Mentorship has been defined as “a dynamic, reciprocal relationship in a work environment between an advanced career incumbent (mentor) and a beginner (protégé), aimed at promoting the development of both.”1 Mentorship is a critical component of career development in medicine and other professions.2–13 In medicine, the importance of mentorship has been described across the professional lifespan.11 Nevertheless, only a few studies have quantified the impact of mentorship on subsequent career development. These studies have shown that junior faculty members with research mentors during their training or in their current positions felt better prepared to conduct research,10 were more productive in research,2,5,10 and were more satisfied with their careers.5,10,13 Even these studies typically assessed a limited range of career development issues, and provided little information about the elements of mentorship that might affect subsequent professional success.
In a previous article, we reported that, among other factors, influential mentorship during training was associated with early research productivity in graduates of a national, primary care research fellowship program.14 In the current article, we describe the nature and consequences of mentorship in this program in more detail. We posed three research questions: What characteristics of research fellows and their training environment are associated with having influential and sustained mentorship?; How do the subsequent careers of fellows with influential and sustained mentorship differ from those with no influential mentor, or those whose mentorship is not sustained?; and What are the attributes of influential mentors?
In the Health Research Extension Act of 1985, Congress required that .5% of funds from the National Research Service Award (NRSA) training program administered by the National Institutes of Health (NIH) be allocated to primary care research.15 Responsibility for this program was delegated by NIH to the U.S. Health Resources and Services Administration (HRSA) in 1988, and ten institutional NRSA programs within academic centers in the United States were funded.15 In 1993–1994, eight renewal awards were made to institutions in the original funding cycle and 15 new programs were funded. In 1998, 28 awards were made for a third five-year cycle, 20 to existing programs and eight to new institutions. All but one of these institutions are schools of medicine; the other is a dental school. Most programs enrolled two or three individuals per year from one or more of the three primary care disciplines (family medicine, general pediatrics, or general internal medicine). Some also enrolled individuals from other medical specialties (psychiatry, rehabilitation), other clinical fields (nursing, dentistry), or nonphysicians with research interests in primary care. The median duration of fellowship training is two years. Most programs support the completion of an advanced degree, such as a Masters in Public Health (MPH). The organization of these programs has been described.16
The design of our survey has been reported previously.14 In brief, all 215 individuals who received HRSA support between July 1988 and June 1997, as identified by HRSA records or the individual NRSA programs, were eligible for the survey. The survey questionnaire was administered by mail in June 1998, with two follow-up mailings at intervals of three weeks, followed by attempted phone or e-mail contacts with individuals who had not replied.
We developed a self-administered questionnaire to assess the demographics of HRSA/NRSA trainees, their duration of training, degrees obtained, and time allocated to course work, research, and other activities during the fellowship. Fellows characterized the educational model of their fellowship as an “apprenticeship” in which the trainee worked on a project closely related to a senior investigator or a strategy of “early independence” in which the trainee developed her or his own research idea and identified advisors to provide input.3 Fellows also described their current professional position, including their employer, academic affiliation and rank, time allocation, hours worked per week, current affiliation with the institution where they had completed the NRSA program, their role on up to three current research projects, and the number of first-authored and coauthored publications since the beginning of fellowship. To validate self-reports of the number of publications, we asked respondents to provide a curriculum vitae. Among the 111 respondents (76%) who did so, there was a strong correlation between self-reports of the number of research publications and the publications listed in the vita (r = .92, p < .001).14
We also assessed two characteristics of the training program that might influence research productivity: the number of NRSA funding cycles that the program had received (either one or two cycles), and whether the program trained clinicians from a single specialty (such as general pediatrics) or multiple clinical disciplines.
Components of Mentorship
We assessed mentorship in several ways. First, we asked the fellows if they had a mentor during training. Among those who did, we asked about the total number of mentors, the hours per month spent with mentor(s) individually and in group settings, and if they were continuing to receive sustained guidance from the mentor at the time of the survey. We also asked, “Of the people you thought of as mentors, was there one who was particularly influential?” Respondents who identified an influential mentor were asked to describe, in an open-ended response, how this person was influential. Finally, we asked if they currently provided mentorship for other individuals, the number of such individuals, and the professional areas in which they provided mentorship.
Most questionnaire items were newly developed for this survey, although some questions were adapted from prior studies.17,18 The survey was approved by the University of Colorado Multiple Institution Review Board.
