Mentorship Interventions in Postgraduate Medical and STEM Settings: A Scoping Review

INTRODUCTION

Mentorship is consistently recognized as an essential component of success in postgraduate science, technology, engineering, math, and medicine (STEMM) training and careers (National Academies of Sciences and Medicine, 2020). A seminal report on mentorship from the National Academies of Sciences, Engineering, and Medicine (NASEM) defined mentorship as “a professional, working alliance in which individuals work together over time to support the personal and professional growth, development, and success of the relational partners through the provision of career and psychosocial support” (p. 37) (National Academies of Sciences and Medicine, 2020). This report also acknowledges that mentorship can improve not only professional outcomes, such as research skills and STEM retention, but also psychosocial outcomes like science identity and self-efficacy (Sambunjak et al., 2006; Fleming et al., 2012; Gershenfeld, 2014; Byars-Winston et al., 2018; Atkins et al., 2020; National Academies of Sciences and Medicine, 2020; Estrada et al., 2022). Mentorship in STEMM is particularly important, as STEMM is a broad and diverse field, ranging from biology education to computer science to medicine, with rigorous and lengthy training pathways that require expertise, persistence, and support. Moreover, STEMM careers can be competitive and isolating, further reinforcing the need for mentorship opportunities to enhance professional success and well-being. These characteristics of the field make it so that individuals at all career stages have the potential to be both mentors and mentees and benefit from mentorship (Ghosh and Reio, 2013). Quality mentorship also promotes inclusive environments that help attract and retain underrepresented (UR) trainees in STEMM (Womack et al., 2020; Estrada et al., 2022). Increasing diversity in STEMM is a priority not only for equity in the workforce, but also for innovation and scientific advancements that result from collaborations of diverse teams (National Academies of Sciences and Medicine, 2020). Overall, there has been a significant focus on mentorship research to explore how mentorship can promote success (Montgomery et al., 2022). Part of this research has included the creation of a variety of mentorship interventions (Sambunjak et al., 2006; Akinla et al., 2018; Sheri et al., 2019; Bonifacino et al., 2021).

Mentorship interventions include efforts, activities, or program elements where mentorship is utilized to improve various career, professional development, and psychosocial outcomes (O’Cathain et al., 2019). In biology and biology education, there are several mentorship studies and interventions that support the importance of mentorship for mentors and mentees (Byars-Winston et al., 2015; Lewis et al., 2017; Pfund et al., 2022; Novick et al., 2023). For example, one intervention that paired biomedical trainees with mentors revealed that the most common topics discussed by mentors and mentees were research and career planning (Lewis et al., 2017). This reveals that cultivating mentorship relationships can lead to productive discussions about career, research, and future plans. Another study that examined mentorship outcomes among undergraduate mentees in the biological and life sciences found that positive perceptions of mentor effectiveness were related to PhD/MD enrollment. Mediators of this relationship were research skills, career knowledge, and research self-efficacy, revealing that mentors’ ability to foster these can improve career outcomes (Byars-Winston et al., 2015). Lastly, mentors who took part in a training focused on improving mentorship skills and cultural diversity awareness were reported as having increased cultural awareness, effective communication, and increased availability by mentees (Pfund et al., 2022). These studies describe how mentorship is relevant for biology educators in succeeding in their own careers, as well as for supporting their students. They also reveal the importance of interventions in understanding how and why mentorship is important.

Examining other existing mentorship interventions reveals that interventions vary significantly from one another in components, such as mentee population, mentorship structure, programmatic elements, mentorship training offered, outcomes measured, and study design. In terms of mentee population, traditionally, mentees are less experienced individuals, and mentors are more experienced researchers and educators. However, many career stages in STEMM can benefit from mentorship, which is why interventions exist for mentees ranging from graduate students to faculty (Thomas et al., 2013; Williams et al., 2023). Mentorship structure is the type of configuration that a mentorship relationship takes. A common structure is dyadic, which is a single mentor to a single mentee (Montgomery and Page, 2018). Other forms are nondyadic and include groups (i.e., single mentor to multiple mentees) or networks (i.e., single mentee to multiple mentors) (Montgomery and Page, 2018). Group mentorship is useful when there are fewer mentors and when dynamics between mentees are important (Montgomery and Page, 2018). Network mentorship is important when mentees may need mentors with differing expertise (Montgomery and Page, 2018).

