Our nation's health care is directly affected by the numbers of internists in practice, yet the percentages of U.S. medical students selecting careers in internal medicine (IM) continue to decline according to the National Residency Match Program (NRMP).1,2 The reasons underlying this shift away from IM are not well understood, but recent studies suggest that student career choice is likely dependent on many factors.3 Further exacerbating this problematic trend is the fact that the demand for internists is increasing with the aging patient population.4
Some academic internists have envisioned a “perfect storm”: a growing demand for internists, a lower percentage of U.S. medical school graduates selecting IM, and young physicians leaving IM to go into subspecialties or other career fields altogether.4–6 Educators, policy experts, and leaders in academic IM have all expressed concern over these trends.1,3,5,6
We recently described factors associated with choosing IM as a career: satisfaction with educational experiences in IM, the nature of patient care in IM, and lifestyle.3 In the present study, we sought to identify further modifiable factors driving medical students' career choice of IM, here focusing on factors that influence students who were initially interested in IM but ultimately chose another field (“Switchers”). We compare these Switchers with students who chose IM careers (“Choosers”) and with those who never considered IM (“Never Considered”).
We conducted a cross-sectional survey of fourth-year medical students at 11 U.S. medical schools. The methods for this cross-sectional survey study have been described in detail elsewhere3 and are briefly highlighted below.
Following a MEDLINE review of the literature about medical student career choice, seven of the authors, who are members of the Clerkship Directors in Internal Medicine, conducted focus groups at seven participating schools of medicine in 2006–2007 with a total of 43 fourth-year medical students to elicit current issues influencing their career choices. The participating schools were Uniformed Services University of the Health Sciences; George Washington University School of Medicine; New York University School of Medicine; University of California, San Francisco, School of Medicine; University of Pittsburgh School of Medicine; Warren Alpert Medical School of Brown University; and Yale University School of Medicine. We then revised items from a 1990 career choice survey7 after reviewing transcripts from these student focus groups. The revised survey was pilot-tested online with the study investigators in the early spring of 2007 before being sent to students to complete. Factor analysis was employed in our earlier study3 to examine the validity of the survey inferences.
The survey3 included 24 questions about demographics, debt, experiences on the core IM clerkship and subinternship (type of hospital, presence of inpatient/outpatient experience, grades), specialties chosen or considered, and IM interest group participation. Twenty-four questions addressed students' perceptions of IM compared with other specialties chosen and considered, using a Likert scale (1 = this issue is much less in IM than other specialties, 3 = same, 5 = much more, 6 = don't know). Thirty-two questions asked students to rate factors that influenced their career decisions regarding IM (1 = very much pushed me away from IM, 3 = no influence, 5 = very much attracted me toward IM).
The electronic survey was sent to fourth-year medical students at 11 schools in the spring of 2007, after they had submitted their choices for the residency match but before match day. Nonresponders were sent up to five follow-up emails. The 11 participating institutions (George Washington University School of Medicine; New York University School of Medicine; Uniformed Services University of the Health Sciences; University of California, San Francisco, School of Medicine; University of Chicago Pritzker School of Medicine; University of Florida College of Medicine; University of Pittsburgh School of Medicine; University of Utah School of Medicine; Warren Alpert Medical School of Brown University; Washington University in St. Louis School of Medicine; and Yale University School of Medicine) represented a range of regions, public/private status, level of research funding, and historical percentage of students matching in IM. Each participating school's institutional review board approved this study.
In this study, we focus on individual survey items that differ between our three defined groups as a means of further understanding why IM is chosen (Choosers), seriously considered but rejected (Switchers), or never considered (Never Considered) as a specialty. We categorized all survey questions into three groups of variables: student demographics, medical school experiences (particularly the IM clerkship), and aspects of the specialty of IM. The primary outcome variable for the analysis was specialty choice defined as Switchers, Choosers, and Never Considered. Students identified their career plans from a list of 46 specialties as defined by the NRMP. Choosers selected IM, IM primary care, or any combination of IM and another specialty recognized as IM by the American Board of Internal Medicine (e.g., IM with dermatology, emergency medicine, family practice, medical genetics, neurology, pediatrics, or psychiatry). Preliminary IM and transitional-year students were not included as Choosers. Because of some disagreement in the literature regarding what fields should be considered primary care versus nonprimary care, we chose to analyze Switcher and Never Considered career choice by both traditionally considered primary versus nonprimary care1 and person-oriented versus technique-oriented specialties.8
Univariate analysis was performed with chi-square. We calculated means and standard deviations (SDs) for variables by group. For comparing three variables simultaneously, one-way ANOVA or Kruskal–Wallis tests were used. All variables from the univariate analysis that demonstrated significant differences (two-tailed P < .01) among the three career groups were retained for nominal logistic regression analysis. With the nominal logistic regressions, we used the likelihood ratio test to determine significance. Because of the large number of variables, we used a two-tailed P < .01 to decrease the likelihood of spurious associations. In our multivariate models, the dependent variables were the three career choice outcomes (Chooser, Switcher, Never Considered). A regression model was created for each of the three variable groups (demographics, medical school experiences, and aspects of IM specialty). Variables significantly contributing to these models were then combined to create an overall model differentiating the three career groups. The contributions of variables are expressed as χ2 and P values. For our final model, we calculated odds ratios and 95% confidence intervals (CIs) for each significant variable. Receiver operating characteristic (ROC) curves demonstrate the variance explained for each career decision path.
