The proportion of women classified as ‘grade A’ researchers (highest calibre researchers based on an international classification system) is merely 23%.6 Men often have superior research performance than women, when using measures such as number of publications, number of citations, research funding (grants) received and remuneration.2 ,7–10 This may be due to systematic discrimination that makes it challenging for women to access resources, and may somewhat be a consequence that women may collaborate less than men and have less established collaboration networks.11 Similarly, gender differences exist in mentoring relationships within academic medicine.12 Men are more likely than women to have positive mentoring experiences that impact their careers.12 ,13 Women are more likely than men to identify lack of mentoring as an obstacle to their promotion, and to consider that same-sex mentors would be more supportive.12 ,14 ,15
Despite these negative findings, there is some encouraging evidence for women in academic medicine. Overall, there has been an increase in the number of female first authors from 11% to 37% and female last authors from 12% to 17% between the 1970s and 2004, respectively.16 Also, in specialties where women are more represented, female authorship has shown a greater rise. Such specialties include paediatrics and obstetrics and gynaecology where the trend has been observed both in the USA17 and the UK.16
This profound disparity of women's representation in academic medicine does matter for three interconnected reasons. First, lack of women in academic medicine results in wasted talent because many women would be keen to add value in this sector.3 Second, female under-representation in academic medicine will inherently bias today's research outline and as a result future clinical practice.3 Finally, the disparate representation of women in academic versus clinical medicine and other fields may reveal a hidden agenda potentially representing a recruitment bias in influential research positions and esteemed specialties.3
In this study, we aimed to evaluate gender disparities in research performance, while considering factors such as mentoring and scientific collaboration, in one of Europe's largest Academic Health Science Centres (AHSCs). Our study used a mentoring perception survey, as well as bibliometric and social network analysis to answer gender-related research questions raised by another recently published study.18 We also discuss various policies and implementations that have been used at the AHSC to tackle the ‘gender gap’ in academic medicine.
Organisational context and sample population
Imperial College London is a public university based in the UK. The university specialises in science, engineering, medicine and business, and has been consistently ranked among the top 10 in the world.19 The Faculty of Medicine (FoM) is one of Europe's largest medical institutions in terms of size of the workforce, number of students and generation of research income. The publicly available university database was used to generate a list of all the professors in the FoM. Their demographic information were extracted including first name, surname, gender and the affiliated division or department for each of the participants included in the study.
Generation of the publication and citation database
SciVerse Scopus (Elsevier, Amsterdam, the Netherlands) Author Identifier was used as a tool to generate, for each professor, the publication list and corresponding citations.20 For each professor, the following research performance measures were identified: total number of publications, total number of citations and the h-index. The study time period was from 1 January 2006 to 31 December 2012. The citations used were collected in March 2016 in order to allow sufficient time for the most recent publications to be cited. To adjust for individual productivity, for each participant the total lifetime publication record preceding the study time period was identified.
Mentoring survey instrument
The mentoring survey instrument administered in this study was used with permission given by the Coalition of Counselling Centres (CCC)/The Mentoring Group, one of four divisions of a not-for-profit corporation: the CCC (Grass Valley, California, USA). The source document for this survey was: Skills for Successful Mentoring: Competencies of Outstanding Mentors and Mentees.21 The survey instrument was based on the conceptual model shown in online supplementary figure S1, which illustrates the shared core skills used by both mentors and mentees, and the unique skills needed by each group. The survey's research documentation offers evidence for construct validity and is presented in detail in the online supplementary appendix.