Mentorship interventions also vary in the detail and structure of their curricula. For some programs, this includes mentorship training, which is instruction about important mentorship concepts and skills given to mentors and/or mentees to optimize their mentorship experiences (Sheri et al., 2019; Stelter et al., 2021). In terms of outcomes, different interventions assess a variety of different outcomes, making comparing studies difficult. Lastly, there are many study designs utilized for mentorship interventions. Existing reviews of mentorship interventions show the variety of study designs. This includes pre-post, where mentorship outcomes are measured prior to and after the intervention, quasi-experimental, where there are preexisting intervention and control groups, and randomized controlled trials (RCTs), where there is a mentorship intervention, and for ethical reasons, usually a waitlist control or control group that is permitted to engage in the mentorship they would normally have (Sambunjak et al., 2006; Gershenfeld, 2014; Bonifacino et al., 2021). These different study designs vary in their certainty of evidence with RCTs generally considered most certain evidence, then quasi-experimental, and pre-post. Choosing a study design is a balance between certainty of evidence, logistics, and purpose (Shadish et al., 2003). Overall, there is no comprehensive examination of these components of STEMM mentorship interventions in postgraduate settings (i.e., post-college trainee, graduate student, faculty or professional), so important questions still exist. For example: What groups of trainees are understudied as mentees in mentorship interventions? What types of mentorship structures are commonly utilized? What outcomes are primarily measured? Answering these questions helps better design more effective interventions. As such, it is important to conduct a review of postgraduate STEMM mentorship interventions, in order to understand best practices and identify gaps in the literature.

Reviews of mentorship interventions that do exist have provided essential information informing research and practice. For example, a review of undergraduate mentorship interventions from 2008 to 2012 shows that there was no support for mentors and mentees in 30% of the studies reviewed (support was defined as “training, resource material, the structure of a course, or ongoing supervision for either the mentor or mentee” [p. 384]) (Gershenfeld, 2014). Only 15% of studies had both mentors and mentees receiving support, and 50% of studies had only mentors receiving support. As such, resulting recommendations suggested additional support and training for mentors and mentees in mentorship programs at the undergraduate level. This example reveals a tangible recommendation resulting from the article, highlighting the value of reviews on this topic. In the postgraduate setting, reviews of mentorship interventions are oftentimes grouped by field or population. There are reviews in academic medicine and nursing (Andrews and Wallis, 1999; Sambunjak et al., 2006). Similarly, there are reviews with only early career researchers, first-year medical students, and UR physicians and medical trainees (Boeren et al., 2015; Akinla et al., 2018; Bonifacino et al., 2021). This reduces the ability to understand the entire state of mentorship interventions in STEMM fields, which is important for identifying best practices and gaps. Lastly, some reviews cover short periods of time or were conducted more than a decade ago, potentially leaving out more recent interventions that may provide important information (Andrews and Wallis, 1999; Sambunjak et al., 2006; Boeren et al., 2015). Overall, there is no overarching review of postgraduate mentorship interventions in STEMM, making it difficult to draw conclusions about programs in this setting.

The purpose of this study is to conduct a scoping review that comprehensively explores the state of mentorship interventions in postgraduate STEMM settings to identify program components (e.g., mentorship structures, mentorship training, etc.), best practices, and research gaps. Our results can guide the creation of future mentorship programs. This review is not only to be used by researchers to design and evaluate mentorship interventions, but also by educators who have the potential to be mentors and who may benefit from a broad understanding of the field. Having educators lead the design and implementation of programs, using results from this study, will enhance the sustainability of mentorship programs, which will improve career and psychosocial outcomes among their trainees.

MATERIALS AND METHODS

This scoping review followed the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines (Page et al., 2021). In general, scoping review methodology was utilized, as opposed to systematic review, as our aim was to determine the scope of the literature on our topic and discuss key characteristics of studies rather than address effectiveness (as is often done in systematic reviews) (Munn et al., 2018). We utilized PubMed and ERIC (Education Resources Information Center) as the databases in this review to ensure we would capture articles that may be in both scientific and educational journals. Next, we developed and refined a search strategy using keywords describing target phenomenon, population, and methodology with the Boolean operators OR, AND, and NOT. For example, the search strategy in PubMed was: mentor* AND (intervention* OR program* OR train*) AND (grad* OR doctoral OR PhD OR postdoc* OR postdoc* OR faculty OR “early career” OR predoc* OR predoc* OR student) AND (outcome* OR evaluate* OR impact* OR assess* OR effect*) NOT (“high school” OR “middle school” OR “K-12” OR college OR undergrad* OR school* OR teacher* OR child* OR adolescent*). The search strategy searched titles and abstracts. In addition to the search strategy, we used the articles that met the inclusion criteria to conduct a reference list search (i.e., backward) and forward search in Google Scholar. We followed this process until no additional articles were found, and the search process, which was started in August 2022, was completed in January 2023. The search process was done by the first author and researchers trained by the first and lead author. The first author has experience with conducting review articles and the lead author has significant, published works on the subject of mentorship (Lambert et al., 2020, 2022; Afonja et al., 2021).

The selection criteria for evaluated articles included peer-reviewed articles published in English from 2002 to 2022. Criteria also included: 1) a postgraduate population as mentees, 2) a program where mentorship is a significant focus, and 3) a description of mentee outcomes (quantitative or qualitative) related to the mentorship intervention. Regarding the requirement that mentorship be a significant focus, this excludes articles that mention that mentorship was a component of the intervention, but do not include any additional information describing details of the mentorship component (e.g., how the mentorship relationship was created, the goal of the mentorship component, etc.). This review also excludes programs where mentorship is solely for training on a specific skill, technique, method, or project. Lastly, studies that solely include outcomes related to feedback on the mentorship relationship or mentorship program (e.g., process evaluations) were excluded.

