The National Institutes of Health has long recognized the importance of cultivating physicians who perform research, for instance, through funding the Medical Scientist Training Program (MSTP) through the National Institute of General Medical Sciences (NIGMS) in 1964.1 Today, at least 118 of 126 U.S. medical schools offer combined MD–PhD programs,2 reflecting the importance the medical community has placed on training those who provide a vital link between scientific advances and clinical applications.3–8
The success of MD–PhD program graduates has been well documented by external and objective means.1,9–16 Such graduates have been adept at navigating the mechanics of academia (grants, promotions, publishing), have served as role models and leaders, and have contributed to the health care missions of our society.1,9–15
Although there is ample information regarding MD–PhDs once they graduate, little is known regarding the educational process and the views of students as they go through training.17,18 A careful evaluation of the educational process should provide useful feedback, a useful basis for proactive improvements. In addition, characterizing the views and plans of students will have other potential benefits such as helping students to better understand their community, policy makers to better allocate resources, and residency directors to better recruit future colleagues.
Previously, our group has reported on the results of a single MD–PhD program student survey.17,18 In this study, we report the results from a similar survey representing views and opinions from 492 students at 13 MD–PhD programs in the United States (one eighth of all MD-PhD trainees in the United States [O. S. Andersen, personal communication, 2003]). In particular, we address a number of issues anecdotally important to students and educators alike: (1) do students at smaller programs have a different experience than those at larger ones, (2) does the student experience change with progression in training, (3) how do students’ clinical interests change over time, (4) what types of future careers do students desire, and (5) how do they feel about the physician–scientist model? To our knowledge, this study represents the first characterization of MD–PhD students at the national level.
In spring 2003, we contacted all 868 MD–PhD students who were enrolled at that time in 15 participating combined-degree programs. Those programs were the Weill Cornell/Rockefeller/Sloan-Kettering Tri-Institutional MD–PhD Program; Emory University School of Medicine MD/PhD Program; the MD–PhD Program at Harvard Medical School; University of California, Irvine, School of Medicine Medical Scientist Training Program; University of California, Los Angeles, Medical Scientist Training Program; University of California, San Francisco, School of Medicine Medical Scientist Training Program; University of Chicago Pritzker School of Medicine MD–PhD Program; University of Connecticut School of Medicine MD–PhD Program; the MD–PhD Program at the University of Florida; University of Illinois, Urbana-Champaign MD/PhD Medical Scholars Program; University of Kentucky College of Medicine MD/PhD Program; University of Rochester Medical Center Medical Scientist Training Program; University of Texas Southwestern Medical Center at Dallas Medical Scientist Training Program; University of Vermont College of Medicine MD/PhD Program; and the Medical Scientist Training Program at Washington University. Institutions were chosen to represent a variety of program sizes (8–161 students; average 67), funding structures (9 of 13 with MSTP funding19), geographic locations (four in the East, and three each in the Midwest, South, and West), and local environments. We asked students to complete an online questionnaire, termed the MD–PhD SAGE (Students’ Attitudes, Goals, and Education). The questionnaire was made up of 29 closed-ended questions and collected information regarding student demographics, year of matriculation, funding status, debt burden, field of PhD thesis, stage in program, primary clinical interest, satisfaction with various aspects of the MD–PhD educational experience, future residency and career goals, and attitudes and views concerning the physician–scientist model. See Appendix 1 for the full survey.
After approval by the University of Pennsylvania School of Medicine IRB, all eligible students were sent recruitment invitations via three group e-mails, each of which included a link to the survey. The final reminder was an individualized e-mail. Five raffle prizes, valued at $100 each, were offered as incentives. E-mail addresses were provided by local program leaders and were used only to document participation in the study. Individual answers were stored on a secure server without linkage to the participant’s e-mail address.
Stratification of participants.
The data were stratified for analyses using three survey items.