We categorized respondents into three groups for the current analysis: those without an influential mentor, those with an influential mentor during fellowship who indicated that this relationship was not sustained at the time of the survey, and those with both influential and sustained mentorship. To identify characteristics that were associated with these categories of mentorship and to assess the impact of such mentorship on the subsequent career, we performed three-group comparisons using Kruskal-Wallis tests for continuous variables, and Mantel-Haenszel chi-square tests (tests for trend) on categorical variables.
We also identified five attributes of the current professional career that were likely to be associated with success in research: a position on an academic faculty, spending 40% or more of professional effort on research, providing research mentorship to others, publishing one or more papers per year, and having a federally funded grant as principal investigator. We conducted chi-square tests or Wilcoxon rank sum test to identify characteristics of the fellows or their training programs that were associated with each of these five outcomes at a statistical significance of p < .10. All variables identified in bivariate analyses were entered, along with the three-level mentorship variable, into logistic regression models. Backwards elimination identified all variables that were independently predictive of each outcome. These variables and the mentorship variable were then entered into a final logistic regression model. The predictive ability of each model was assessed using the c-index, a statistic that represents the probability that a randomly chosen individual with the outcome of interest is correctly classified by the model, compared to an individual without that outcome. We performed all analyses using the Statistical Analysis System (SAS) version 8.1 (SAS Institute, Cary, NC).
For the qualitative analysis of mentorship domains, one of us (DSM), an experienced qualitative researcher, reviewed the responses to the open-ended questionnaire item about the ways in which a mentor had been influential, divided the responses into codeable text segments, and identified themes based on those responses. One of us (JFS) independently reviewed papers that described the qualities of mentors and identified domains of mentorship.2–13 DSM and JFS then agreed on a common set of mentorship domains, independently coded text segments, discussed disagreements, and recoded. The intercoder agreement (calculated as the number of identical codes divided by the total number of text segments) was 91%.
Of the 215 individuals who participated in the HRSA/NRSA program between July 1988 and June 1997, 146 (68%) completed the questionnaire. Using HRSA administrative data, we found that respondents and nonrespondents did not differ in age, race or ethnicity, or clinical discipline. The response rate was 63% for male program graduates and 75% for female program graduates (p = .06). No other data on nonrespondents were available.
Of the respondents, 51.0% were men and 84.1% were white/non-Hispanic. Their mean age was 38 ± 5 years. General internal medicine was the most common discipline (36.3%), followed by general pediatrics (30.8%), family medicine (22.6%), and other clinical or nonclinical disciplines (10.3%). These individuals completed 2.3 ± 0.6 years of fellowship training, and responded to the survey 3.9 ± 2.4 years after completion of their fellowship program.
Two participants did not respond to any of the questions about mentorship, while seven did not indicate whether they had an influential mentor during fellowship. Thus, the analysis was based on between 139 and 144 respondents, depending on the question, with a overall response rate to one or more mentorship items of 65%. A total of 134 individuals (93.1% of those responding to the question) reported having a mentor during fellowship. Fellows had a median of two mentors, with a range of 0–6. The fellows spent a median of four hours per month with their mentor(s) individually, and a median of six hours per month with mentors in group settings. Seventy-one respondents (49.3%) reported sustained mentorship at the time of the survey. Ninety-six former fellows (66.7%) reported that they currently provided mentorship to others, with a median of two current protégés (range 0–41). These fellowship graduates most often provided mentorship in research activities (62.5% of those providing mentorship), followed by clinical practice (52.1%), research teaching (33.3%), clinical teaching (29.2%), and administration (20.8%).
One hundred and two respondents (73.4%) indicated that they had a mentor who was “particularly influential.” Few characteristics of the fellows or their training programs distinguished individuals who had influential and sustained mentorship from those with influential but not sustained mentorship relations, or from those with no influential mentorship (see Table 1). Individuals who reported influential but not sustained mentorship had finished their fellowship over one year earlier than the other groups (p = .03). Those who reported influential and sustained mentorship had spent almost twice as much time individually with their mentor during training as either of the other groups (p = .02). Individuals who spent more time individually with their mentor reported spending a higher proportion of their fellowship in teaching than those who spent less individual time with their mentor (12.2 ± 9.6% versus 8.6 ± 7.5%, p = .02) but a comparable proportion of time conducting research (37.4 ± 18.6% versus 37.9 ± 15.6%, p = .86).