After applying the search strategy in the aforementioned databases, articles were collected and uploaded to Rayyan, an online platform for conducting systematic reviews. After duplicates were removed, the screening process began. Two reviewers independently screened titles for inclusion criteria, then abstracts, and then article full texts. If there were any disagreements, another member of the research team evaluated the full text to make a final decision. The entire process is summarized in Figure 1. For the final articles, the research team summarized: study design, setting, intervention details, sample details, and outcomes (i.e., outcomes were categorical in nature). The certainty (or quality) of evidence (including the risk of selection bias) (Shadish et al., 2003) was determined by the study design and using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) approach as a guide. Because this study focuses on mentorship interventions, we sought to use a certainty of evidence evaluation method focused on interventions (Guyatt et al., 2008). In the GRADE approach to assess certainty of evidence, RCTs without critical limitations constitute high-certainty evidence. Observational, prospective cohort studies, and RCTs with significant limitations constitute low-certainty evidence. Additionally, methods, such as sensitivity analysis and statistical heterogeneity which are often applied to reviews of clinical trials, were not utilized in this project due to the nature of the review and the variability of the study designs. Lastly, this review was not registered, and a formal protocol was not prepared.

RESULTS

The search initially revealed 2379 articles and 51 duplicates were removed, yielding 2328 unique articles. There were 109 articles (101 studies) that met the criteria for this scoping review. Table 1 displays information about all studies that met the criteria for the scoping review. Of all articles, 12 (11.0%) were published between 2002 and 2009, 69 (63.3%) were published between 2010 and 2019, and 28 (25.7%) between 2020 and 2022. Most studies (81.2%, n = 82) were set in the United States. The other settings were as follows: Australia (Chanchlani et al., 2018), Canada (Jefferies and Skidmore, 2010; Morrison et al., 2014; Zhang et al., 2017; Croke et al., 2021; Mahmood et al., 2021; Rajendran et al., 2022), China (Lin et al., 2015; Zhang et al., 2019), Europe (Rodríguez-Carrio et al., 2020), India (Singh et al., 2014), Latin America and Asia (Santos et al., 2020), Lesotho (McGuire et al., 2020), Sweden (Jangland et al., 2021), and United Kingdom (Dutta et al., 2011; Hunsberger et al., 2013; Eisen et al., 2014; Ahmed et al., 2018; Wright et al., 2020). Interventions were affiliated with the following institutions: 53.5% (n = 54) medical centers, 29.7% (n = 30) universities, 9.9% (n = 10) professional organizations, and 7.9% (n = 8) were nationwide programs that did not sample specifically from any type of institution (one study sampled from both medical centers and universities). The most common field of study (as identified by article authors) was medicine (63.4%). Other fields are displayed in Figure 2. In terms of study design, posttest only was the most represented design comprising 34% (n = 37) of articles. Only 5.5% (n = 6) of articles were designed as RCTs. A specific intervention length was not stated for 26.7% (n = 27) of studies. For those that did have a formal length (n = 74), 51.3% (n = 38) were 1 year, 24.3% (n = 18) were less than 1 year, and 24.3% (n = 18) were greater than 1 year. Other findings from this review are organized by topic and discussed further below.

Mentee Population

Most studies were designed for a specific mentee population with 45.6% (n = 46) of studies stating they were for faculty, 12.9% (n = 13) of studies stating they were for medical students, 15.8% (n = 16) for physicians/residents, 11.9% (n = 12) for nurses, 6.9% (n = 7) for graduate students, 6.9% (n = 7) for early career scholars (or early-stage scholars), and 4.0% (n = 4) for postdoctoral scholars (some studies stated they were for more than one mentee group; details in Table 1). Of all the studies, 18.8% (n = 19) were specifically for UR groups (Files et al., 2008; Buchwald and Dick, 2011; Dutta et al., 2011; Harris et al., 2012; Varkey et al., 2012; Welch et al., 2012; Mayer et al., 2014; Jean-Louis et al., 2016; Lewis et al., 2016, 2017; Margherio et al., 2016; Ockene et al., 2017; Vermund et al., 2018; Voytko et al., 2018; Petersen et al., 2020; Thorpe et al., 2020; Eiland et al., 2021; Tagge et al., 2021; Jadi et al., 2022; Naidoo et al., 2022; Zheng et al., 2022). Of the 19 studies for UR groups, most explicitly included gender minority trainees (including field-specific gender minorities, such as men in nursing and women in science) and racial/ethnic minority trainees in their definition of UR groups (which were defined by the articles themselves). Only one included sexual minority trainees (Vermund et al., 2018), one included trainees with disabilities (Margherio et al., 2016), and one included trainees from low socioeconomic backgrounds (Lewis et al., 2016, 2017).

Some, but not all studies designed for UR groups, described programmatic elements that addressed UR group membership. Those that did had some notable components. For example, the Native Investigator Development Program (NIDP) was a career and research mentorship program in the United States that required Native investigators to work on Native health projects alongside a mentor (Buchwald and Dick, 2011). Other programs had workshops, panels, or directed conversations that discussed race and gender issues (Pololi et al., 2002; Varkey et al., 2012; Welch et al., 2012; Voytko et al., 2018). Lastly, another U.S. program delineated that one of the mentor responsibilities was to link their UR mentees with other groups, such as “women-at-risk scientific committees and the HPTN [HIV Prevention Trials Networks] Black Caucus, a community/investigator consortium that focuses on racial/ethnic and sexual orientation issues” (Vermund et al., 2018).