All analyses were performed on JMP software (SAS Institute, Cary, North Carolina) or SPSS (version 16.1, Chicago, Illinois).
Of the 1,177 of 1,439 fourth-year medical students who completed the survey (82% response rate), 274 (23%) chose IM (Choosers), 399 (34%) seriously considered IM but switched to another field (Switchers), and 504 (43%) never considered IM (Never Considered). These percentages are similar to 2007 NRMP match data for U.S. students, which showed that 22.3% of students matched to IM or IM combined fields.9
Responding students' demographic characteristics by career choice groups are shown in Table 1. Mean (SD) age was 27.7 (3.2) years, and respondents were almost evenly split regarding gender. Our results are comparable to all U.S. medical students completing the Association of American Medical Colleges Graduation Questionnaire in 2007, among whom 82.9% were 24 to 29 years old and 49.6% were men.10 There were no significant differences among Switcher, Chooser, and Never Considered groups regarding age, gender, race, marital status, debt burden, and having children (all P > .05). The only difference between groups was that a higher percentage of Choosers came from private schools (P = .005).
Career choices for the Switcher and Never Considered groups are shown in Table 2.
No significant differences in career choice were seen between groups using either traditional definitions of primary care versus nonprimary care or person versus technique orientation.
Timing of career decisions differed significantly across the three groups (P < .001). Switchers made their career decisions later than Choosers or Never Considered. Few Switchers and Choosers reported making their specialty decisions before entering medical school (7%, n = 19 and 12%, n = 32, respectively); in contrast, 29% (n = 146) of the Never Considered group made their career decision prior to medical school. Fifty-two percent (n = 207) of Switchers made their career decision during the core clerkship year and 41% (n = 164) after the core clerkship year. For Choosers, 47% (n = 129) made their decision during and 41% (n = 112) after the IM clerkship. Forty-four percent (n = 202) of Never Considered chose their career during the clerkship year and 27% (n = 136) after the clerkship.
Medical school educational experiences
Results from questions relating to respondents' educational experiences in medical school are shown in Table 3. In univariate analysis, Choosers were more satisfied with their core IM clerkship (mean [SD] of 4.23 [1.0] versus 3.99 [1.2] versus 3.74 [1.2] for Choosers, Switchers, and Never Considered groups, respectively; P < .001; scale of agreement with statements about satisfaction: 1 = strongly disagree, 3 = neutral, 5 = strongly agree). Choosers were more likely to have obtained honors grades in their IM clerkship than were Switchers and Never Considered (55% versus 49% and 38%, respectively; P < .001). However, Choosers reported feeling less influenced by their grades when making their career decision than did the other two groups (mean [SD] of 2.17 [0.93] versus 2.25 [0.82] versus 2.54 [0.67] for Choosers, Switchers, and Never Considered, respectively; P < .001). Choosers were more satisfied with their subinternships in IM than were Switchers and Never Considered (mean [SD] of 4.34 [1.02] versus 4.12 [1.04] versus 3.96 [1.14], respectively; P < .0001). Choosers also reported that their core IM clerkship made careers in subspecialty IM more attractive (mean [SD] of 2.63 [0.63] versus 2.32 [0.79] versus 2.10 [0.76] for Choosers, Switchers, and Never Considered, respectively; P < .001; scale of 1 = less likely to choose IM, 2 = neutral, 3 = more likely to choose IM). In the regression analysis of medical school experiences associated with career choice, three items remained significant: satisfaction with core IM clerkship, core IM clerkship made a career in subspecialty IM seem attractive, and participation in an IM interest group influenced career choice.
Aspects of the specialty of IM
Students' perceptions regarding numerous aspects of the specialty of IM (perceptions and influences) revealed differences between the three career groups (Table 4). Switchers' responses were found to be between those of Choosers and Never Considered. Choosers had the most favorable impression of the six significant items in the regression analysis: type of patient internists see, high intellectual challenge of IM, perceived satisfaction among IM residents, a feeling of being recruited to IM, continuity of care in IM, and research opportunities in IM. These items were rated most highly for Choosers, followed by Switchers and then Never Considered.