- Program size was categorized as small (six programs with 8–32 students), medium (four programs with 60–93 students), or large (three programs with 138–161 students). These divisions represent the most obvious clusters based on size, maintaining enough individuals in each group for reasonable comparisons.
- Students’ stage in program was categorized as pre-grad school, grad school (prethesis), grad school (thesis), clinics (postthesis), or completed all requirements; this categorization is based on a general paradigm used by MD–PhD programs.
- Clinical interest was defined with three categories: surgery or surgical: primarily surgical or procedure-based fields (primary clinical interest in general surgery, otolaryngology, neurosurgery, plastic surgery, orthopedic surgery, ophthalmology, obstetrics–gynecology, urology); medical: primarily based on medical management (internal medicine, pediatrics, neurology, physiatry, emergency medicine, family medicine, psychiatry); and lifestyle and diagnostic: primarily diagnostic or anecdotally considered especially conducive to a favorable lifestyle (dermatology, anesthesiology, pathology, radiology, radiation oncology). In one analysis, the latter two categories were combined and labeled nonsurgery.
Statistical analysis and nonresponders.
Statistical analysis was performed using Stata SE 8.2 (StataCorp LP, College Station, TX) and JMP 5.1 (SAS Institute Inc, Cary, NC). Comparisons were made with ANOVA and Student t test for normal and continuous variables, rank-sum tests (Mann-Whitney U or Kruskal-Wallis) for nonnormal/continuous variables, and contingency table analysis for proportional data (chi-square). Satisfaction and agreement scores are reported as averages. Standard deviations (SD), 95% confidence intervals (CI), and P values are reported as indicated (significance set at P < .05).
Finally, we randomly selected 50 nonresponders via multiple e-mail solicitations to mitigate concern about nonresponder bias. Abbreviated questionnaires were sent to these nonresponders that collected data about demographic characteristics, year of matriculation, funding status, debt burden, field of PhD thesis, stage in program, primary clinical interest, and satisfaction with program; similar statistical analyses were performed on these data. To assess the likelihood of a “clinically” important bias (a priori determined to be a difference of 1.5 on the eight-point overall satisfaction scale; >80% power with alpha of 0.05 and sample size of 50), the nonresponders’ satisfaction ratings were compared with those of the responders.
Results and Discussion
Demographics and background
Surveys from 492 students at 13 MD–PhD programs were returned (57.6% response rate); two programs were excluded from the analysis pool because of nonadherence to the recruitment protocol. Of the 492 respondents, 20 students did not fit the paradigm and were excluded from the stage-based analyses. The average age of students was 26.6 years (SD 3.1); 159 (32.3%) respondents were women, and 56 (11.4%) considered themselves part of a racial minority group underrepresented in medicine (Table 1). One hundred forty-six (29.7%) had one year or more of full-time research experience before matriculation in their MD–PhD program (Table 1), suggesting significant commitment to research before enrolling. Students at all stages of training were reasonably represented: 119 (24%) from pre-grad school; 246 (50%) from grad school (prethesis and thesis); 107 (23%) from post-grad school (clinics and completed all requirements). Generally, students expected to complete their programs in approximately eight years (SD 0.9). Three hundred ninety-three (79.9%) students were receiving full funding for their studies; accordingly, only a small proportion (48; 9.8%) expected their educational debts to be greater than $50,000 on graduation (Table 1).
Our response rate was consistent with rates reported for surveying similar populations.20 However, the sizable nonresponse rate raised concerns for nonresponder bias. On a scale of 1 (most satisfied) to 8 (most dissatisfied), the average overall satisfaction among responders (2.7 ± 1.8 SD) and nonresponders (3.4 ± 2.2 SD) was similar according to the criterion of <1.5 that we set, suggesting no important responder-based bias. The two groups did not show statistically significant differences in gender and minority proportion, year of matriculation, funding status, debt burden, field of PhD thesis, stage in program, or primary clinical interest. Those who did not respond were, on average, 0.9 years older (P < .01).