Generating the network for analysis (measures of scientific collaboration)
The individual publication files for each professor were inserted into a unified list, which was saved as a comma-separated value (.csv) file. Each of these files was loaded into a network analysis software named NetworkBench (Network Science Center, Indiana University, Bloomington, Indiana, USA).22 The co-authorship network was then created. The nodes of the network were the authors, and links were assumed to exist between two nodes when the corresponding authors had co-authored one or more scientific publications. Degree, betweenness and Eigenvector were used as measures of network centrality. Local clustering coefficient and constraint were used to quantify network social capital.23
The position of an author in the co-authorship list is an indicator of each individual’s contribution to the publication task (citation).24 ,25 These were categorised in six position-related variables (see online supplementary appendix).18 ,23
Descriptive statistics were used to present demographic, research performance, network and authorship characteristics as well as mentoring variables from the survey. Each variable was assessed for normality assumptions using Kolmogorov-Smirnov test. Bivariate correlation (Spearman's rank for non-normal distributed data) was subsequently used to determine the relationship between independent variables (demographic, networking, authorship and mentoring) and dependent variables (research performance) and these were presented graphically in a correlation matrix. Data that followed a non-normal distribution underwent logarithmic (log10) transformation before using models of univariate and multivariate regression analyses to assess associations between independent and dependent variables. Finally, Mann-Whitney U test was used to establish if there were any gender differences between dependent variables such as research performance and independent variables such as mentoring survey scores. Statistical analysis was performed using IBM SPSS Statistics software V.18.0 and Microsoft Excel 2010. Statistical significance was considered where p<0.05.
A total number of 104 responses were received out of 215 professors that were surveyed, which equated to a 48% response rate. Out of the 104 respondents, 70 were males (67%). Online supplementary table S1 illustrates the thirty six statements from the survey instrument used to self-assess mentoring skills, online supplementary table S2 how the total scores for quality and frequency of mentoring skills were calculated, table S3 summarises the descriptive statistics for the demographic, research performance, network and authorship variables, and online supplementary table S4 shows the same data when the cohort was divided into male and female groups. The same applies to online supplementary figures S2 and S3, respectively, where pie charts show the distribution of individual and institutional variables expressed as percentages (%, n=104).
Male respondents had a significantly higher number of previous publications than females (mean 131.07 (111.13) vs 79.60 (66.52), p=0.049). There were no other statistically significant differences with regards to demographic, research performance, network and authorship variables.
The distribution of survey scores for males and females was similar for the quality and frequency of shared core, mentor-specific and mentee-specific skills. Figure 1 illustrates a box plot showing distribution of total survey scores for quality and frequency of mentoring skills and gender (male=blue, female=pink). Figure 2 shows radial charts highlighting ‘subtle’ differences between survey scores and gender, although none of these differences was statistically significant.
Online supplementary table S5 shows the results of the univariate regression analyses between measures of research performance and individual, scientific field, social capital and authorship position variables.18
The statistically significant results from the univariate regression analyses were encompassed to build multivariate models (36 in total) taking into account multicollinearity issues (variance inflation factor>10).
Initially, multivariate models were constructed to determine any relationships between the three measures of research performance and the quality and frequency of mentoring skills. Based on these, further multivariate models were generated by including gender as an independent variable. In all multivariate models, there were no statistically significant relationships between quality or frequency of mentoring and research performance.
Moreover, multivariate models were created for each mentoring skill that had a statistically significant association with research performance in the univariate analysis including gender as an independent variable. Again, there were no statistically significant relationships between total survey scores for quality or frequency and research performance. However, model 13 (see online supplementary table S6) demonstrated that the quality of managing the relationship had a statistically significant positive influence on number of publications (B=1.100, SE=0.382, p=0.007) even when gender was included in model 23 (see online supplementary table S7) (B=1.203, SE=0.386, p=0.004). Model 18 (see online supplementary table S8) demonstrated that frequency of providing corrective feedback had a statistically significant positive influence on number of publications (B=0.700, SE=0.277, p=0.016) even when gender was included in model 28 (see online supplementary table S9) (B=0.701, SE=0.288, p=0.020). Model 34 (see online supplementary table S10) demonstrated that frequency of building trust had a statistically significant positive influence on number of publications (B=0.761, SE=0.340, p=0.032) even when gender was included in model 37 (see online supplementary table S11) (B=0.705, SE=0.343, p=0.048).
