Most agree that mentoring is crucial for the development of physician–scientists1–8; however, few agree about how best to provide mentoring at all levels of academic medicine. Although mentoring is important, few outcome data exist to endorse any one mentoring approach.9 One study has shown that mentoring is an effective means for attaining academic advancement, but that study favored a “political” model of efficacy, rather than a “performance” model.10 Another has demonstrated that mentorship increased the protégé's subsequent involvement in research, but not necessarily the protégé's number of publications.2 Before one can consider the effectiveness of mentoring, one must define the parameters of the process.
The definition of the mentor–protégé relationship as “dynamic and reciprocal”11 is attractive but may not be feasible in many circumstances, in part because of limited availability of mentors and in part because of the difficulty of ensuring mutual compatibility. The quest for “role models” as mentors can become problematic for the most desirable mentors; sought-after attributes of a particular mentor may not be present in the majority of faculty.12 Also, when mentors are scarce, avoiding the use of a supervisor as a mentor is difficult,13 but such supervisor–supervisee mentoring relationships can create conflicts of interest; the power differential between mentor and protégé can lead to abuse of the ideal dyadic relationship.5 To address a scarcity of suitable mentors, some medical educators recommend peer or collaborative mentoring arrangements, which resulted in some success.14,15 Thus, although there is much enthusiasm for the concept of personalized, facilitated professional development for young academics, there is little in the way of guidance about how to establish effective mentoring.
One critical aspect of the Albert Einstein College of Medicine's Clinical Research Training Program (CRTP), which began in July 1998, is strong mentoring. The two-year training program, funded in part by the National Institutes of Health (grants K30-HL004110 to P.M., 2000–2008; RR RR025748 to Harry Shamoon, MD, PI, 2008–present), entails a didactic course of study in statistics, epidemiology, data analysis, research ethics, grant writing, and scientific communications, and it culminates in a master of science in clinical research methods degree. The capstone of the training program is a thesis project. We refer to CRTP participants as scholars because words such as student or trainee seem inappropriate to describe professionals, including senior-level faculty, who are already advanced in their clinical careers. To ensure that adequate expertise and supervision are available for each scholar, the program requires all scholars to have a mentor as they complete their thesis. Some scholars enter the program with a mentor already selected. The program faculty, who constitute the Steering Committee, review these mentors and, when needed, suggest additional mentors to complement the expertise of the originally selected mentor. For those scholars without mentors, the Steering Committee selects a specific mentor who meshes professionally and scientifically (i.e., has similar research interests).
Beginning in 2001, the program faculty has held separate annual workshops for protégés and mentors (i.e., scholars) to help each group arrive at a common definition of best practices of the mentor–protégé relationship. The program faculty schedule these workshops for scholars within their seminar block. Seminars occur on most Tuesday mornings from 9:30 to 11:30 am for the duration of the CRTP, and they include presentations by senior-level investigators on topics in clinical research (e.g., how to establish an investigative team and how to negotiate for a faculty position) as well as work-in-progress seminars given by each scholar. During the workshop on mentor–protégé relationships, scholars complete a directed exercise, given by one of us (N.S.), in which they describe their expectations for mentors in detail. Scholars next evaluate these expectations critically and develop creative solutions to typical challenges (e.g., handling a very busy mentor or determining when they need additional mentoring). For mentors, a two-hour breakfast meeting (to accommodate off-campus faculty) is held each fall. During this meeting, program changes, general expectations for mentors, and program deadlines are reviewed in a lecture format, followed by interactive discussion of topics including suggestions for program improvement, specific problems encountered with protégés, and overall progress of the clinical research enterprise on the national scene.
To further define and refine the mentor–protégé relationship, one of the program faculty (N.S.) performed in-depth interviews with each protégé in the graduating classes of 2001 and 2002 to determine their perceptions of the quality of and their satisfaction with the mentoring. The author scheduled an appointment for all of the CRTP scholars and interviewed them individually for at least 30 minutes, asking a series of open-ended, nonjudgmental questions about their experience of the mentor–protégé relationship: whether their mentorship was sufficient for the preparation of their thesis, whether they desired additional mentoring, whether they expected to have an ongoing relationship with their mentor after they completed the program, and whether they perceived any problems with their mentors. The author undertook this effort as a quality assurance measure for the program and used the resulting data to help inform the expansion of mentoring within the CRTP. The author took notes at each interview (the interviews were not recorded) and drew general themes from these interviews to elucidate the strengths and weaknesses of the available mentor pool. The results of these interviews have not been published.
Since 2001, at the annual breakfast workshops for mentors, the CRTP faculty have consistently encouraged mentors to increase their involvement with their protégés, and they have worked to make the mentors aware of all the changes and additions to the program. These changes (e.g., advancing the deadlines for IRB approval of protocols, increasing the recommended numbers of meetings between mentors and protégés, and enforcing adherence to the schedule of project completion) have helped develop a cadre of highly experienced mentors who are joined by new, first-time mentors each year.
