Griffith, Charles H. III MD, MSPH; Georgesen, John C. MA; Wilson, John F. PhD
Many researchers have tried to determine what influences students' specialty choices.1,2,3,4 Some have postulated that the primary influences are students' personal characteristics (e.g., controllable lifestyles, indebtedness); others have suggested medical school characteristics (e.g., orientation toward research). Medical educators, however, have focused on educational influences (e.g., curriculum, primary care experiences, and faculty role models), since these influences are more readily modifiable than are such factors as an institution's relative research intensity or a student's long-held values. We, for instance, previously looked at the influence of clinical role models and found that students who worked with the lowest-rated attending physicians during their third-year internal medicine clerkship were much less likely to choose a residency in internal medicine.5
The literature on students' specialty choices, our work included, has had one generally unacknowledged limitation: the range of specialty choices is not limitless for below-average and sometimes even average students. For example, in a report from our own institution regarding the relationship of medical students' achievement and specialty choices for graduates from 1984 to 1994, students from the bottom half of those classes rarely matched in the competitive specialties of that era—diagnostic radiology and the surgical subspecialties.6 Most of the lowest quartiles of the classes chose residencies in family practice, internal medicine, and psychiatry. To be frank, many highly competitive specialties are realistic options only for students with higher academic achievement. Several studies have suggested this in their data, although they have generally not focused on this phenomenon in their text. For example, low MCAT scores have been important predictors of primary care residency choice in regression analyses of large data sets.7,8
Researchers and medical educators who would like to see the brightest and best medical students entering broad-based generalist specialties must investigate influences on the specialty choices of those specific students. Therefore, we sought to determine the relationship of the “quality” of clinical role models in internal medicine (both attending physicians and residents) to the residency choices of the best medical students—those who truly had choices. Our formal hypothesis was that excellent medical students who worked with the highest-rated internal medicine role models during their third-year medicine clerkships would be more likely than excellent medical students who did not have such exposure to choose residencies in internal medicine.
In a prospective cohort study at the University of Kentucky, we collected data for all 169 students who rotated on the third-year internal medicine clerk-ship in academic years 1993–94 and 1994–95 and for their 62 supervising attending physicians and 89 supervising residents. In 1997, we published an article5 that looked at the relationship between the quality of the attending physicians and their students' residency choices and demonstrated that students were less likely to choose primary care residencies if they had worked with low-rated attending physicians during their medicine clerkships. The current article represents a data analysis that both focuses and extends the findings in that previous report. We now focus only on “better” students and extend the scope to include the association between the quality of residents and students' specialty choices. By “better” students, we mean students who scored in the top 30% of their two-year cohorts on the USMLE 1. Our cohort for this study, then, was the 52 students who scored 215 or higher on the USMLE 1. We chose USMLE scores because, unlike class rank or membership in Alpha Omega Alpha, they reflect a nationally recognized standard of academic excellence.
At the University of Kentucky, the inpatient internal medicine clerkship consists of two four-week rotations on the general medicine inpatient services, one at the university hospital and the other at the affiliated Veterans Affairs hospital. (Ambulatory general medicine experiences are in a temporally separate primary care clerkship and were not included in this study.) During the clerk-ship, students are randomly assigned in pairs to one of nine general medicine teams (we do not honor students' requests to be assigned to specific teams or attending physicians; attending physicians and residents are assigned in a process independent from the student assignments). Each team consists of an attending physician, a supervising junior or senior resident, two interns, and two students. A substantial number of the interns are “rotating” on the medicine services from other disciplines (i.e., neurology, anesthesiology, obstetrics and gynecology); because of the heterogeneity of their background disciplines, we did not hypothesize that they would influence internal medicine residency choice.
The resident generally leads the daily management rounds in which the team discusses and visits every patient on the service, with interns and students verbally presenting the patient at the bed-side or outside the patient's room to the resident. The residents are encouraged to teach the students and interns on the management rounds, usually focusing on the patient being discussed. The residents also work closely with the students after the management rounds, especially during new patient evaluations and when completing daily patient care duties. The attending physicians some-times participate in or observe the daily management rounds, but are encouraged to allow the residents to be the focus of presentation and decision making. The attending physicians are expected to conduct separate formal teaching rounds with the team three times a week, one to two hours per session, focusing on an individual patient or topic.
