Both the Institute of Medicine1 and the Association of American Medical Colleges2 have called for increased racial and ethnic diversity of the medical profession. A necessary precursor to attaining such professional diversity will be a substantial increase in the racial/ethnic diversity of students selected to enter medical school. However, we have seen recent declines in the number of students applying to medical school from underrepresented minority racial and ethnic groups (URM), which include African Americans, Latinos, and Native Americans.3
The decline in URM applicants and graduates has been even more acute in California. After enactment of the state's Proposition 209 in 1996, prohibiting the use of race or ethnicity in admissions decisions by state medical schools, the combined number of URM students entering the five University of California medical schools dropped from a high of 117 in 1992 to 63 in 2000. There were equally sharp declines in the number of URM students applying to medical school.4
Private universities in California, such as Stanford University, are not constrained by the dictates of Proposition 209. In both its undergraduate admissions and its medical school admissions, Stanford has continued to support the goals of attracting and maintaining a diverse student population. During the three years of our study, an average of 108 students per year (29%) of the 367 incoming freshman undergraduates at Stanford who expressed interest in careers as physicians were from one of the three URM groups.
Despite the diversity of the cohort of incoming premedical students, at some point in the four-year university experience, the premed “pipeline” of URM students at Stanford leaks badly. For every 100 white or Asian freshmen at Stanford who indicate at the beginning of their freshman year that they are premed, between 110 and 125 graduating white or Asian students will apply to medical school. For every 100 African American, Latino, or Native American freshmen who indicate that they are premed, fewer than 50 will apply to medical school. At some point in their undergraduate experience, 50 or more seemingly qualified URM students leave the premedical “pipeline” at Stanford every year.
There have been relatively few data published on why students at other institutions leave premedical studies. Lovecchio and Dundes5(p 719) reported a study of premed attrition at one small college, concluding that “most former premed students admitted organic chemistry had played a significant role in the change in their career plans.” In a more general study of why students leave science, math, and engineering majors (SME), Seymour and Hewitt6 (p 392) “posit that problems which arise from the structure of the educational experience and the culture of the discipline … make a much greater contribution to SME attrition than the individual inadequacies of students or the appeal of other majors.”
Through a longitudinal study at Stanford of three cohorts of incoming freshmen who self-identified as premed, our report addresses the questions of when and why the loss of interest in premedical studies among URM students takes place.
Selection of study participants
Stanford University administers a survey to all incoming freshmen that addresses students' academic and professional aspirations. One question asks, “At this point in your life, if you are thinking of pursuing a graduate degree, in which area(s) would you do so?” The university's office of institutional research provided us with identifying information on all incoming students in the fall quarters of 2002, 2003, and 2004 who responded to this question by indicating medicine as a probable area of study. This group of 1,101 students formed the initial study population. The Stanford Office of Human Subjects Research reviewed and approved our research protocol.
Survey instrument and administration
We developed a survey instrument, which we administered electronically using Remark Web Survey software (Gravic, Inc., Malvern, Penn). Approximately one week after freshman orientation each of the three years, we sent an e-mail to each incoming student in the study population, asking him or her to participate in our study. Those who agreed to participate clicked on a Web link, which took them to the consent page of our survey. We sent out two additional waves of e-mails to those students not responding to our initial request for participation, at two-week intervals.
After linking from the e-mail to our questionnaire, students were asked to do the following: “Please choose a whole number between 1 and 10 from the Interest Scale which best describes your current interest in being premed.” The students were then shown a 10-point scale, with the following prompts located at the numbers indicated:
10 So committed to premed that nothing can stop me
7 Probably will be premed
4 Probably will not be premed
1 Absolutely no interest whatsoever in premed
Students could indicate only a single whole number in response. In earlier piloting of our survey instrument on the cohort of freshmen entering Stanford in 2001, students consistently recognized that “being premed” meant to undertake a course of premedical studies in preparation for applying to medical school. The survey also asked students to indicate which, if any, immediate family members are doctors.