In our final regression model, we combined timing of career choice and the significant variables from medical school educational experiences and nature of patient care in IM regression models (Table 5). The six contributing items are shown in Table 5. Of these, one describes students, two describe medical school experiences, and three describe aspects of IM practice. The most influential item in the model was the type of patients that internists see. We calculated adjusted odds ratios. Using an alpha of .05, all but one item (“IM interest groups influenced my career choice”) had an odds ratio >1 (range for lower CI 1.15–3.2, range for upper CI 3.0–7.5). In this final model, we also performed odds ratios comparing Never Considered with Choosers and Switchers as well as Choosers with Switchers and Never Considered. With our odds ratios, satisfaction among attending internists had a 95% CI that crossed 1 for the former (comparing Never Considered with Choosers and Switchers) and all but “IM interest groups influenced my career choice” had CIs >1 in the latter (comparing Choosers with Switchers and Never Considered). This model had areas under the ROC curves of 0.94 for Choosers, 0.81 for Switchers, and 0.84 for Never Considered, meaning that the majority of variance (94%, 81%, and 84%, respectively) was accounted for in our models.
Major decisions such as career choice likely result from the complex interplay of past experiences, current perceptions, and future aspirations.10 We sought to explore the differences regarding IM career choice between Switcher, Chooser, and Never Considered groups. Our multivariate regression results accounted for over 80% of the variance in each group.
Although implications of our study findings for the size of the IM workforce are noteworthy, we wish also to point out that increasing the quantity of IM physicians by attracting Switchers will not necessarily increase quality, because Switchers' attitudes may be less positive about IM than their Chooser peers'. To increase the IM workforce in a meaningful way, we need to address shortcomings that are modifiable so that quality can follow increased quantity. If all the Switchers had chosen IM, the number of students entering IM in our sample would have more than doubled (274 to 672). Further, our findings indicate that IM loses significant numbers of students to other primary care fields. Roughly equal percentages of Switchers and Never Considered went into a primary care field other than IM. Our findings are consistent with a prior study that found that early interest in surgery and psychiatry seems more “stable” than interest in other fields.11 This finding is also consistent with a recent cross-sectional and longitudinal study that found that only 30% of those initially interested in primary care remained interested at all longitudinal time points.12 Although prior studies have tended to categorize specialty choices as “primary care” versus “nonprimary care,” our finding that one in three students in the Never Considered group entered another primary care field contradicts the notion that all primary care fields are alike. Likewise, with grouping by person-oriented versus technique-oriented specialties,8 there were no significant differences between the Switcher and Never Considered groups (Table 2). The reasons why these groups chose another primary care (or person-oriented) field are not clear. Perhaps these students were not (or were less) attracted to specific aspects of the IM practice, including the types of patients an internist sees, complexity of practice, and controllability of lifestyle in terms of work hours, practice environment,3,13 and/or perceived satisfaction among IM residents. We believe that further informing medical students about IM and how it contrasts with other primary care fields is a potentially modifiable factor. Those in other primary care fields earn similar amounts of money, and thus it is unlikely that income or debt influenced decisions. A high percentage of Switchers choose fields that tend to offer a more controllable lifestyle and higher pay. Finally, although convincing students to choose IM is a goal of IM organizations, this path may not necessarily be a solution for the primary care workforce problems because so many IM residents later subspecialize.
Timing of career decisions differed between the three groups, suggesting opportunities during medical school for efforts to attract Switchers to IM. Nine of every 10 Switchers chose their field during or after the third-year clerkships. Thus, interventions to attract Switchers should likely be focused on the clerkship years. Focusing efforts to attract students to IM during the clerkship year would represent a potentially modifiable factor. However, merely changing the clerkship without addressing larger profession and practice issues is unlikely to attract more students to IM. This point was illustrated in our final model (Table 5) which showed that the most influential item swaying students toward or away from IM was the type of patients that internists see. Understanding this negative perception deserves further study.