At the end of each month's rotation, before the students get their clinical grades, they confidentially evaluate their residents and attending physicians. The evaluation forms, described in detail elsewhere,5,9 were created at our institution. Briefly, the form that evaluates attending physicians consists of 16 items (scored on a five-point Likert-type scale: 1 = strongly disagree to 5 = strongly agree) concerning such things as the students' assessments of their attending physician's teaching ability and skills, rapport with students, residents, and patients, and overall rating as a teacher and a role model. The form that measures the quality of residents contains nine items (scored on the same Likert-type scale) regarding resident teaching, management of service and involvement of students, rapport with patients, students, interns, and ancillary personnel, and role modeling. Though the forms are reliable and internally consistent (coefficient alphas of .96 for the attending physician form and .95 for the resident form), the individual items have high inter-item correlations (which is not surprising for teaching evaluation forms). Because of this lack of item discrimination, we used the mean rating across the items on each form to indicate a role model's “quality.” Each attending physician and resident received an overall score, which was the mean of the ratings from all the third-year students that instructor had precepted during the two-year period.
Conceptually, our model for any influence that attending physicians or residents might have is not necessarily the average ability of all the clinical role models a student works with, but rather, that significant influence occurs from exposure to a particularly outstanding (or particularly poor) role model. For this study, we a priori defined “best” attending physicians and residents as those receiving the highest 20% of overall teaching evaluations in the two-year period, and “worst” as those receiving the lowest 20%. We chose these cutoffs because we believed most attending physicians and residents (over 50%) are neither best nor worst, and that the top 20% is not such a rarefied cohort as, say, the top 5% or 10% (and therefore a cohort perhaps more attainable through faculty development).
Our outcome of interest was internal medicine residency choice, as determined by match-day results. Analysis was with multiple regression approaches derived from the general linear model.10 The dependent variable was specialty choice (dichotomized to an internal medicine residency versus all other specialties). Independent variables included “dummy” coded variables for the different categories of housestaff (exposure to “best” resident or no; “worst” resident or no) and attending physicians (exposure to “best” attending or no; to “worst” attending or no).
We also wanted to investigate the possibility of rater bias, specifically, that an excellent student already oriented to internal medicine might rate all his or her internal medicine attending physicians and residents high, resulting in a spurious association of internal medicine residency choice and high instructor ratings. To investigate this possibility we performed Pearson correlation analysis of excellent students' mean evaluations of each of their two attending physicians and two residents. A lack of significant correlation among a student's evaluations of internal medicine faculty and housestaff would indicate that the students did discriminate in their ratings of their instructors and did not rate them uniformly high or low due to bias.
The original cohort of 169 third-year students completed 291 of 338 possible evaluations (86%) regarding 62 faculty (4.7 evaluations per faculty member). Those students also completed 320 of 338 possible evaluations (96%) for the 89 residents (3.6 evaluations per resident). After calculating the 20% at the top and bottom of each group, we found that the 12 “best” faculty had precepted 62 students and had received a mean rating of 4.80 (SD, 0.31) on the five-point scale. The 12 “worst” faculty had precepted 56 students and received a mean rating of 3.62 (SD, 0.71). The 18 “best” residents had worked with 69 students and received a mean evaluation of 4.80 (SD, 0.09). The 18 “worst” residents had worked with 56 students and received a mean rating of 3.49 (SD, 0.32).
In the multiple regression analysis including all the various categories of exposure or no exposure to best or worst instructors, independent predictors of internal medicine residency choice were exposure to highly rated attendings (F = 5.68, p = .02) and exposure to highly rated residents (F = 5.28, p = .03), suggesting the influences of attendings and residents were independent. Exposure to low-rated faculty or residents was not associated with internal medicine residency choice.