Demographic and background characteristics revealed little difference between student populations at small, medium, and large programs (Table 1). Higher proportions of students from the small group reported more than one year of full-time research before matriculation (P = .03), and those from large programs reported the lowest proportion with full funding (P < .001). The latter is likely to reflect partial institutional support for students beyond the limits of the more comprehensive MSTP funding. Although conclusive statements cannot be made, students from small, medium, and large programs seem to have more demographic similarities than differences.
Satisfaction with education
Students seem to have high overall satisfaction with their MD–PhD programs. On a scale of 1 (most satisfied) to 8 (most dissatisfied), the respondents had an average score of 2.65 (SD 1.76, median and mode of 2). The overall satisfaction score varied from stage to stage in the program, as shown in Figure 1. The highest satisfaction was during the first stage of training (pre-grad school, 2.19), was lowest during the thesis stage (2.94, P < .01 compared with pre-grad school), and then rose slightly by program completion (completed all requirements, 2.67, P < .05 compared with pre-grad school, P > .05 compared with thesis stage). There was no statistically significant change between any two adjacent stages.
Two trends deserve comment. One, it is difficult to ascribe importance to the statistically significant but small downward trend in satisfaction from the first to the fifth/last stage of training. The investigators arbitrarily but a priori determined an average change of 1.5 to be of “clinical” importance. Even if we were to assume that our threshold was incorrect, a change of 0.5 is less than one third of a standard deviation and, thus, represents a small effect size.
Two, on the other hand, the pattern of change warrants consideration. Specifically, the lower satisfaction during the grad school (thesis) stage with a “rebound” in the clinic stage closely reflects anecdotal evidence. Our thought is that the predictable progress of early stages is replaced by more unpredictable thesis work. Compounding this is the graduation and departing of former classmates (fellow MD-only matriculants). Program directors and advisors should not only take note of this observation; they should direct resources and time to support students during the thesis period.
To identify specific aspects of MD–PhD training that may be in need of improvement, we also asked students to rate various components of their programs. Students seem to have relatively high satisfaction with their clinical and research education but relatively lower satisfaction with more ancillary information concerning their futures after MD–PhD graduation, obtaining grants/funding, and coordination between the MD and PhD components of their programs (Figure 2, components listed in order of satisfaction, highest to lowest among all students). Some variability was noted on the basis of program size, with 4 of the 11 comparisons being statistically significant and all differences favoring large programs (see note, Figure 2).
Again, two observations are worth mentioning. One, students seem to be more satisfied with the basics of their education, clinical and research. They seem less satisfied with the more ancillary aspects of being part of an integrated and career-oriented MD–PhD program: coordination between medical and graduate school and information concerning future residency and career choices. Programs should take note and attempt to improve these aspects of student training. Two, all four of the components found to be statistically different on the basis of program size favored the large programs and concerned these ancillary aspects of MD–PhD training. Smaller programs should take note and continue to improve these aspects of their education.
Negative thoughts about training
We explored two prominent “negatives” commonly expressed by students in the form of I’m thinking of quitting or I can’t be here any longer; I have to get out of here in eight years. Overall, 120 (24.4%) respondents responded yes to the question, “Have you ever seriously considered leaving your MD–PhD program?” A breakdown by stage revealed a significant difference between the pre-grad school (11/119, 9%) stage and the prethesis (13/57, 23%, P < .05) and thesis (57/189, 30%, P < .001) stages. These rates are likely to underestimate true differences because they do not include students who have already left. Although we can speculate that students are at greatest risk of leaving training during the transition from the pre-grad school to the grad school period, we cannot know until we have an accurate measure of dematriculation (data not released by individual programs).
