Effectiveness of University of California Postbaccalaureate Premedical Programs in Increasing Medical School Matriculation for Minority and Disadvantaged Students FREE

A racially and ethnically diverse physician workforce is important for increasing access to care for underserved populations, improving the cultural competence of the workforce, and enhancing the educational experience of all medical students.^1- 5 However, many ethnic groups remain underrepresented among physicians in the United States. In 2000, blacks, Latinos, and Native Americans comprised more than 25% of the US population but only 7% of the nation's physicians.¹ Moreover, students from these underrepresented ethnic groups constitute a decreasing proportion of US medical students. In 1994, underrepresented minorities comprised 15.5% of students matriculating into Liaison Committee on Medical Education–accredited medical schools, but by 2000 this proportion had decreased to 13.8%.² Students from lower-income families are also much less likely than those from higher-income families to matriculate into medical school. Sixty percent of US medical students come from families in the top quintile of household income.⁶

One strategy for increasing the diversity of the physician workforce is to implement interventions to support the academic achievement and health career aspirations of minority and low-income youth. A wide variety of educational “pipeline” interventions exist in the health professions, ranging from assisting elementary schools to develop curricula in science and mathematics to providing support for the hiring of minority faculty members at medical schools.^2,7 Among these programs are postbaccalaureate premedical programs targeting minority and disadvantaged students. These programs enroll college graduates, most of whom have previously applied unsuccessfully for admission to medical school, and provide an enrichment experience with the aim of making students more competitive medical school applicants. Currently, more than 75 academic institutions offer nondegree postbaccalaureate premedical programs, many focused on minority and disadvantaged students.⁸ Other postbaccalaureate programs focus on preparation for dental school.⁹

Postbaccalaureate programs receive funding support from a variety of sources, including federal Title VII funds administered by the Bureau of Health Professions, state governments, private foundations, and sponsoring institutions. Federal support for these types of educational pipeline programs in the health professions has recently been markedly reduced. For example, federal funding of the Title VII Health Careers Opportunity Programs decreased from $35.6 million in federal year 2005 to $4.0 million in federal year 2006, a reduction of 89%.¹⁰

Case studies of postbaccalaureate premedical programs suggest that many participating students subsequently gain admission to medical school.^11- 14 However, we are not aware of published well-controlled studies of the educational outcomes of programs designed exclusively for postbaccalaureate students. One well-controlled observational study found a positive effect on medical school matriculation following a summer premedical enrichment program for minority students, but this program primarily included students still enrolled in college, and the study did not separately examine the relatively few postbaccalaureate students in the program.¹⁵ Another study of a preadmission premedical program for minority students found that participating students were more likely to matriculate into medical school than the overall population of minority students applying to US medical schools but did not adjust for possible underlying differences in the academic or other characteristics of the program participants and comparison group.¹⁶ Inclusion of a valid control group is essential for determining the degree to which positive program outcomes may be attributable to the program intervention itself rather than to a program's selection of students with more favorable underlying attributes that make them more likely to eventually succeed in being admitted to medical school.

To address this deficit in the existing evaluation literature, we conducted an investigation of University of California (UC) postbaccalaureate premedical programs. We measured program outcomes—matriculation into medical school—and included a control group to account for possible selection effect.

The University of California operates medical schools at the Davis, Irvine, Los Angeles, San Diego, and San Francisco campuses, each of which administers a postbaccalaureate premedical program conforming to relatively standardized admissions procedures and curricula. Students admitted to the postbaccalaureate program are not guaranteed admission to a UC medical school on completion of the program. All 5 UC postbaccalaureate programs accept students who have previously applied to medical school without gaining admission. Two campuses also accept first-time medical school applicants, but these students represent less than 20% of all program participants.

The programs implement a rigorous application process designed to mimic that used by the American Medical College Application Service (AMCAS). Applicants must submit undergraduate transcripts, a personal statement, Medical College Admissions Test (MCAT) scores if available, and letters of recommendation. Following an initial evaluation based on these application materials, a group of applicants is invited for interviews and the final cohort is selected from this group. The postbaccalaureate programs enroll applicants who are California residents from disadvantaged backgrounds and who meet minimum academic criteria. Additional preference is given to students who express a strong interest in practicing medicine in underserved communities. Applicants with the highest grade point averages (GPAs) and MCAT scores are not necessarily ranked highest for admission, as the programs aim to identify students who might benefit the most from an enrichment program.