Network diagrams were created using specialised software for picturing large-scale networks (http://visone.info). The size, shape and colour of the nodes representing the authors was changed in the graphic, so that several attributes such as gender, research performance and quality of mentoring could be visualised. The biggest connected section of the co-authorship network of healthcare researchers in the FoM, Imperial College London is illustrated in figure 3.
This is the first study in healthcare research to investigate the relationship between mentoring perception, scientific collaboration and research performance in the context of gender. This research area is crucial for personal and organisational development and relies on the social and managerial skills of the academic healthcare professionals rather than just their scientific skills.
The most important finding from this study relates to the gender gap at Imperial College London, which was found to be ‘subtle’. No significant qualitative or quantitative differences were found between males and females in terms of mentoring skills and mentoring perception (quality and frequency). The same applies to research performance and social capital.
It is important to understand how this may be achieved. The mainstay of this success is thought to relate to a fundamental change in the workplace ethos through promotion of ‘a culture conductive to women's academic success’. The literature has clearly shown that when support in the workplace is low, women will underperform irrespective of the work demands imposed on them.26
It becomes apparent that in this type of situations the first thing to concentrate in changing is not the working hours but the workplace culture to ensure this is as supportive as possible to female academics. Identifying those cultural barriers and systematically addressing them can only achieve this.27 In Imperial College London, a number of measures have been introduced over the recent years to promote this culture change. These led to the prestigious Athena Scientific Women's Academic Network (SWAN) Silver Award showing organisation-wide commitment to gender equality across Science, Technology, Engineering, Medicine and Mathematics.
Various policies and implementations were introduced. Data from central college records were retrieved and analysed. Following this, the Academic Opportunities Committee (AOC) organised workshops across college campuses to hear the views of FoM staff and PhD students on culture and organisation, career progression, development opportunities, mentoring, working environment, communication and recruitment-retention-progression of research/academic women. The findings were discussed, consulted on best practice, specific actions that could be taken identified and implementation began. The AOC Chair presented the findings and recommendations to the Management Board, who agreed with all recommendations. The FoM members were then consulted on the recommendations through an online survey and campus visits by the Head of Department.28
As an example, a negative finding related to the falling percentage of female academics with progressively increasing seniority. Numerous actions were implemented to address this. A register was established for routinely recording number of clinical and non-clinical career development fellows. Following this, an anonymous survey was launched to identify the reasons for this with questions such as “What do you think are the reasons for the steady decline of female academics with progression?” and “What can the department do to help?” Subsequently, changes were implemented to improve recruitment of females. Changes included having a good female-to-male balance among those featured in a video clip on the website telling what it is like to work in the FoM, switching to an open style of advertisement and putting applicants in touch with female members of staff. Events were also organised by the Postdoc Development Centre such as ‘Aiming for a Lecturership’ or “Planning for an Academic Career and Interview,” which were well attended by the departmental postdocs, 68% of them were female. Refreshments followed each of these events to provide opportunities for networking.28
A mentoring scheme for researchers and academic staff was also established, ensuring that the mentoring team included a good number of females. Moreover, an annual ‘Research and Family Life’ panel discussion was introduced. For every new female academic staff starting in the department, an informal meeting was arranged soon after their arrival to give them the opportunity to discuss about life and progression and put them into contact with other female academics.28
As a testament to the success of the above interventions in addition to the Athena SWAN Silver Award, in the 2011 Imperial Staff survey, 93% of female and 92% of male academic staff reported they felt they had been offered fair and equal access to training and development.28
A meta-analysis assessing the effect of gender in the mentoring literature showed that gender effects may occur more often among mentors than among mentees.29 The findings of this study contradict this meta-analysis because there was no difference in mentoring perception (frequency and quality) between males and females, which is also in accordance with the study by Ragins and Cotton that assessed the ‘willingness to mentor’ among genders.30 The same study showed that rank is a predictor of ‘willingness to mentor’ and that high-ranking individuals have stronger motivation to participate in mentoring activities.30 In the current study, female professors showed similar perception that can be a mentoring resource for organisational success because women are exposed to less opportunities compared with men in terms of: (1) receiving mentoring, (2) providing mentoring, (3) high ranking positions and (4) experiencing time constraints.30
Strengths of the study
The main strength of this study is first the extensive variety of confounders of research performance included. Second, the fact that quantitative measures of social capital were calculated, which allowed relevant adjustments in multivariate analysis. Third, the sample size was considerable (ie, more than 100 professors from one of the top 10 universities in the world), which emphasises the importance of this study for academic healthcare sectors in high-income countries. Finally, it is important to note that the study took place in the healthcare sector and in particular within an AHSC where, to our knowledge, no similar analysis has ever been conducted. It involved a true multidisciplinary project involving a team consisting of clinicians, statisticians and academics.