Mentorship Relationship

Structure.

The mentorship relationships in these interventions took many different forms. The most common structure in this review was dyadic mentorship, which is one-on-one mentorship between a single mentor and mentee (National Academies of Sciences and Medicine, 2020), comprising 53.7% (n = 54) of the studies. Dyadic mentorship in these studies was mostly vertical (i.e., between a less experienced mentee and a more experienced mentor). Examples of these types of relationships were junior graduate students with senior graduate students, medical students and residents, postdoctoral researchers and faculty, and early career faculty and senior faculty. Other dyadic mentorship relationships were peer-based (i.e., horizontal), where both participants in the relationship were at similar training levels, and there is reciprocity in mentoring (Eisen et al., 2014; Greene, 2021). The participants in these programs were early career faculty, graduate students, and residents.

Another structure that was seen in this review was group mentorship, which is defined as mentorship delivered to “more than two people in which the interactions were simultaneous and collaborative” (Huizing, 2012). This was a part of 26.7% (n = 27) of studies. The most common form of group mentorship was peer group mentorship (i.e., horizontal). In peer group mentorship, a group of peers would meet to provide mentorship to each other. In some cases, there were peer group facilitators/mentors who were senior peers that would provide some vertical mentoring (i.e., senior PhD students [Lewinski et al., 2017; Bôas Fávero et al., 2018] or senior faculty [Files et al., 2008; Varkey et al., 2012; Mayer et al., 2014; Fleming et al., 2015; Ockene et al., 2017]), but senior mentors met at the same time as the peer group. In other articles in this review, group mentorship was more focused on vertical mentorship (Singh et al., 2014; Hall et al., 2018; Santos et al., 2020; Womack et al., 2020). One common format of this was coaching programs (Williams et al., 2016a, 2016b, 2017; Thorpe et al., 2020; Womack et al., 2020). Coaches were independent, external parties that received mentorship training, and coaches in these programs met with groups of trainees to provide an intensive level of mentorship (Womack et al., 2020). Although vertical mentorship was implied as the main goal, there may have been peer mentorship that also took place.

The last type of structure that was identified was network mentoring (20.8%, n = 21 studies). This includes a single mentee at the center of the mentoring relationship, and the mentee has multiple mentors with differing roles (National Academies of Sciences and Medicine, 2020). In one form of network mentoring, a mentee has several senior mentors (Eiland et al., 2014; Hickey et al., 2014; Kohn, 2014; Byington et al., 2016; Chen et al., 2016; Libby et al., 2016). Examples of these are the Robert Wood Johnson Foundation Nurse Faculty Scholars program in the United States where mentees had three mentors that they met with: “a primary nursing mentor (a senior nurse scientist/leader) in his or her school of nursing, a senior non-nurse researcher outside the school (but within the applicant’s home institution), and a national nursing mentor” (p. 2) (Hickey et al., 2014). In another program, the mentee had four types of mentors: senior, scientific, peer, and staff mentorship (Byington et al., 2016). The I³ Mentoring Program based in Indiana was another unique program that concentrated on one-on-one mentorship from a senior faculty “super mentor,” but also included a “super mentor” committee, a professional grant writer, and a biostatistician (Santos et al., 2020). Some of these network relationships also focused on peer mentors, as well as senior mentors (Jefferies and Skidmore, 2010; Porat-Dahlerbruch et al., 2021; Naidoo et al., 2022). One example was called the constellation model where a variety of mentorship relationships were facilitated for the mentee, including faculty-student mentorship, one-on-one peer mentorship, and peer group mentorship (Porat-Dahlerbruch et al., 2021). Lastly, some programs did not provide information about who was in the mentoring network (Jean-Louis et al., 2016; Zhang et al., 2017).

Program Components.