We used a test–retest methodology to assess the reliability of our 10-point interest scale. We requested approximately 15% of the respondents in one admission cohort, identified randomly, to complete a duplicate survey three months after having completed the initial survey. Twenty-four of the 38 repeat subjects completed the retest survey. The Pearson correlation coefficient between students' responses at these two times was 0.69 (P <. 001).
In addition to the data obtained from the student survey, the office of institutional research provided us with the following data on each student: gender, principal racial or ethnic group (as identified by the student), zip code of family residence, SAT verbal score, and SAT math score. We categorized students into the following principal racial/ethnic groups: white (not Hispanic), African American, Latino (Hispanic), Asian, and Native American. We used the median household income for the zip code of the student's family residence to estimate the household income of the student's family. This estimation method has been shown to provide an appropriate estimate of family income when actual family-level data are not available.7 We excluded students whose families lived outside the United States and students who indicated “other” for race/ethnicity rather than one of the five principal groups, giving a final study population of 1,056 students.
We maintained a list of those students who responded to our initial survey request, which we referred to as Time One (T1). We sent each of these students an e-mail between two and four weeks from the end of the freshman year, asking them again to link to our survey and again to respond to the question about their current level of interest in premedical studies, using the same 10-point visual scale. We sent two additional rounds of e-mails to those students who had responded to our initial survey at the beginning of the year but had not yet responded to our end-of-year survey. We referred to this second round of the survey at the end of freshman year as Time Two (T2). Finally, we repeated the survey at the end of the sophomore year, which was Time Three (T3). We report data on the 362 students who responded to all three surveys.
In the initial pilot study of our survey instrument, we tested for possible response bias by telephoning 61 students who did not respond at T2. On the phone we asked them to rate the strength of their interest in premedical studies; we then compared their responses with those of 143 students who had responded and participated in the T2 survey. The mean interest level at T2 of the initial responders was 6.73 (95% CI: 6.36–7.11), whereas that of the nonresponders was 6.44 (95% CI: 5.77–7.12).
Follow-up interview design and administration
We conducted a follow-up interview with 68 of the 362 responding students, administered between the end of the student's sophomore and senior years (Table 1). In selecting students for interviews, we attempted to include 20% of respondents and to balance students who had reported decreases in their levels of interest in premed with those who had reported increases in their levels of interest. We further divided these two groups into URM and non-URM, and then we randomly selected students from each group for interviews. In doing so, we oversampled URM students.
Students had the choice of in-person or phone interviews. In either case, we provided students with the approved consent form (written for in-person, verbal for phone). Each interview was recorded and subsequently transcribed verbatim. A URM interviewer interviewed all URM subjects; a non-URM interviewer interviewed the non-URM subjects.
The interview contained the following questions.
1. “What were the factors that led to the [increase] [decrease] in your level of interest in being premed?”
Before finalizing our interview scripts, we had conducted an initial pilot study in 2001, consisting of informal interviews with several premedical students enrolled in the undergraduate classes offered by one of the authors (D.A.B.). In those pilot-study interviews, subjects often mentioned particular courses that they said discouraged their interest in medicine. On the basis of those preliminary findings, we included as a follow-up to question 1:
2. “Were there any specific courses at Stanford that discouraged your interest in medicine?”
We then asked the following questions:
3. “What do you see as the key resources to help students succeed as premeds?” (If an interviewee asked what was meant by resources, the interviewer said, “Resources would be programs, offices, people, materials.”)
4. “What skills do you feel are most important to help students succeed as premeds?” (If an interviewee asked what was meant by skills, the interviewer said, “Skills would be abilities, talents.”)
5. “What things at Stanford would you change to make the premedical experience more positive for students?”