Limitations of the study
The most important limitations of this study are that it is based on self-perceptions of mentoring and there is a potential for selection bias. A more robust study would include a 360-degree mentor evaluation. In addition, the survey instrument has not been extensively validated in the academic and healthcare sector. Finally, the outcome measures for research performance, such as number of publications, citations and h-index, are not the only metrics that can be used to characterise performance. Alternative performance measures for use in future studies include: financial parameters (nominated salary, bonuses and royalties), grant income, nominations for senior Medical Research Council and Wellcome Trust fellowships, patents and innovation metrics, which are all representative of global performance for knowledge management organisations and have implications for business and wider societal impact and organisational resilience.
This is the first study in healthcare research to investigate the relationship between mentoring perception, scientific collaboration and research performance in the context of gender. It illustrates that though there still is a gender gap in academic medicine, a series of initiatives can prove effective in marginalising it. These include the Athena SWAN charter, new recruitment and advertisement strategies, examining women's experiences of academic appointments and promotions processes, exploring clinical and non-clinical academic trajectories, setting up a Research and Family Life forum and establishing mentoring circles for women and projecting female role models. These findings may offer guidance on patterns of scientific collaboration to improve research performance and show other AHSCs, where the gender gap remains wider and how to address those differences in their respective institutions.
Despite improvements, the ‘gender gap’ in academic medicine remains.
This gender gap consistently favours male over female academics.
Changing the workplace ethos towards ‘a culture conductive to women's academic success’ is fundamental and should be the first step in addressing the gender gap.
Successful strategies include new recruitment and advertisement strategies, setting up a ‘Research and Family Life’ forum, establishing mentoring circles for women and projecting female role models.
Current research questions
Are there any gender differences in patterns of collaboration such as interdisciplinary co-authorship?
How can the strategies described to marginalise the ‘gender gap’ in academic medicine be implemented on a larger (nationwide) scale?
Is there variation in the gender gap in academic medicine between high-income and middle-income countries?
Is there any evidence of a ‘racial gap’ in academic medicine?
Professor Sian Harding is Institute Lead for Women for Athena Scientific Women's Academic Network, London, UK. Professor Sevdalis is funded by the National Institute for Health Research via the ‘Collaboration for Leadership in Applied Health Research and Care South London’ at King's College Hospital NHS Foundation Trust, London, UK. The views expressed in this article are those of the authors and not necessarily those of the National Health Service, the National Institute for Health Research or the Department of Health.
Contributors TA, VP, GG and HA contributed to the conception and design of the study, the analysis and interpretation of data and the drafting of the paper. SP contributed to the conception and design of the study, and the interpretation of data. LH and NS contributed to the conception and design of the study, and to the acquisition of data. SH and AD contributed to the conception and design of the study, and to data analysis and interpretation. All authors contributed to the critical revision of the paper, interpretation of the data and approved the final manuscript for submission.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.