The program components of mentorship interventions varied. We categorized interventions as 1) only facilitating the mentoring relationship (e.g., leaving it up to participants how to operate in the relationship) or 2) facilitating the mentoring relationship and requiring additional activities. Out of all studies, 48.5% (n = 49) focused only on facilitating the mentorship relationship. On the other hand, 51.5% (n = 52) also had additional required activities for participants. One notable category of mentorship activities in studies was research training. One program in the United States (two separate studies/sets of participants) had women in the group peer mentorship relationship all work together on a research proposal (Files et al., 2008; Varkey et al., 2012; Mayer et al., 2014). Others had mentees develop individual research projects and/or grants, and a part of the mentor’s responsibilities was assisting with these projects (Buchwald and Dick, 2011; Reid Ponte et al., 2015; Libby et al., 2016; Williams et al., 2016b, 2016a, 2017; Hall et al., 2018; Mazerolle et al., 2018; Padek et al., 2018; Vermund et al., 2018; Vasylyeva et al., 2019; Jacob, et al., 2020a,b; Wright et al., 2020; Brownson et al., 2021). Another category of mentorship activities was workshops (Zeind et al., 2005; Daley et al., 2006; Welch et al., 2012; Eisen et al., 2014; Byington et al., 2016; Chen et al., 2016; Jean-Louis et al., 2016; Spence et al., 2018; Blanco and Qualters, 2020; Prionas et al., 2021). Some programs had workshop series for participants that included general professional and personal development skills, such as stress management, negotiation, and gender and power issues. Other types of activities included mock interviews (Richter et al., 2019), attendance at conferences (Wright et al., 2020), and informal social events (Kohn, 2014). Overall, activities seemed to be viewed favorably by participants based on qualitative results from interventions. For example, one paper noted a participant saying: “The program presented a rare opportunity to ‘get some skills to deal with issues that we are never taught about, such as gender, power, and so forth’” (p. 383) (Pololi et al., 2002). In another study, based at Lenoir Memorial Hospital in North Carolina, several spoke about a specific activity, called “Circle of Trust” “that made it okay to be vulnerable, reveal their fears, and honestly tell others what they were experiencing (p. 237)” (Scott and Smith, 2008).

Mentorship Training

Mentorship training in this review includes training workshops and written materials for mentors and/or mentees to function in their relationships. Mentorship training does not include education that was exclusively about the study itself (e.g., describing/orienting to the program, overview of responsibilities of the mentor and mentee, etc.). In this review, most (50.5%, n = 51) of the programs did not have any training, 15.8% (n = 16) of programs had only mentor training, 5% (n = 5) had only mentee training, and 30.7% (n = 31) had both.

Workshops, which we have defined as any workshops, meetings, or presentations facilitated or delivered by the study team to support participants in their mentoring relationship, were a part of 78.7% (n = 37/47) of mentor training. Workshops included instruction on topics, such as the importance of mentoring, developing mentoring skills, and strategies to create a good mentoring relationship/environment. In terms of mentee training, approximately 66.7% (n = 24/36) of the education included workshops. Mentee workshops mostly focused on developing mentorship skills, exploring personal and professional goals, and brainstorming topics to discuss with mentors.

Mentorship training that took the form of written materials included books, articles, and manuals to support mentors and mentees in their mentoring relationships. Of all mentor and/or mentee training, written materials were a part of 38.3% (n = 18/47) of mentor training and a part of 41.7% (n = 15/36) of mentee training. They were similar in topic area to workshops and usually delivered (electronically or in-person) to mentors or mentees at the beginning of the program. Examples of published written materials for mentors included “Every Other Thursday: Stories and Strategies from Successful Women Scientists” (Petersen et al., 2020) and Cohen’s “Manager’s Pocket Guide to Effective Mentoring” and the “Principles of Adult Mentoring Inventory” (Tracy et al., 2004). Others were more general and included: mentoring literature, “mentor tips,” and “mentee tips” (Quaas et al., 2009) or a mentoring tool kit (Moorthy et al., 2016). One specific written material for mentees included Cohen’s “Mentees Guide to Mentoring,” (Tracy et al., 2004).

Some notable mentorship training that was seen in this review included mentors meeting as a group biweekly with a study advisor to discuss mentorship (Bôas Fávero et al., 2018). In addition, one program at the Medical University of South Carolina had a litany of resources: monthly Tools for Mentors and Mentees series where a faculty member with expertise in mentoring presents on a topic (e.g., balancing work and life, developing a teaching portfolio, preparing grant budgets), an annual Mentorship Training Symposium that focuses on a particular theme (e.g., mentoring in team science, promotion, communication skills, negotiation skills), and a Mentor Training Course offered twice per year to faculty who wish to improve their mentoring skills that was modeled after the University of Wisconsin’s mentor training program for clinical and translational researchers (Bonilha et al., 2019). Another program based at two universities in Korea conducted significant hours of mentor training (12–13 h), and only those who completed the training were eligible to be mentors (Hur et al., 2018). In one program at academic medical centers in New York, a theory of motivation, Social Determination Theory, was used as a foundation for training (Lewis et al., 2016). Lastly, two programs outsourced their mentorship training to external consultants: Christine Pfund for mentor and mentee training (Padek et al., 2018; Jacob, et al., 2020a,b; Brownson et al., 2021) and David Clutterbuck for mentor training (Dutta et al., 2011).

Outcomes of Mentorship

The included articles had different approaches to evaluating the outcomes of the mentorship interventions. Of all articles, 42.6% (n = 43) utilized only quantitative measures, 6.9% (n = 7) utilized only qualitative measures, and 50.5% (n = 51) utilized both. Outcomes were categorized into the following groups: 1) career, 2) professional development, 3) research productivity, 4) satisfaction, 5) psychosocial, and 6) mentorship process evaluation. These groups were based on prior literature (National Academies of Sciences and Medicine, 2020), as well as preliminary work done where we listed all study outcomes and categorized them by theme. Further descriptions of the group are in Table 2. Of all articles, 25.7% (n = 28) measured career outcomes, 58.7% (n = 64) measured professional development outcomes, 23.0% (n = 25) measured productivity, 20.2% (n = 22) measured satisfaction, 38.5% (n = 42) measured psychosocial, and 51.4% (n = 56) captured mentorship process evaluation.