We analyzed data from each of the interview questions separately. For each question, two members of the research team took a sample of the transcripts and closely read the responses in those transcripts. The researchers then discussed these responses in depth. On the basis of their examination and discussion of this first sample of transcripts, the researchers identified an initial set of response categories. The two team members then took a second set of transcripts, closely read the response in each transcript, and identified response categories in this second sample. This was done to determine whether, and to what extent, the categories in the initial set of transcripts also represented the data in the second set of transcripts. After this initial coding process with two samples of transcripts, all transcripts (including those in the first two samples) were analyzed using the response categories derived from this process.
Having coded all the transcripts and recorded the frequency of principal response categories, we then undertook a series of comparisons for each question:
- We compared responses of students who had reported a loss of interest in premedical studies between T1 and T3 with those who had reported a gain in interest in premedical studies during this time frame.
- We compared the responses of URM students with those of non-URM students.
- We compared the responses of women students with those of men students.
Of the 1,056 students in our study population, 151 were African American (14.3%), 146 were Latino (13.8%), and 28 were Native American (2.7%) (Table 2). At T1, the mean responses for African American students, Asian students, and Latino students were within 0.25 points of each other, with the mean for white students approximately 0.5 points lower and that for Native American students approximately 0.5 points higher. By the end of freshman year (T2), the mean response had fallen somewhat for all five groups (the three URM groups, white students, and Asian students), falling the most steeply for Native American students. By the end of sophomore year (T3), the level of interest of Asian and Native American students had declined somewhat, that of white students had shown a moderate decline, and that of African American and Latino students had declined sharply.
Figure 1 compares the mean declines in interest level for individual students between T1 and T3, sorted by student race/ethnicity. Asian students had the smallest decline, and the three URM groups had the largest decline.
We used ordinary least-squares regression to evaluate the associations between each of the demographic variables and the value of the change in level of interest between T1 and T3 (Table 3). In the first step of the analysis (Model One), we entered a variable indicating URM status and a variable for female gender. Both URM status and female gender are independently associated with a larger decline in interest. In a separate analysis, we tested for an interaction effect between URM status and female gender. We found no interaction; entering the interaction term weakened the fit of the model (data not shown).
In the second step (Model Two), we entered median household income by zip code as a marker of family socioeconomic status. Although the coefficient for income was not significant, its inclusion resulted in the URM variable no longer having a significant association, suggesting an association between URM status and median household income. We tested for such an association, and we found that mean household income in zip codes of URM families was $62,439 (95% CI: $57,311–$67,577), whereas that of non-URM families was $83,332 (95% CI: $79,490–$87,175).
In Model Three, we added the variables for SAT scores and ratio, and the variable for the number of family members who are doctors. In this model, the only variable with a significant association is family members who are doctors. Female gender no longer has a significant association with the change in level of interest. URM status continues without a significant association. We find no significant association for either individual SAT scores or the ratio of SAT scores. Whereas the explained variance (as measured by the value of R2) has gone up in Model Three compared with Model Two, the overall fit of the model (as measured by the F value) has gone down. In a separate analysis, we used cross-tabulation analysis to determine whether (1) URM students and (2) women students tend to report fewer family members as doctors. We found both associations to be significant (data not shown).
The results of our analyses suggest three principal findings:
- Between T1 and T3, the premedical interest level of URM students declines more than the interest level of non-URM students; this decline is independent of gender.
- Between T1 and T3, the premedical interest level of female students declines more than the interest level of male students; this decline is independent of URM status.
- Both the number of family members who are doctors (a significant association) and family income (a weak but nonsignificant association) act as intervening variables that replace either URM status or gender as significant predictors of change in level of interest, in the models we tested.
Results of interviews
Factors that led to [increased] [decreased] level of interest in being premed.
Among all respondents, the most frequent response category for this question was courses taken at Stanford, mentioned by 36 of the 68 students we interviewed (53%) (Table 4). (Recall that students were not asked specifically about the influence of courses they had taken until the next question.) The second-most-frequently mentioned influence was contact with physicians, identified by 13 students (19%).