The program faculty sought to define those characteristics that would be most desirable in an “ideal” CRTP mentor–protégé relationship and to evaluate how closely both mentors and protégés in the CRTP program felt the relationships realized those ideals.
The CRTP faculty prepared an online, anonymous survey for both mentors and protégés of the program at the Albert Einstein College of Medicine. One of the authors (N.S.) developed the initial survey, and the specific questions and responses were refined by the Steering Committee over several meetings into a final document. The final survey was administered through the Informatics and Study Management Unit within the Department of Epidemiology and Population Health. The faculty administered the survey in 2006 to program graduates of the classes of 2000 through 2006 and their mentors. The surveys were available online from August 16, 2006 to January 5, 2007. Mentors and protégés received an initial e-mail to prompt their response (CRTP staff keep e-mail addresses current to track the academic trajectory of all graduates). Respondents were directed to a Web link that provided only the appropriate survey (mentor or protégé). Nonresponders received weekly reminders beginning in September 2006 for up to 12 weeks. The program faculty designed the surveys to contain seven parallel statements (Table 1), to which either the mentor or protégé responded, along with minimal information about departmental affiliation and numbers of publications. Numbers linking mentors with protégés were prospectively prepared by the CRTP staff to preserve anonymity. The Epidemiology Informatics and Study Management Unit of the Department of Epidemiology and Population Health obtained results directly, and staff removed all numeric codes before releasing the data to the investigators. The survey administrators reported to one of us (N.S.) total percent responses of mentors, protégés, and matched pairs. The respondents received no incentives for completing the survey.
Respondents indicated their levels of agreement on a five-point Likert scale (1 = strongly agree, 2 = somewhat agree, 3 = neither agree nor disagree, 4 = somewhat disagree, and 5 = strongly disagree). One additional response was “not applicable.” One of us (A.M.) calculated, using STATA Version 10.0 (College Station, Texas), the percent agreement for each of the seven statements in the survey and generated descriptive statistics for both mentors and protégés. To assess the agreement between paired mentors and protégés, she dichotomized responses into “agree” (strongly agree and somewhat agree) and “do not agree” (neither agree nor disagree, disagree, or strongly disagree), and she calculated kappa statistics to obtain the percent agreement expected above chance, excluding any pair in which either the mentor or protégé responded “not applicable.” The author also calculated kappa statistics stratified (at the median) by total number of mentor publications (median = 73) and total number of protégé publications (median = 5).
The nonparametric Mann–Whitney U test compared the median response scores (1 indicating strong agreement, 5 indicating strong disagreement) for each of the questions among the 50 protégés who responded as part of a complete mentor–protégé pair to the median response scores of the 22 protégés who responded but whose mentor did not. This test also compared the median response scores to each of the questions among the 50 mentors who responded as part of a complete mentor–protégé pair with those of the 11 mentors who responded but whose protégé did not. Chi-square analyses yielded very similar results to Mann–Whitney U tests, and therefore only the Mann–Whitney U test results are reported herein.
The Committee on Clinical Investigation of the Albert Einstein College of Medicine considered this survey to be exempt research.
Characteristics of the sample
Limited descriptive data are available because we designed the survey to be anonymous. We administered the survey to 188 individuals (94 mentor–protégé pairs). Of these 94 pairs, neither the mentor nor the protégé of 11 pairs responded, the mentor but not the protégé of another 11 pairs responded, and the protégé but not the mentor of 22 pairs responded. This resulted in 133 completed surveys—61 completed by mentors and 72 completed by protégés—However, because 13 mentors completed surveys for more than 1 protégé (range of 2–5) and 6 protégés had 2 mentors for whom they completed a survey, there were 40 unique mentors and 66 unique protégés who completed surveys. Of the completed surveys, there were 50 mentor–protégé pairs for which both the mentor and the protégé completed a survey. The overall response rate was 70.7% (133/188), and the response rate for matched mentor–protégé pairs was 50%.
Table 2 lists the mentor–protégé pairs who responded by medical specialty. Mentors overwhelmingly reported their degrees as MDs. One mentor held an MD/PhD, and five had earned PhDs in the fields of epidemiology, neurosciences, pediatrics, psychiatry, and surgery (urology).
Matched pair responses.