As the interventions in this review utilized a variety of study designs, it is difficult to examine outcome effectiveness and the certainty of evidence. However, we used an adapted GRADE approach to categorize studies as “low” or “high” certainty. Most articles were “low” certainty due to their study designs (e.g., lack of control group, randomization, etc.), and three were considered “high” certainty (Lewis et al., 2016, 2017; Williams et al., 2016b; Prionas et al., 2021). The “high” certainty studies were RCTs and are described below.

The first RCT at 11 institutions in New York included 150 mentor/mentee dyads randomized into four groups: 1) mentor training, 2) peer mentoring, 3) mentor training and peer mentoring, or 4) control (Lewis et al., 2016, 2017). At the end of the program, there was no significant difference in “need satisfaction with mentor” scores between groups. In terms of topics discussed, mentor-training dyads were more likely to discuss both teaching and work-life balance than participants with no mentor training. Peer mentoring participants were more likely to discuss both clinical care and career planning than participants without peer mentoring. Another RCT was based in the United States and examined those who took part in coaching groups (n = 60) and a control group (n = 60) (Williams et al., 2016b). Analysis showed that perceived achievability of a career in academia increased in the coaching group but decreased in the control group. In addition, the perceived desirability of a career in academia decreased significantly less in the coaching group than in the control group. The last “high” certainty study explored in-person mentorship versus telementorship (Prionas et al., 2021). Overall, there were no differences between in-person and telementorship groups in respect to confidence in clinical knowledge, career planning, networking/exposure, sponsorship, and wellness/coping. However, over time, for both groups, there were significant increases.

Additional Innovative Elements of Mentorship Interventions

Through analysis of included studies, some unique intervention aspects emerged, which did not fall into the aforementioned results categories. We have summarized them in Table 3. For example, one unique aspect of one of the included studies was the usage of mentor champions (Bonilha et al., 2019). In this study, mentor champions were “senior faculty who were responsible for leading the mentoring program in each department.” Mentor champions helped those who did not have a mentor facilitate a mentoring relationship. All mentor champions also met quarterly to share information about the mentoring program across departments.

Another aspect of these studies was extra potential benefits (e.g., incentives, compensation, rewards) provided as part of the program. Most studies did not have benefits other than from the mentorship program itself. However, some studies had unique potential benefits. In one study, mentors were compensated and “received a monthly salary equivalent to US$70 per mentee per month” (p. 3) (Lin et al., 2015). Another unique potential benefit was an annual mentoring award established by the program (Bredella et al., 2021).

The program by Zeind et al. also focused on sustainability and incorporated the mentorship program into the university culture, with buy-in from representatives of the Faculty Council and Faculty Development Committee, as well as various academic leaders “(i.e., President, Provost/Vice President of Academic Affairs, Deans, Chairs) and representatives of faculty and student committees and organizations (e.g., promotion committee, faculty council, academic advising, academic support, counseling, etc.)” (p. 4) (Zeind et al., 2005). Lastly, some programs measured elements of mentees’ social network, which is a method to examine the process and potential outcomes of mentorship (Buchwald and Dick, 2011; Brownson et al., 2021).

DISCUSSION

This scoping review fills a gap in the literature by providing a comprehensive assessment of mentorship interventions in STEMM over the past two decades. This review is the first step in creating consensus about the most effective elements of mentorship programs. There are several takeaways from this study that can be used to inform future mentorship efforts at the research and institutional level. These include a need for additional interventions for UR groups and postdoctoral scholars, increased mentorship training (particularly for mentees), further work on the consistency and quality of outcome measurements for mentorship programs, and increased RCTs. We have also developed a conceptual framework (Figure 3) that serves as a guide for designing mentorship interventions and includes components of intervention design that emerged from this study.

This framework is interactive and there are seven steps, with accompanying questions, that can be sequentially followed to outline a mentorship intervention. We were guided by existing theories for mentorship, including social network theory and ecological systems theory (National Academies of Sciences and Medicine, 2020). This framework can be employed by universities, medical centers, and organizations outside of academia. It can also be used to evaluate existing interventions. This can be both at the organizational level or in future academic articles that assess mentorship interventions. Step 1 depicted in Figure 3 describes both the setting of the intervention, as well as the modality. For nationwide interventions, consider how a national sample will be recruited and maintained. In Step 2, the mentee population will be considered. As identified in this paper, a smaller number of articles are designed for UR mentee groups in the field. As such, this may be an important consideration for designing future studies. Step 3 prompts those using this figure to decide the mentorship structure. Once the structure is decided, it will help determine who the mentor(s) are (i.e., faculty members, peers, both, etc.). Afterward, it is important to consider the mentor/mentee match process. There are various ways to match mentors and mentees. They can be paired randomly, (Singh et al., 2014) chosen by the mentor, chosen by the mentee (Caruso et al., 2016, 2019), or chosen by a third party, such as the researchers (Nagarur et al., 2018). It is important to consult articles with similar settings (Table 1) to decide which method to match mentors and mentees. Results of this study reveal the necessity of including mentorship training, especially for mentees, and we describe this further in the “Mentorship Training” section below. Consequently, Step 5 is a prompt to consider what mentorship training will be provided and how it will be delivered. Step 6 builds out the intervention. This step comes after significant consideration of important mentorship program components. In Step 7, outcomes are explored. We encourage the use of existing, validated measures, which we discuss further in the “Outcomes” section below. Lastly, Step 8 focuses on deciding what study design is feasible. Although RCTs were mentioned as a gold standard, in some settings, this may not be possible. As such, a rigorously designed pre-post study, longitudinal, or non-randomized controlled trial may be sufficient. To add rigor to these study designs, it may be important to consider what the intervals of outcome measurement are (i.e., outcomes measured every 6 mo, participants followed up with 1-year postintervention, etc.).