The way in which a student reacts to specific courses taken at Stanford as a freshman or sophomore seems to have an important influence on that student's ongoing level of interest in continuing in premedical studies. An especially positive response to those courses will be associated with an increase in interest, while an especially negative response will be associated with a decrease in interest. This contrast is illustrated in the following four responses to the question, “What were the factors that led to the [increase] [decrease] in your level of interest in being premed?”
Information. I took biology at Stanford—the bio core—so understanding the way they actually presented material in [the] medical school curriculum, as well as the topics covered.
—URM male, interest level increased
The realization that I could handle the academic workload. You hear all those rumors about “weeder” courses, but I saw that I could do it. I feel I have a decent grasp of what I expect and what is expected of me.
—non-URM male, interest level increased
Everyone says [Stanford's premed courses are] more like a weeding-out process than anything [else], and I just ended up being one of those people.
—URM male, interest level decreased
I think I experienced the same distaste [as other premeds] for the large premed classes, like the biology core. You think that no one wants to support you; they're just out to get you.
—non-URM female, interest level decreased
One third of students whose interest had increased during the period of this study identified courses they took as contributing to that increase, whereas three fourths of students whose interest had decreased during the period of this study identified courses they took as contributing to that decrease. The current courses offered to premeds at Stanford seem substantially more likely to discourage students' interest in medicine as a career than to encourage that interest.
Having had contact with a physician during the first two years of college also seems to affect the levels of students' interest. Examples of this contact include speaking with a physician about career options or observing a physician in a clinical setting. In this case, that contact seems more likely to encourage students in maintaining their interest (8 of 35 students, 23%) than to discourage that interest (5 of 33 students, 15%). An exposure to research also seemed to play a role in increasing interest in premed for some students.
It should be noted that among the 33 students who lost interest in premed, 11 (33%) reported that their interests had simply changed, without identifying a specific contributing factor. It is difficult to know whether this change in interest is in reaction to or independent of the other negative experiences reported.
In comparing the responses of URM students with those of non-URM students, and female students with male students, the pattern of responses is generally the same. Experience with courses was mentioned most frequently, with those losing interest substantially more likely to mention the effect of courses. The exception is male students regardless of ethnicity, who mentioned the influence of courses at about the same rate for those whose interest increased and those whose interest decreased.
Specific courses mentioned by students.
Of the 35 students we interviewed whose interest in premed increased, 26 (74%) still identified at least one course they took during their first two years at Stanford as tending to discourage their interest in careers in medicine (Table 5). Of the 33 students whose interest in premed decreased, 28 (85%) identified at least one course they took during their first two years at Stanford as tending to discourage their interest in careers in medicine. Whereas there were minor differences among men and women (men reported courses that discouraged them somewhat less often), the responses between URM and non-URM groups were quite similar.
When we then look to see what specific courses students mentioned as discouraging their interest in medicine, we see a striking pattern: students identified chemistry courses between four and five times more often than the next category, biology. Other courses, such as physics and math, were mentioned only rarely. It is also instructive to note that among students whose interest in premed had decreased, students often mentioned more than one chemistry course as having contributed to that decline.
Among the 54 students in Table 5 who identified a chemistry course as tending to discourage their interest, organic chemistry was the course mentioned most frequently (19 students). Sixteen students mentioned chemistry as a generic subject without mentioning a specific course, and 16 mentioned inorganic chemistry. These findings confirm those of Lovecchio and Dundes5 described above. However, our findings suggest that the discouraging effects of studying chemistry in general as part of the early premedical curriculum are more extensive than the effects of organic chemistry alone.
Below, we provide excerpts from the interview texts that are typical of students' responses to the question, “Were there any specific courses at Stanford that discouraged your interest in medicine?”
Chem 33 and Chem 36 [both organic chemistry] kinda discouraged me. It was difficult to get a helping hand.