In bivariate analyses, influential mentorship, particularly when sustained, was associated with several characteristics of the subsequent careers of respondents (see Table 2). Those with influential and sustained mentorship were most likely to report that their training had prepared them well for their current position (p = .0009), and spent more time conducting research (p = .007). Fellows with influential mentorship, sustained or not, were more productive in publications and in obtaining grants from any source as a principal investigator (see Table 2). The percentage of respondents who provided research mentorship to others increased from 36% among those with no influential mentor to 73% among those with both influential and sustained mentorship (p = .008).
On multivariate analyses, influential but unsustained mentorship was significantly associated only with publishing one or more paper per year (odds ratio [OR] = 4.0) (see Table 3). Influential and sustained mentorship was significantly associated with spending 40% or more effort on research (OR = 2.7), providing research mentorship to others (OR = 8.9), and publishing one or more papers per year (OR = 5.2). Individuals with influential and sustained mentorship also had a twofold increase in the odds of being on a full-time faculty (OR = 2.1) and of being a principal investigator on a federal grant (OR = 2.1), although neither association reached statistical significance. Among the other variables assessed, only spending 40% or more effort on research as a fellow was consistently associated with these key markers of research development, appearing in four of the five models (see Table 3). Unexpectedly, fellows who reported spending more than four hours per month individually with their mentor were significantly less likely, in multivariate analyses, to spend 40% or more of their time in research or to provide research mentorship (see Table 3). A possible explanation for this negative association is that fellows had more intensive interactions with their mentors if they chose to emphasize activities such as teaching during their fellowship, as noted earler.
Ninety-five individuals provided open-ended descriptions of the ways in which their mentor was influential. These responses could be categorized qualitatively into three domains: attributes of the relationship between the mentor and the protégé (such as their advice and listening skills), professional attributes of the mentor (such as their professional eminence), and personal attributes of the mentor (since as kindness and caring). These domains are described in Table 4, with representative attributes from the literature and quotes from open-ended responses. Respondents most frequently indicated that attributes of their relationship with their mentor were influential (69% of those responding), followed by the professional attributes of the mentor (61%) and the personal attributes of the mentor (17%).
Our prior finding that influential mentorship was associated with early research productivity prompted this in-depth analysis of mentorship among graduates of primary care research fellowships.14 Fellows with influential and sustained mentorship reported better career preparedness, more protected time for research, and a higher likelihood of providing research mentorship to others, while influential mentorship, even if not sustained, was associated with enhanced production of papers and acquisition of grants (see Table 2). The importance of mentorship in career development was confirmed using multivariate analyses (see Table 3). The key attributes of mentorship were the establishment of meaningful relationships with the protégé and the professional status of the mentor (see Table 4).
The Domains of Mentorship: A Preliminary Conceptual Model
Although many commentators have described specific characteristics of effective mentors in medicine,2–13 few studies have proposed a conceptual framework to facilitate research in mentorship.19,20 We developed a preliminary model of mentorship domains from the responses to our qualitative question about the attributes of influential mentorship that was corroborated by an independent review of the literature on medical mentorship.2–13 Three domains of mentorship arose from this analysis: attributes of the relationship between mentor and protégé, professional attributes of the mentor, and personal attributes of the mentor. The quantitative analysis added a fourth domain—the amount of time spent individually with the protégé—which was not associated with research success but may be important in other professional activities. Because this survey was not designed primarily to investigate the domains of mentorship, other important components may exist that we could not identify (such as the style or content of mentorship).
The model derived from our study has areas of overlap with other proposed models of mentorship. In a study of university faculty members, Corcoran and Clark19 focused on what we have termed the professional attributes of the mentor, and found that highly productive faculty members were more likely than their less productive colleagues to receive specific assistance from their graduate school mentors in obtaining positions, grant applications, collaboration in research, and critique of writing. Bower and colleagues20 surveyed 18 participants in a junior faculty development program in a department of family medicine about the attributes of their mentors, and found that the most effective mentors were more highly rated in areas such as support, challenge, and guidance—attributes that our model assigns to the domain of the mentor–protégé relationship. Thus, we feel that the model proposed here does provide a preliminary conceptual framework to guide further research in this area.
Although the literature on mentorship in medicine often emphasizes personal characteristics of the mentor such as availability, intelligence, and generosity (see Table 4), these attributes were seldom mentioned by our respondents. This discrepancy may be due to the orientation of the survey toward professional rather than personal development. The finding that individuals with influential but not sustained mentorship have been out of their training for a longer period of time is easily explained by the evolution of fellows toward independence in research or by a diminution in the intensity of mentorship over time.1 The absence of statistically significant associations between influential or sustained mentorship and other easily measurable attributes of the fellow (such as age, race, or gender) or the characteristics of their training program (see Table 1) further emphasizes that the qualities of the mentor and the relationship between mentor and protégé are the key determinants of a mentor's impact.