Figure 1 demonstrates the percentages of students choosing an internal medicine residency depending on their exposures to “best” versus “worst” internal medicine instructors. For the 52 “excellent” medical students, 19 worked with at least one highly rated resident or attending, but no low-rated instructor; 14 worked with at least one low-rated resident or attending, but no high-rated instructors, nine students worked with neither a high-rated nor a low-rated instructor; and ten worked with both a high-rated instructor and a low-rated instructor. As presented, six of the 19 (32%) excellent students who worked with one high-rated instructor and no low-rated instructor chose a medicine residency. Of the 23 students who did not work with a highly rated instructor (14 worked with a low-rated instructor, and nine worked with neither high nor low), none chose a residency in internal medicine. Of the ten students who worked with both a high-rated instructor and a low-rated instructor, three (30%) chose a medicine residency. There was no difference in USMLE 1 scores of excellent students who were exposed to high-rated instructors versus those who were not so exposed (225 ± 8 versus 223 ± 6; p = .90), suggesting that students of similar academic achievement were in the two cohorts. Figure 1 also compares specialty choices depending on exposures to high-rated and low-rated instructors for students not defined as excellent (i.e., not in the top 30% of USMLE I scores, or scores less than 215). As can be seen, for these students, who in general have a narrower spectrum of residency choices, the choice of a medicine residency occurred more frequently than for excellent students, but was independent of high- or low-rated instructor exposure.
Our investigation did not detect any rater bias. The median correlation of the four ratings by excellent medical students (two attendings, two residents per student) (r = 0.12, p = .40) suggests that the students discriminated in their ratings of instructors and did not rate them uniformly high or low despite their relative proclivity to select an internal medicine residency. In comparison, the median correlation of the ratings of the four instructors of the “not excellent” students was similar (r = 0.07, p = .44), suggesting that excellent students discriminated in their evaluation of instructors similarly to other students.
Our results confirmed that exposure to excellent internal medicine clinical role models, both faculty and residents, is associated with the greater likelihood that an “excellent” medical student will choose a residency in internal medicine. In the random nature of the assignment of students to teams of residents and attendings, if an excellent student by chance had the good fortune to work with just one outstanding role model (out of the four possible instructors), that student had approximately a 30% chance of choosing an internal medicine residency. Dramatically, in the two-year study period, not a single excellent medical student who was not fortunate enough to have such an exposure chose an internal medicine residency (23 students, and not a one choosing medicine, versus 30% of their colleagues).
Several studies have cited clinical role models as being important influences on students' residency choices,3,8,11,12,13,14,15 (including negative role models, who can drive students away from some specialties5,13). In addition to faculty, resident role models have been occasionally cited as influential.11,12 However, for the most part, those studies have been retrospective, so it is difficult to determine whether their data arose from actual influence or from students' fond remembrances. The one prospective study noted that students exposed to a generalist attending physician role model in their internal medicine clerkship were more likely to choose general medicine, but the study measured this with a post-clerkship survey and did not determine actual residency choices.15
Our prior report from this data set was the first prospective study to document an association of the relative ratings of clinical role models with their students' specialty choices.5 That report focused on all students and noted primarily the association between exposure to a low-rated faculty member and a student's being less likely to choose a medicine or primary care residency. In contrast, the current study focused on the “excellent” medical students, who generally are the ones with the most realistically broad array of specialty choices. To our knowledge, it is the first study to do so. Further, this report is the first to document prospectively the influence of excellent residents on students' specialty choices. We found that, for excellent students, excellent teaching does influence specialty choice, and that this influence is tangible and not merely fond remembrance.
Our findings have several implications. First, exposure to excellent clinical role models appears to be an important predictor of specialty choice for excellent medical students, but not so for less-than-excellent students. But, as we have suggested, a broad array of specialty choices is generally not available to less-than-excellent students, and this restriction of opportunities may have masked the relation between instructor quality and specialty choice. Perhaps the difficulty in documenting modifiable influences in specialty choice has been that researchers are studying the wrong population, including too many students whose choices are limited. This problem may also explain the heterogeneity of influences and the sometimes contradicting nature of specialty-choice determinants in the literature. Therefore, researchers interested in specialty choice should consider emulating our method, focusing on the influences on the specialty choices of only the better students, those who truly have choices. Second, our study demonstrates that excellent clinical role models can have measurable effects on student outcomes. As academic medical centers increasingly are asked to justify and account for the costs of medical education, studies that demonstrate the influence of faculty on student outcomes, such as this study, help provide such a justification for the added costs of a high-quality teaching program or of faculty development.