—non-URM male whose interest level increased
—URM male whose interest level increased
Chem 31 [inorganic chemistry] and math—first quarter calculus—huge anonymous classes with bad TAs.
—non-URM female whose interest level decreased
… [T]he chem core, Chem 31/33. It's tough and I think there are a lot of students. There's not a lot of professor–student contact. I felt the professor [was] somewhat abrasive at times.
—URM female whose interest level decreased
Skills and resources to help students succeed as premeds.
When asked about the skills that help students succeed as premeds, students responded most often with factors such as determination, perseverance, hard work, time management, and ability to focus. Whereas an occasional student would mention factors such as humility, spiritual grounding, or compassion, the clear pattern was for students to see success as based on the force of one's will to succeed. This pattern remained consistent in comparing URM with non-URM groups, and women with men.
When asked to identify the most important resources that help students succeed in premedical studies, respondents consistently identified three principal factors: having high-quality advisors, having the support of other students, and having the support of faculty. We noted two interesting patterns in these responses:
- Whereas men and women were equally likely to mention the importance of the university's advising system (30%–40%), women were nearly three times as likely to mention the importance of peer support (14 of 46 women, or 30%, versus 4 of 34 men, or 12%).
- In comparing URM students with non-URM students regarding key resources, we identified no significant differences in response. However, of the nine African American students we asked, only one gave any response; the others were unable to identify resources that are conducive to success.
Students seem to be describing a combination of personal strength plus institutional support as key to success for premedical students. Men and women look to different aspects of the institution for support—men to the university's advising system reinforced by contact with faculty, women to the advising system reinforced by the support of their peers. However, the experience of African American students may not be fully reflected in these responses, because so few responded to our question about key resources.
Students' suggestions for changes to make the premedical experience more positive.
Students were consistent across interest level, across gender, and across URM status as to the principal change they would recommend to improve the premedical experience: improve the university's advising system.
Students also recommended changes such as increasing their exposure to practicing physicians, improving teaching (presumably in premedical courses), and fostering more structured peer-support systems. However, when it came to the university working to improve the premed experience, improving the advising system took priority. The following excerpts from responses to the question, “What things at Stanford would you change to make the premedical experience more positive for students?” illustrate this point.
I think the accessibility to advisers. I think I would change the system of premedical advising at Stanford…. At a lot of other schools … you get direct access to a premed adviser … on a one-to-one basis by name from a very early start in your undergraduate career.
—URM male whose interest level increased
You know, it might be helpful if the premed advisers were better…. The freshman advisers are not that helpful.
—non-URM male whose interest level increased
I don't really know, because it's very hard for me to distinguish what I'm lacking from what the university is doing wrong. And the premed advisers, although they try to be helpful, end up putting a lot of unnecessary pressure on students.
—non-URM female whose interest level increased
Positive advising—advising that says, “You're committed [to premed] and we're going to help you.”
—non-URM female whose interest level decreased
Although only a small number of students recommended improvements in teaching as a means to improve the premedical experience, the comments of those who did mention it hold particular relevance in light of the factors discussed above that discourage continued interest in premed.
Definitely change chemistry. But I'm not sure how I'd change it—just very discouraging.
—non-URM female whose interest level increased
I don't know if it would be possible to change them, but—like the chemistry classes—tailoring them more towards … making it premed.
—URM male whose interest level increased
I felt that the chemistry courses were designed to weed people out and allows a kind of disconnect between the courses that the people had to take for the premed requirements and the actual type of medicine or career interest they wanted to pursue.
—URM female whose interest level increased
Among entering students at Stanford University who initially are premed, women and URM students are less likely to maintain their interest in a medical career through four years of undergraduate studies. Although gender and ethnicity each seem to exert an influence independently, there is an important gender skewing that makes these effects cumulative. Whereas 61% of the non-URM students in our study sample were female, 74% of URM students (88% of African American students) in our sample were female.