The Impact of Mentorship on Career Development
The early academic “outcomes” of publication and acquisition of federal grants were comparable between fellows with influential but not sustained mentorship and those with both (see Table 3). However, fellows who had an ongoing relationship with their mentors reported that they were better prepared for their careers, devoted more time to research, and provided more mentorship for others, all potential indicators of long-term research success. Since productivity in publication and grant acquisition over the first few years of a faculty career is likely based on research conceived and conducted during the fellowship, it is possible that, over a longer term, sustained mentorship may also be associated with greater research productivity.
The finding that fellows with influential mentors were substantially more likely, even early in their own careers, to provide research mentorship to others (see Table 3) may be due to the adoption of the attributes of the mentor by the protégé, who models the mentoring process in subsequent relationships.1 Alternatively, fellows with the capacity to form sustained relationships with mentors may also have a stronger inclination to attract protégés of their own. The inverse association between receiving influential research mentorship and subsequently providing mentorship to others in clinical teaching (see Table 2) suggests that the impulse to mentor others is not extinguished in individuals who did not have an influential research mentor, but may be redirected to other professional domains.
Limitations and Implications for Research
Our study had several limitations. Because of the small number of graduates from the HRSA/NRSA fellowship program in primary care research and the 65% response rate to mentorship questions, we lacked the statistical power to identify potentially important findings (such as the lower rate of mentorship reported by women) as statistically significant. Further, the median length of follow-up after the fellowship was only four years, preventing us from assessing any longer-term effects. Since the respondents reflected on their fellowship experience retrospectively, their assessment of their mentors and the influential attributes of mentorship may have changed since they completed their training. In addition, because the fellows did not identify their mentor(s), we could not assess the direct participation of the mentor in the fellow's publications or grants. In the laboratory sciences, the apprenticeship model of fellowship education has been much more common than the “early independence” model adopted by most of these primary care fellows. The impact of a mentor on career development may be different where an apprenticeship model predominates. Finally, an observational study using self-reported survey data cannot establish causal linkages between mentorship and professional outcomes. The ability of research fellows to identify and sustain a relationship with their mentor may be related to other personal or professional attributes that enhance their subsequent productivity in research. In the light of these limitations, future studies of research mentorship should be prospective, larger, longer in duration, more representative of the broad range of medical research, and should use consistent measures to facilitate comparison between studies. Such studies should also be guided by a well-specified conceptual model to ensure that all important domains are assessed.
Implications for Training Programs and Funding Agencies
Our finding that physicians with influential and sustained mentorship were more successful in the early stages of their research careers has important implications for research training programs and the organizations that fund them.21 Influential mentors were professionally successful in their own right, yet willing to establish meaningful relationships with their protégés, spend substantial time individually with them, and encourage them to explore topic areas that may have been outside their own immediate research interests. To accomplish this, they were somehow able to make mentorship a priority despite the many competing time demands and fiscal pressures of academic life. Fellowship programs should identify potential mentors not just on the basis of their professional eminence, but also on their willingness to devote time to their protégés. The programs should instruct fellows proactively about the characteristics of influential mentors, and should establish an expectation with both mentor and protégé for regular, individual meetings. Programs should explicitly acknowledge that, at least in primary care research, a group of mentors with complementary skills, rather than a single mentor, may be necessary to provide all the components of professional development.
While research training programs should attempt to support time for their faculty mentors, the major responsibility for adequate funding of mentorship ultimately falls on the federal agencies and other organizations that support these programs. A recent report of the Institute of Medicine15 and our survey of the directors of primary care NRSA training programs16 have pointed out that the success of federal research training efforts such as the NRSA program may be jeopardized by their lack of support for faculty time. Although newer NIH programs that encourage mentorship, such as K24 midcareer awards, do provide an important mechanism to support mentorship, and thereby help ensure the continuity of the clinical research enterprise in the United States, advocates for clinical research must continue to emphasize that adequate support of faculty time in all training grants can most reliably enhance the mentorship that is essential for the career development of new clinical and health services researchers.
Support for this research was provided by the Bureau of Health Professions, U.S. Health Resources and Services Administration, Contract #97–0452(P)-BHPR.