What leads Switchers to choose another field? The majority of items in this multivariate analysis follow an expected progression of being most attractive to Choosers and least attractive to Never Considered. The “dose response” findings for Choosers to Switchers to Never Considered groups for most items suggest that increasing efforts directed at Choosers could also influence additional Switchers to consider IM. These items in Table 4 represent potentially modifiable factors. Our regression results suggest that Switchers' decisions were negatively influenced by the types of patients internists see. This concern accounted for the largest proportion of variance in the current ROC analysis and in our prior paper.3
The core IM clerkship seemed to make a career in subspecialty medicine more attractive for Switchers as well as Choosers. Being exposed to subspecialists in IM was important in our multivariate analysis. Perhaps exposure to physicians who are expert in their subspecialty and who work in a different practice environment than general IM is important. Subspecialty practice environments often allow physicians to focus on fewer problems per visit, frequently have longer appointment lengths, and receive higher reimbursement. These areas represent factors that could potentially be modified to attract students to IM.
Multivariate nominal regression analysis characterizes the powerful role that residents can play in students' career choices regarding IM (Table 4). The “dose response” pattern (Never Considered, Switchers, Choosers) in the multivariate models was seen with all items with two exceptions—Switchers, more so than Choosers and Never Considered, (1) were the most deterred from IM by a perceived low satisfaction among IM residents and (2) found research opportunities in IM to be more neutral in terms of attracting them toward careers in IM. Thus, residents may exert a more important influence on IM career choice than has been previously cited,7 which represents another potentially modifiable factor. Whereas residents' satisfaction was seen as neutral for Choosers, it had a negative influence for the other groups. Further, all groups were pushed away from IM by their perceptions of practicing internists' and IM residents' job satisfaction. This finding may reflect the emerging “lifestyle” issues for career selection and the need for positive role models.13–15 We believe this is an important area for emphasis to increase the number of internists. Programs such as Residents as Teachers and the Stanford Faculty Development Course could improve residents' attention to the learning environment and improve teaching satisfaction which would also likely improve their students' perceptions of their satisfaction. These programs represent potentially modifiable factors. Research opportunities in IM did not attract Switchers toward IM; this finding contrasts with prior work in this area.15
IM interest groups seem to be important to career selection, though attendance may reflect already strong interest in IM. Interest groups offer the opportunity to showcase examples of all of the items listed in Table 3—attractive aspects of IM associated with career pathways. Mentors, information about IM, research opportunities, and allowing the medical student to see intellectual challenges and satisfied internists as well as residents may be the reasons for this item's persistence in both univariate and multivariate models, which represent potentially modifiable factors. Given our finding that most students select their career during the third year, perhaps more IM interest groups and other activities that can expose students to positive role models should be directed to this year. However, students in all three groups perceived that recruiting efforts minimally affected their career choices; they may prefer to receive information but feel that recruiting (trying to actively persuade a student to go into IM) is biased.
Two items that have previously been reported to potentially influence career decisions did not seem to influence career selection for our three groups—debt and participation in an IM ambulatory rotation. From our models, debt seems to affect all three groups equally, and importantly it doesn't seem to lead to an increased percentage in Switchers or Never Considered, perhaps because IM is the pathway to some high-paying subspecialties, such as cardiology and gastroenterology.
Our study has several limitations. We did not survey all U.S. medical students, although we had a large sample size and high response rate from a diverse group of schools. Because we did not study students longitudinally, recall bias may have influenced responses regarding specialties previously considered but not chosen, and we cannot determine whether students will change their specialties later. Data were based on self-report; we did not confirm students' stated specialty choice with match results. Though it seems that subspecialty experience may be important for Choosers and Switchers, our study was not designed to further ascertain the potential impact of this phenomenon, and we believe that this should be explored in future studies. Though published in the literature, some may not agree with the division of specialties as person- or technique-oriented.8 We only sought to determine what is associated with IM specialty choice; our findings may or may not be germane to specialty choice for fields outside IM. Finally, our findings represent associations with selection of IM as a career choice and not necessarily causation.
The percentage of variance explained by group is noteworthy and suggests that our findings may incorporate some important aspects of career decisions. A few potentially modifiable educational experiences and aspects of our specialty distinguished Switchers from the other two groups. These items suggest ways that educational experiences and aspects of the specialty could be redesigned to improve the attractiveness of careers in IM—moving some from “almost” internists to internists.
The Shadyside Hospital Foundation of Pittsburgh, Pennsylvania, and the American Board of Internal Medicine Foundation funded the study. These organizations did not participate in the study's design; conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
This study was approved by the institutional review boards at all participating institutions: George Washington University School of Medicine; New York University School of Medicine; Uniformed Services University of the Health Sciences; University of California, San Francisco, School of Medicine; University of Chicago Pritzker School of Medicine; University of Florida College of Medicine; University of Pittsburgh School of Medicine; University of Utah School of Medicine; Warren Alpert Medical School of Brown University; Washington University in St. Louis School of Medicine; and Yale University School of Medicine.
The views expressed herein are those of the authors and do not reflect the official policy of the Department of Defense or other federal agencies.