Mentee Population

Results from this study identified gaps in mentorship interventions for certain groups of mentees. One group of trainees where a gap exists is those from UR groups. Only 18.8% of interventions were designed for UR groups in STEMM. This is notable because there are well-documented disparities in career achievements, where UR researchers (compared with well-represented researchers) are less likely to receive funding and be hired as an assistant professor (Ginther et al., 2011; Gibbs et al., 2016). As such, mentorship programs fill an important function for UR trainees. Evidence reveals that they not only help scholars cultivate a robust science identity, which is an important predictor of retention for UR trainees, but they can also offer the guidance, vigilance, and support needed to navigate the “hidden landmines” that UR trainees face along their career paths (Berhe et al., 2021). Among the interventions in our study, the definition of UR groups varied. There were few programs designed for trainees who are gender and/or sexual minorities, from low socioeconomic backgrounds, and with disabilities, though evidence suggests that they face significant barriers to achievement in STEMM (da Silva Cardoso et al., 2016; Hughes, 2018). In addition, few studies mentioned program activities that address UR group membership or had mentor trainings specific for working with mentees from UR groups. One established, evidence-based example of this practice is called culturally aware mentoring (CAM), which is described as “mentoring practices in which mentors recognize their own culturally shaped beliefs, perceptions, and judgments and are cognizant of cultural differences and similarities between themselves and their mentees” (p. 87) (Byars-Winston et al., 2018). Researchers have created a CAM curriculum, and studies of this curriculum’s implementation reveal that mentors have increased confidence, intentions, and skills for addressing cultural diversity in their mentorship (Byars-Winston et al., 2018; Womack et al., 2020). As such, results from this study and research on cultural awareness in mentorship show a need for additional interventions for UR groups with rigorous trainings and activities to create a more inclusive mentorship environment where identities are discussed.

This review shows that mentorship interventions in STEMM have overlooked postdoctoral scholars, with only 5.5% of studies in this review designed for this population. This training stage is defined as a period of mentored research for a recent Ph.D. graduate designed to prepare the researcher for an independent research position. It is often considered a transition point, with a unique set of circumstances unlike any in the academic career pipeline, including: a recent award of a doctorate, temporary position, full-time research, preparation for a research career, the supervision of a senior scholar, and freedom to explore own research (Åkerlind, 2007). Research suggests that it is a training stage where some talented scholars previously committed to academic research careers decide on career paths outside of academia (Lambert et al., 2020). Furthermore, research reveals that this decline in academic research career intentions is more pronounced for UR groups in STEMM (Lambert et al., 2020). If a goal is to diversify the STEMM research workforce to achieve equity and improve national scientific advancement, retaining postdoctoral scholars, especially from UR groups, is important. In this review, only two studies were designed specifically and only for postdoctoral researchers. These studies incorporated components specifically for postdoctoral scholars, like exploring possible career paths, filling gaps in doctoral training, and contract negotiations (Reid Ponte et al., 2015; Rodríguez-Carrio et al., 2020). These may lead to retention in the field, as they better equip postdoctoral scholars with the skills and information to navigate a research career. These studies may serve as a model for increased interventions among postdoctoral researchers, with future research including not only creating additional interventions, but considering postdoctoral scholars who are also from UR groups.

Mentorship Structure

Most mentorship structures in this review were dyadic. This is surprising since there is significant research discussing the benefit of having multiple mentors and a mentorship network (Montgomery, 2017; National Academies of Sciences and Medicine, 2020). These benefits include having more comprehensive mentoring, additional resources and connections, and improved career success (Yun et al., 2016; McBride et al., 2017). In this review, all studies that had mentorship networks were published after 2010. Mentors in these studies were given a variety of roles, including primary mentor, external mentor (outside of the mentee’s institution), scientific mentor, career counselor, biostatistical mentor, writing mentor, or peer mentor. Of particular interest are the mentors with specific roles, such as the career counselor, biostatistical mentor, and writing mentor. These roles were given to mentors with expertise in certain areas, allowing mentees to receive targeted training that is not assumed to be a skillset of a single mentor. These serve as examples of mentor roles that are important for trainee success. Many of these should be noted by primary research advisors and institutions so that mentees can be connected with those who fulfill such roles. Ultimately, this review not only provides evidence of a changing landscape for mentorship structures, but also a collection of studies that utilize mentorship networks as a template for future interventions and efforts.