We also asked students whether “pressure to complete your MD–PhD program in a defined period ever adversely affected your educational experience.” Overall, 62 (12.6%) students answered in the affirmative. Evaluated by stage, there were significant differences between the pre-grad school (5/119, 4%) and prethesis (8/57, 14%, P < .05) and completed-all-requirements (11/52, 21%, P < .001) stages. Concern among students regarding length of training time may not be appropriate or controllable, because ample time must be given to training. However, efforts should be made to identify and minimize external, unnecessary sources of time pressure. Also, educational redundancies and transition inefficiencies between the MD and PhD phases must be minimized; this will require mechanistic and political cooperation between parties involved. Neither “negative” showed significant differences related to the size of the program (P > .05, data not shown).
Attainment of philosophical goals
Part of the rationale for integrating MD and PhD training is to find synergy between the two sometimes disparate educational processes. This synergy was expressed by responses to the statements, (1) “your ability to recognize clinical problems that can be approached scientifically” and (2) “your ability to design experiments/approaches to solve clinical problems.” Students were asked to rate their ability on these fronts with a satisfaction scale of 1 (most satisfied) to 8 (most dissatisfied). Overall satisfaction with ability to recognize clinical problems was 2.5 (SD 1.5), and satisfaction with ability to design approaches to solve problems was 2.8 (SD 1.6). Program size was not related to students’ self-ratings (P > .05, data not shown), and there were no significant differences between any two adjacent training stages (P > .05). However, there were statistically significant differences between the early training (pre-grad school) and late training (completed all requirements) groups with regards to both recognizing problems and designing solutions (respectively, 2.7 versus 2.0, P < .05; and 3.0 versus 2.0, P < .05) (Figure 3). The noticeable skew may reflect students’ generally positive self-assessments. Nonetheless, the data do suggest a gradual increase in satisfaction with advancement in training, regarding the two philosophical goals of an MD–PhD education.
Looking beyond graduation, we asked students to identify their primary clinical interest among 20 choices (available through the National Residency Matching Program21), other residency options (nuclear medicine and combined medicine–pediatrics were among those not listed), and no residency. The vast majority—462 students—chose from among the 20 choices, with 24 (4.9%) choosing other and 6 (1.2%) choosing no residency. The most common choices were consistent with previously reported assessments of MD–PhD graduates1,10–15 and included internal medicine (135, 27%), neurology (65, 13%), pediatrics (52, 11%), and pathology (31, 6%). In contrast to surveying former graduates retrospectively, we had the unique opportunity to examine students while their preferences were being formed. At the beginning of training, medical, lifestyle, and surgical clinical interests represented 70%, 13%, and 8%, respectively, of all 492 respondents. A smaller proportion of students at more advanced stages of training (48% versus 70%, P < .01) claimed an interest in a medical residency (Figure 4). As a result, at more advanced stages of training, there was a larger proportion of students choosing surgical and lifestyle residencies (respectively, 8% to a final 23%, P < .01; and 13% to a final 27%, P < .05; Figure 4).
These differences represented an almost 30% decrease in medical residency interest, a more than twofold increase in lifestyle interest, and a nearly threefold increase in surgical interest between the first and last stages of MD–PhD training. One simple explanation is that MD–PhD students enter programs with the greatest interest in the few specialties that they have had the most experience with (clinical, research, and exposure to MD-PhD role models)—most likely medical specialties. With continued experience, they find their eventual interests. The implication is that teaching institution, departments, and individual instructors/mentors can play a role by proactively providing exposure to their field and to MD–PhD role models.
The question then arises as to when this exposure is best given. There was a significant jump in the degree of interest in their chosen primary clinical field (where 1 = greatest and 8 = least) between the postthesis (1.8, CI 1.35–2.25) and completed-all-requirements (1.2, CI 1.08–1.32) stages of training (P < .001). Although there is a statistically significant difference between the completed stage and every other stage (P < .001), there are no such differences between any two other stages (P > .05; pre-grad school 1.8, CI 1.6–2.0; prethesis 1.9, CI 1.58–2.31; thesis 1.9, CI 1.71–2.11). Though we cannot say for certain that a confidence score of 1.8 (pre-grad school) is moldable whereas 1.2 (completed) is not, it seems reasonable to assume that a student applying for residencies through the Match will be much more difficult to redirect than a first-year student would be. As mentioned above, interest scores between the pre-grad school to postthesis stages did not differ significantly. Taken together with the general trend (Figure 4), one interpretation of the data is that MD–PhD students are searching for their clinical calling for the majority of their training, solidifying their interests only at the very end. Although our data cannot pinpoint the optimal timing of “exposure” to various clinical specialties, it does suggest that it is almost never too late to educate students regarding residency choices.