Of the 50 matched pairs (i.e., both the mentor and protégé of a pair responded, hereafter simply “matched pair”), ratings for the questions demonstrated an overall high level of agreement that the mentors demonstrated the ideal qualities of a good mentor as mentioned in the survey. Fifty-four to seventy-two percent of mentors and protégés strongly agreed with all of the statements except 6 and 7. More than 50% of mentors also strongly agreed with Statements 6 and 7, whereas more than 50% of protégés either strongly agreed or somewhat agreed with Statements 6 and 7. A detailed description of the percent agreement based on responses using the five-point Likert scale is provided below for each statement. Additionally, Table 1 provides the percent agreement of the responses dichotomized into “agree”(1 or 2 on the Likert scale) or “do not agree” (3–5 on the Likert scale). A comparison of paired and unpaired responses appears at the end of the Results section.
Statement 1 (regarding the mentor's availability ). Of the 31 matched pairs in perfect agreement for this question (62% of 50 total matched pairs), both the mentor and the protégé of 30 pairs reported that they strongly agreed on the mentor's availability, and both members of the 31st pair somewhat agreed. Of the remaining 19 matched pairs (38%), the mentor's and the protégé's responses differed by at least one point (range 1–3) on the Likert scale.
Statement 2 (regarding the mentor's assistance in helping the protégé meet thesis deadlines). Of the 19 matched pairs in perfect agreement for this question (38% of 50 total matched pairs), both the mentor and the protégé of 14 pairs strongly agreed. Both the mentor and protégé of 2 matched pairs agreed somewhat, both the mentor and protégé of another 2 pairs neither agreed nor disagreed, and both members of the final matched pair disagreed somewhat. In 30 matched pairs (60%), the mentor's and protégé's scores differed by at least one point (range 1–3) on the Likert scale. In 1 matched pair, at least one respondent reported “not applicable.”
Statement 3 (regarding the mentor's guidance of the protégé to an appropriate thesis pathway). Of the 27 matched pairs in perfect agreement for this question (54% of 50 total matched pairs), both the mentor and the protégé of 23 pairs strongly agreed, and both members of 4 pairs agreed somewhat. In another 21 matched pairs (42%), the mentor's and protégé's scores differed by at least one point (range 1–3) on the Likert scale. In 2 matched pairs, at least one respondent reported “not applicable.”
Statement 4 (regarding the mentor's ongoing advice as the protégé worked on his or her thesis). Of the 27 matched pairs in perfect agreement for this question (54% of 50 total matched pairs), both the mentor and the protégé of 24 pairs strongly agreed, both members of 2 pairs agreed somewhat, and both the mentor and the protégé of 1 pair neither agreed nor disagreed with the statements. In another 22 matched pairs (44%), the mentor's and the protégé's responses differed by at least one point (range 1–3) on the Likert scale. In 1 matched pair, at least one respondent reported “not applicable.”
Statement 5 (regarding the mentor's actions to help the protégé succeed). Of the 27 matched pairs in perfect agreement for this statement (54% of 50 total matched pairs), both the mentor and the protégé of 26 pairs strongly agreed, and both members of the remaining pair agreed somewhat. In 21 pairs (42%), the mentor's and protégé's responses differed by at least one point (range 1–3) on the Likert scale. In 2 matched pairs, at least one respondent reported “not applicable.”
Statement 6 (regarding the mentor's assistance in helping the protégé negotiate for his or her current position). Of the 25 matched pairs in perfect agreement for this statement (50% of 50 total matched pairs), both the mentor and the protégé of 19 pairs strongly agreed, both members of 5 pairs agreed somewhat, and both the mentor and the protégé in the remaining pair strongly disagreed. In 12 matched pairs (24%), the mentor's and the protégé's responses differed by at least one point (range 1–4) on the Likert scale. In 13 matched pairs, at least one respondent reported “not applicable.”
Statement 7 (regarding the mentor's and protégé's continued contact with each other). Of the 17 matched pairs in perfect agreement for this statement (38% of 50 total matched pairs), both the mentor and the protégé of 16 pairs strongly agreed, and both members of the other pair strongly disagreed. In 23 matched pairs (46%), the mentor's and protégé's responses differed by at least one point (range 1–4) on the Likert scale; however, in 15 of these 23 pairs, the mentor strongly agreed and the protégé somewhat agreed. In 10 matched pairs, at least one of the respondents of the matched pair selected “not applicable.”
B. Unpaired responses.
Median responses to each of the questions among the 50 protégés who responded as part of a complete mentor–protégé pair did not differ from the median responses of the 22 protégés who responded but whose paired mentor did not (Table 1). However, the median response to each of the questions among the 50 mentors who responded as part of a complete mentor–protégé pair demonstrated significant differences for Statements 3–7 when compared with the 11 mentors who responded but whose paired protégé did not (P < .01), indicating slightly less agreement (median score 2 [unpaired mentor] versus 1 [paired mentor]).
C. Stratification by total number of publications.
We observed no substantial differences in percent agreement when we stratified the pairs by median total number of mentor publications or median total number of protégé publications (Table 3).