Mentorship Training

This study revealed the need for increased mentorship training, especially mentee training. The importance of training mentors in STEMM is well understood and has resulted in various mentor training programs and research (Pfund et al., 2006, 2014, 2017; Stelter et al., 2021). However, there has been less of a focus on mentee training (National Academies of Sciences and Medicine, 2020), which is also evidenced by this review. Studies of mentee training that do exist reveals important benefits, such as increased knowledge and skills in managing mentoring relationships (Zerzan et al., 2009; Lee et al., 2015; Risner et al., 2020; Manuel and Poorsattar, 2021). One of these trainings designed for postgraduate (specifically postdoctoral trainees) is called Mentoring Up, and “each module of this curriculum is considered from the perspective of the postdoc as a current mentee attempting to get the most out of their relationship with their PI/Mentor, and then flips to consideration of the concepts as a future and/or current mentor” (Risner et al., 2020). A review of mentor training in medical settings describes how mentor training not only increased communication skills, the ability to work with diverse trainees, mentoring skills, and mentoring effectiveness, but also increased productivity, promotions, and awards (Sheri et al., 2019). By this logic, mentee training may not only be effective in improving the mentorship relationship, but also in providing benefits that may or may not be related to the mentorship relationship itself. Stelter et al. describe priorities for mentor training, and this information can be adapted to also create effective mentee training (Stelter et al., 2021). Future mentorship interventions should consider addressing the gap found in this review, incorporate evidence-based mentee training, and consider evaluating its effects.

Outcomes

There was variability in the types of outcomes measured in mentorship interventions. In this study, the following categories of outcome measurement emerged (career, professional development, psychosocial, productivity, satisfaction, qualitative, mentorship). However, no category was assessed in a majority (over 50%) of articles, revealing low consistency of outcome measurements in the field. We emphasize the importance of collecting outcome measurements that are endpoint outcomes, such as career and professional advancements. We found that many studies in our review examined outcomes related to endpoint outcomes, such as the quality of the mentorship relationship and number of publications. However, it is important to design long-term studies that can examine endpoint outcomes themselves. This would add beneficial information that could be used to support mentorship programs at the policy and funding level. In addition, within each of the outcome categories, there was little consensus about the ways to measure each outcome. In the NASEM report, authors state that low consistency is “often related to a lack of valid measures at various levels or from various perspectives” (p. 138) (National Academies of Sciences and Medicine, 2020). As such, there is a need to utilize common assessment methods. The Center for the Improvement of Mentored Experiences in Research (CIMER) has a service that will assess mentorship programs, and, because mentorship spans many fields, we encourage the creation of a public library where researchers from all disciplines can search for validated measures. We also encourage the use of both quantitative and qualitative methods to assess outcomes. Both numerical and rich qualitative data provide essential information to assess outcomes, refine programs, and apply for funding.

Study Design

Results from this study revealed that there is a need for additional high certainty studies in the mentorship field, such as RCTs that are designed to assess the effectiveness of mentorship interventions. This will promote an evidence base of mentorship best practices, optimizing the future of mentorship programs. Oftentimes organizations have limited resources for mentorship interventions, so using an evidence-based, effective program is important. We acknowledge that there are situations when RCTs may not be possible. When this is the case, we encourage researchers to employ other rigorous study designs. This may include pre-post studies with sufficient sample size/power, quasi-experimental studies with matched groups, quasi-experimental studies utilizing regression discontinuity design, and longitudinal studies that follow mentees for years (Shadish et al., 2003).

Though this study provides essential information to synthesize the mentorship intervention literature, limitations exist. In terms of data collection, we utilized a specific search strategy and acknowledge other keywords may have elicited additional articles that met inclusion criteria. Furthermore, only PubMed and ERIC were utilized as databases for this study, and other relevant articles may have been present in other databases for specific fields. Lastly, there are limitations in each included article’s study designs, methods, and analysis that create overall limitations to the body of evidence collected, and studies must be viewed through a critical lens.

CONCLUSIONS

In conclusion, to our knowledge, this study is the most comprehensive review of mentorship interventions across STEMM. We identified and analyzed over 100 studies, and several themes emerged relating to best practices and gaps in the mentorship intervention literature. These themes include a need for additional interventions for UR groups and postdoctoral scholars, robust mentorship training (especially for mentees), further work on outcome measurements for mentorship programs, and increased use of RCTs. Prioritizing UR groups promotes the retention of a diverse STEMM workforce, which is important for scientific advancement and innovation. Furthermore, more rigorously designed studies with robust mentorship training and consistent outcome measurements can promote the creation of consensus statements about mentorship interventions that will optimize future programmatic efforts. The conceptual framework created from this study’s results can be used by researchers and educators to design and evaluate mentorship interventions in a variety of STEMM settings to promote a stronger and more diverse workforce.

ACKNOWLEDGMENTS

REFERENCES