Future career goals
The future career desires of MD–PhDs in training were consistent with those reported in the literature concerning program graduates.1,10–15 Just over half of the respondents, 277 (56%) students, chose research as their primary professional activity. Approximately one third (161, 33%) of the respondents chose patient care, and another 33 (7%) named teaching as their primary activities. However, by current definitions, 44%, or almost one half of respondents, indicated that they do not wish to become physician–scientists because they do not wish to perform research as their primary professional activity. This suggests either that programs are failing in their mission to train physician–scientists through MD–PhD programs, or that the definition of the physician–scientist (and with it the implication that research activities should necessarily trump clinical ones) is incorrect.
Consistent with their extensive academic training, 444 (90%) respondents would like to obtain a position at an academic center. Their other choices were none/other (14, 2.8%), research institute (13, 2.6%), government (11, 2.2%), hospital/clinic (8, 1.6%), and industry (2, 0.4%). Ninety-nine percent of respondents expressed a desire to perform some type of research: 257 (53%) chose disease oriented, 133 (27%) chose basic science, and 76 (15%) indicated clinical or epidemiologic to describe their future primary research type. Taken together, the “typical” MD–PhD graduate seems headed for an academic center and will conduct laboratory research as his or her primary professional activity. However, we should be mindful of the fact that 33% of respondents prefer patient care as a primary activity and that 15% prefer clinical/epidemiologic research to more traditional laboratory research.
Relevance of the physician–scientist model
When asked whether they agreed with the definition of the physician–scientist (an MD who performs research as his or her primary professional activity), 292 (59%) respondents agreed. In contrast, 41% did not: 129 (26%) disagreed, and 71 (15%) were undecided. Surgery students were more likely to disagree (25, 36%) and less likely to agree (34, 49%) to the definition than were nonsurgical students (94 [24%] disagree and 244 [62%] agree; P < .05) (Table 2). These findings suggest that the definition may be too narrow to encompass the varied/undetermined career aspirations of MD–PhD students—in particular, those interested in surgical careers. Alternatively, some students may not have realized that most physician–scientists are individuals with only an MD degree, so the disagreement may have been a function of misperception.
When asked whether being successful as a researcher or as a clinician required more than 50% of one’s time, 366 (74.4%) respondents agreed with the statement about researchers, but only 152 (30.9%) agreed with the statement regarding clinicians (P = .001). We hypothesized that the surgery student group would consider clinical time (time performing surgery, etc.) more important than would nonsurgery students; however, both groups ascribed similar importance to clinical time (Table 2). In fact, both groups thought that research would require a greater time commitment than clinical time (54% for research and 33% for clinical among the surgery group, P < .01; and 79% for research and 32% for clinical among nonsurgery, P < .001). A significantly smaller proportion of surgery students felt that successful research would demand majority time (54% versus 79% for nonsurgery, P < .001, Table 2).
Students reported their degree of agreement with the statement, “An MD–PhD graduate should become a physician–scientist (assuming the traditional definition)” (scale of 1–8, where 1 = strongly agree and 8 = strongly disagree). The squarely middle-of-the-road response by respondents (4.2, SD 2.0; median and mode of 4) suggested ambivalence on this issue.