In the interviews we conducted, students in our study tell us that their interest in becoming a physician was affected most profoundly by their early college experience in the chemistry classroom. We did not identify clear differences in the influence of chemistry courses when we compared URM and non-URM students with similar levels of interest in premed, or women and men. What we did find was a striking difference in the impact of chemistry between students whose interest in continuing as premeds increased during the course of their early college experience and those whose interest decreased. Combining these results with the analysis of our survey responses, we can reasonably conclude that the negative influence of chemistry courses on continued interest in premed is experienced more so by women and URM students. In light of the fact that 74% of URM premedical students at Stanford are women, it is not at all surprising that Stanford's URM students are less than half as likely as non-URM students to persist in their interest in premed and eventually to apply to medical school.
A number of students used the term weeder course to describe their experience in the chemistry classroom. Students perceive success in chemistry as essential to gaining admission to medical school. Although Stanford University publishes no formal premedical curriculum, a number of students have told us that if, rather than beginning chemistry in the first quarter of college, a freshman premed waits until the second quarter to begin, he or she is “off track” in premed (i.e., behind the other premedical students).
This belief—that early success in chemistry as an undergraduate is essential to success in medical school—is a century old. In 1910, Abraham Flexner8(p 24) wrote the following concerning the premedical curriculum: “The normal rhythm of physiologic function must remain a riddle to students who cannot think and speak in biological, chemical, and physical language.” When Flexner wrote these words, few students approaching medical school had any exposure to chemistry.
Writing in 1924, Sinclair Lewis9 echoed Flexner's view in describing the challenge facing young Martin Arrowsmith, his fictional premedical student.
In 1904, Martin Arrowsmith was an Arts and Sciences Junior preparing for medical school…. Martin's father and mother were dead, leaving him only enough money for his arts and medical courses. The purpose of life was chemistry and physics and the prospect of biology next year.9(p7)
The pedagogy of the chemistry classroom has evolved in the 100 years since the Flexner Report, and it has a substantial record of success in training students for careers involving the application of chemical knowledge. It seems neither reasonable nor desirable to change that pedagogy in response to the negative experience of premedical students. The university may, however, want seriously to consider establishing an alternative pedagogical approach to the teaching of chemistry that is better suited to students interested in the health professions.
The principles of chemical knowledge essential for success in medical school and medical practice represent a subset of the chemical principles essential for success as a chemist. To develop an alternative curriculum and pedagogical approach to that curriculum may hold the promise of providing a more focused, and more appropriate, knowledge of chemistry, suitable to the needs of future health professionals. Such an alternative approach may also contribute to enhancing the racial and ethnic diversity of the medical profession, a goal on which there is widespread agreement.
The students' responses also give direction as to steps the university can take to reduce the loss of interest many students experience. Although not intending to do so, the advising system for undergraduates may reinforce students' early negative experiences. Rather than responding to a student who has had an early negative experience in the chemistry classroom by discouraging that student from continuing in premed, advisers could work to support these students, sharing with them the knowledge that negative experiences in chemistry are common and should not be taken as a disqualifier for entry into medical school. Such advice must also be coupled with information on available alternatives for finding a more positive experience in chemistry and other premedical sciences. The university may also need to invest in more formalized and more extensive systems of peer support among premedical students, and in creating mentoring opportunities with practicing physicians.
This study involves students from only one university, a university which is among the most selective nationally in its undergraduate admissions. Many if not most of the students who lost interest in premedical studies would be likely to succeed in medical school, given an appropriate academic preparation. It is therefore difficult to generalize our findings to other colleges and universities. Our findings do point to possible explanations, however, as well as to possible interventions that may have relevance for other colleges and universities experiencing a similar decline in interest among URM students in pursuing careers in medicine. We encourage those institutions to gather similar data on their students to assess the extent of this relevance.
This research was funded by a grant from the California Wellness Foundation.
The authors also gratefully acknowledge the assistance of the staff of the Center for Teaching and Learning at Stanford University in administering our Web-based survey.