Overall, most mentors and protégés who responded to the survey expressed strong agreement with each other about the mentoring that the mentor provided within the CRTP. This overall agreement reflects the likelihood that mentors demonstrated basic mentor attributes as delineated by the program faculty. In fact, the overall high level of agreement with the statements made detecting the degree of agreement among the mentor–protégé pairs difficult (most responses were confined to the “strongly agree” and “agree” categories). Nonetheless, these initial findings are encouraging and align with the general literature on this topic that seeks to define attributes of successful mentors.14–18
Findings in the areas regarding long-term aspects of the mentor–protégé relationship (Statements 6 and 7) are noteworthy: Mentors and protégés both agreed in paired analyses that mentors helped protégés negotiate for their academic positions after program completion, and they agreed that the mentor served as a source of future support for the protégé. These aspects of mentoring go beyond the immediate goals of the CRTP program in that we originally conceived the program mentors to serve as “thesis mentors.”19 Again, such findings are encouraging in that these sentiments reflect the essence of “generativity,” that is, the protégé's success is a mentor's “ultimate award,” which is believed to be an important component of mentoring.20
Interestingly, the level of the experience of the mentors and the protégés, as measured by publication record, did not strongly correlate to the level of agreement with the statements about ideal mentoring. Although not an ideal measure, numbers of publications may be a proxy for academic rank of the mentors. Because the survey was anonymous, the mentors' academic ranks are unknown. This finding nonetheless suggests that the skills of a good mentor may not be exclusive to the highest levels of faculty achievement or research commitment—a theory some have discussed in the literature.14,15
Agreement with the seven statements was greater among paired mentors and protégés than among unpaired mentors (mentor answered the survey but protégé did not). The 22 protégés who responded but whose mentors did not also strongly agreed overall with the statements about their mentor's mentoring. However, unmatched mentors expressed slightly, but nonetheless significantly, less agreement with Statements 3–7 than did matched mentors, indicating that they felt they did less as mentors than their matched colleagues felt they did. It is tempting to speculate that the lack of response of their protégés reflected less engagement, implying that the quality of an unpaired mentor's relationship with his or her protégé was less satisfactory than when both partners responded as a pair. This implication makes the findings for Statements 4, 6, and 7, which did not focus exclusively on the CRTP program, especially interesting. These three statements focused on the mentors' work with their protégés beyond the confines of the program, that is, providing ongoing advice (Statement 4), helping to negotiate for faculty positions (Statement 6), and extending long-term support (Statement 7). Possibly, mentors who responded to the survey but whose protégés did not were more likely (than matched mentors) to view themselves as responsible solely for guiding their protégés through the CRTP program. This expectation may have led to the lack of response on the part of their protégés. On the other hand, protégés who did not respond but whose mentors did may have had similar expectations for their mentoring but were simply less engaged somehow. This lesser engagement may be geographic (i.e., graduates may have moved to other institutions) or research based (i.e., graduates' research interests may have moved in a different direction from their original thesis work).
This study has several limitations. Despite the survey's simplicity and an overall response rate of 74.4%, matched mentors and protégés both responded in about half of the cases, and thus these data are somewhat less robust than the overall data. The use of an anonymous survey limits our ability to conduct a more detailed multivariable analysis geared toward generating hypotheses about which characteristics of mentors might make them more effective. Although the program faculty assessed the survey's face validity before implementing it, they did not pilot the survey before deploying it. The high level of initial agreement indicates that mentors and protégés agree that mentors exhibited the attributes of a successful mentor, as conceived by its program faculty, within Einstein's CRTP. This level of agreement augurs well for the clinical research community at Einstein and helps to establish a baseline set of assumptions about adequate mentoring and how it is perceived by protégés. Ultimately, a more detailed description of mentors may well provide useful information to further facilitate the program goals by linking specific mentor attributes to specific outcomes.
In summary, we have found that within the confines of a CRTP, protégés and mentors perceived that the mentors provided effective mentoring, including making both short- and long-term commitments to their protégés. Hopefully, these findings will initiate a further dialogue to develop a “code of mentoring”16 that medical educators can define and articulate across institutions. An ongoing evaluation of the adherence to these principles of mentoring, the refinement of these principles, and a better articulation of the various aspects of effective mentoring may be beneficial to trainees and mentors alike. More quantitative study is needed to determine the characteristics of the most effective mentors.
The authors gratefully acknowledge the assistance of Ms. Mindy Ginsberg with the conduct of the online survey and Ms. Ann Marie McCauley for her secretarial and administrative support.
This work was supported in part by RR025748, 49, and 50 and CD41978 to Dr. Santoro.
The survey was considered exempt research by the Committee on Clinical Investigation of the Albert Einstein College of Medicine.