There does not seem to be any practical reason that a physician–scientist must perform research as a primary activity, and it should not be preached as such. David Ricardo (considered by many to be the father of modern economics) would argue that specialized activities are best left to the specialist for maximum benefit/efficiency.22 The MD clinician is best at patient care and should perform that as a primary activity. Similarly, the PhD researcher is best at research and should perform that as a primary activity. A physician–scientist is best equipped to provide connections between clinical medicine and research and should perform that function as best able—be it primarily in the arena of research, patient care, administration, or health policy. Similarly, MD–PhDs are perfectly poised to be leaders in bringing together the benchside and bedside; in that respect, they are physician–scientists regardless of whether research is their primary activity.
Methodologically, we were unable to obtain (from the NIGMS or the MD–PhD Directors Association) a pool of students from which to draw a random sample. Therefore, we opted to select by program characteristics such as size and location, to have diverse representation. In addition, we obtained responses from approximately one eighth of all MD–PhD students in the United States (O. S. Andersen, personal communication, 2003), a large enough sample to reduce the likelihood of selection bias.
Our response rate was low but typical of our survey population.20 Sampling analysis of the nonrespondents indicated that important nonresponder bias was unlikely in our dataset. Still, we cannot completely rule out the influence of nonresponder bias on our findings. Also, our response rate represents a conservative value, because we were unable to verify or document an actual refusal to participate. The potential subjects may not have received our e-mails for various reasons or may not have seen them in their inbox.
By design, our survey was a single-time-point cross-sectional study. Although this method does not allow one to measure change and outcomes over time, it does allow inclusion of a greater number of subjects (for the same amount of resources) and abrogates the effects of the passage of time (such as changes in curriculum, program funding or size, or losing subjects to follow-up).
Statistically, we did not adjust the significance level to compensate for the performance of multiple tests based on the same data stratifications. We did, however, provide levels of P values (not significant at P > .05, and significant at P < .05, P < .01, and P < .001), thus allowing readers to make their own conclusions.
Finally, although survey the instrument has been previously used17,18 and developed according to recommended procedures for item development and piloting,23 the questions focused on opinions and self-report of intentions and not on objective measurements. It was not designed to measure success or to predict the future actions of graduates but, rather, to gather the thoughts of MD–PhDs in training in the hopes that the data could be used to improve on their education and future careers.
Conclusions, implications, and future directions
It should be emphasized our findings are a reflection of MD–PhD students’ needs and interests. That investigation should be based on clinical observation is a lesson learned from Goldstein and Brown,6 physician–scientists who continue to teach us the importance of instigating investigation based on clinical observations. We designed and implemented this survey to answer basic questions that students had about fellow students in MD–PhD programs, their interests and motivations, debates concerning the definition of a physician–scientist, whether surgical specialties should be considered, the frequency with which students considered leaving their programs, why students became less social for at least a year during PhD thesis work, and whether students in large programs had experiences different from those of students in medium or small programs.
There are a range of basic implications of our study. The majority of students are satisfied with their education and intellectual progress, but the graduate school years may be the most difficult for students and require extra vigilance on the part of leaders and mentors. Although there are some sized-based variations, students seem to undergo equivalent educational experiences at larger and smaller programs. There is a significant difference in students’ clinical interests between the first and last stages of training, with a gradual movement away from material that is initially familiar. Accordingly, students may not make their final clinical choice until very late in training. Finally, although most students wish to perform research or provide clinical care within an academic setting, they are diverse in their opinions, and many are ambivalent about what it should mean to be a physician–scientist.
Certainly, we should continue to monitor and improve the education of MD–PhD students and others pursuing combined degrees, but we should do so with two modifications: (1) longitudinal studies should be carried out beginning in medical/graduate school, with continuation into the careers of graduates, and (2) the effort to collect data from students should be more concerted and widespread, utilizing the resources of the American Association of Medical Colleges and the NIGMS. Our society has made a significant investment into educating and training MD–PhDs; we have a responsibility to ensure that it reaps the rewards.
The authors thank Tessa Sundaram, MD–PhD candidate at the University of Pennsylvania, for her computer expertise and for helping to develop approaches for analysis. This work was supported in part by the Combined Degree and Physician Scholars Program at the University of Pennsylvania School of Medicine.