Several limitations to our study should be taken into account when interpreting our results. First, this study is from a single institution, in a single clinical clerkship, in but two academic years. Future studies should consider the influences on the specialty choice of excellent students from many institutions, across several clerkships, over many years. Second, our measure of the “quality” of attending physicians and residents was based solely on students' evaluations. A more comprehensive measure of instructor quality might include, for example, attending physicians' evaluations of residents (and vice versa), as well as peer evaluations. However, one could argue that there are none better to judge the impressions that faculty and residents make on students than the students themselves. Third, a possible confounder of concern was rater bias: internal medicine-oriented excellent students might rate their instructors higher, creating a spurious association of instructor rating and internal medicine specialty choice. However, our investigation did not detect such a bias. Further, most of the excellent students did not choose residencies in internal medicine. For faculty members or residents to achieve high ratings (top 20%), they had to be rated highly by generally all of their students whether the students chose an internal medicine residency or not. For these reasons, we believe rater bias does not explain our results.
Despite these limitations, we conclude that clinical role models, both attending physicians and residents, can have measurable impacts on the specialty choices of excellent medical students—those students who truly have choices in the matter.
1. Meurer LN, Bland CJ, Maldonado G. The state of the literature on primary care specialty choice: where do we go from here? Acad Med. 1996;71:68–77.
2. Meurer LN. Influence of medical school curriculum on primary care specialty choice: analyses and synthesis of the literature. Acad Med. 1995;70:388–97.
3. Campos-Outcalt D, Senf J. Watkins AJ, Bastacky S. The effects of medical school curricula, faculty role models, and biomedical research support on choice of generalist physician careers: a review and quality assessment of the literature. Acad Med. 1995;70:611–9.
4. Pathman DE. Medical education and physicians' career choices: are we taking credit beyond our due? Acad Med. 1996;71:963–8.
5. Griffith CH III, Wilson JF, Haist SA, Ramsbottom-Lucier M. Relationships of how well attending physicians teach to their students performance and residency choices. Acad Med. 1997;72(10 suppl):S118–S120.
6. Rubeck RF, Witzke DB, Jarecky RK, Nelson B. The relationship between medical students' academic achievement and patterns of initial postgraduate placement. Acad Med. 1998;73:794–6.
7. Kassebaum DG, Szenas PL, Schuchert MK. Determinants of the generalist career intentions of 1995 graduating medical students. Acad Med. 1996;71:197–209.
8. Martini CJM, Veloski J, Barzansky B, Xu G, Fields SK. Medical school and student characteristics that influence choosing a generalist career. JAMA. 1994;272:661–8.
9. Griffith CH III, Wilson JF, Haist SA, Ramsbottom-Lucier M. Do students who work with better housestaff in their medicine clerkship learn more? Acad Med. 1998;73:557–9.
10. Cohen J, Cohen P. Applied Multiple Regression Correlation Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates, 1983.
11. Burack JH, Irby DN, Carline JD, Ambrozy DM, Ellsbury KE, Stritter FT. A study of medical students specialty-choice pathways: trying on possible selves. Acad Med. 1997;72:534–41.
12. Lieu TA, Schroeder SA, Altman DF. Specialty choices at one medical school: recent trends and analysis of predictive factors. Acad Med. 1989;64:622–9.
13. Mutha S, Takayama JI, O'Neil EH. Insights into medical students' career choices based on third- and fourth-year students' focus-group discussions. Acad Med. 1997;72:635–40.
14. Ambrozy DM, Irby DM, Bowen JL, Burack JH, Carline JD, Stritter FT. Role models' perceptions of themselves and their influences on students' specialty choices. Acad Med. 1997;72:1119–21.
15. Henderson MC, Hunt DK, Williams JW. General internists influence students to choose primary care careers: the power of role modeling. Am J Med. 1996;101:648–53.