All 5 UC programs offer preparation in studying, test taking, writing personal statements, and interviewing, as well as opportunities for clinical or research experiences. All programs conduct an intensive summer session that focuses on MCAT preparation, followed by at least 2 semesters of academic work in which the participants enroll in undergraduate science courses to strengthen their academic record. Throughout the year-long program, students also attend seminars taught by medical school faculty on topical issues such as health disparities and career pathways in medicine, as well as group events intended to foster peer support among participants. Programs also provide access to learning specialists on an as-needed basis for assessment of possible learning disabilities.

Data on postbaccalaureate program participants were obtained from a centralized database maintained by the UC Office of the President, Department of Academic Advancement. This database includes baseline information on each student at the time of program application, including name, overall college GPA, cumulative MCAT score for students who previously took this examination, undergraduate major, race/ethnicity, sex, date of birth, and the highest level of education attained by each of the applicant's parents. Race/ethnicity was self-reported by students in their program applications. Some programs used open-ended questions, while others listed racial/ethnic categories as check boxes (with varying use of “black” and “African American”). This database also includes follow-up information indicating whether the student was accepted to and matriculated into a medical school by the 2005 academic year.

For the group of controls, we selected college graduates who had applied to UC postbaccalaureate premedical programs but did not participate in a UC program. All of the controls had been denied admission to a UC postbaccalaureate program, except for 4 who were admitted but did not attend. Specific reasons for rejection were not available. For each control, we used applicant files stored at 4 of the UC programs to abstract data for the same baseline variables included in the program participant database. A unique list of controls was created, eliminating duplicate entries for students who had applied to more than 1 UC program. Applicant files were not available for the program administered at the Los Angeles campus; however, because almost all controls had applied to more than 1 UC program, it is unlikely that a large number of additional controls would have been identified who had applied only to the Los Angeles program.

To ascertain medical school matriculation outcomes for the controls, we provided identifying information for each control to the Association of American Medical Colleges (AAMC) Division of Medical Student Services and Studies. Using students' name, sex, and birth date, AAMC staff searched the AAMC computerized files to determine which students had matriculated into allopathic medical schools by the 2005 academic year. Due to AAMC policies regarding data confidentiality, the office could not release information on which of the nonmatriculating controls had applied to medical school through the AMCAS. However, the office did provide aggregate information on the number of controls who had submitted applications through AMCAS. We also contacted the Association of American Colleges of Osteopathic Medicine to obtain comparable data on controls. However, the association does not maintain a centralized database on applicants and matriculants, and we were therefore unable to ascertain whether controls had matriculated into an osteopathic medical school. We were not able to obtain information on matriculation in non-US medical schools for either participants or controls.

We restricted this study to participants and applicants for UC postbaccalaureate programs in the 1999-2000 through 2002-2003 academic years to allow adequate follow-up duration. During this period, there were 706 unique UC postbaccalaureate applicants, 268 who participated in the programs and 438 who did not. From this sample, we excluded 3 participating students and 42 controls for whom baseline data were missing for GPA, sex, or whether the student had previously applied to medical school, resulting in an analytic sample of 265 students in the participant group and 396 in the control group (Figure). There were no statistically significant differences between the included and excluded controls in baseline GPAs and MCAT scores, although the excluded students were somewhat older and more likely to have applied in earlier program years. We included students for whom data were missing for race/ethnicity, age, and highest level of parental education. We also included 129 students who did not have data recorded for baseline MCAT scores because 123 were first-time applicants and would therefore have not been likely to have taken the MCAT by the time they applied to a postbaccalaureate program. MCAT data were missing for 28 of the 265 participants and 101 of the 396 controls. Missing data among students included in the study were addressed in the analyses as described below.

We compared postbaccalaureate program participants and controls on baseline characteristics and on the proportion of students in each group who had matriculated into an AMCAS-participating US medical school. We then divided the pooled sample (ie, program participants and controls) into groups depending on whether they had matriculated into medical school and compared the characteristics of the matriculants and nonmatriculants. For these bivariate analyses, tests for the significance of differences across groups were performed using 2-tailed t tests of differences in means and χ² tests of differences in proportions, with significance set at P<.05.

We developed logistic regression models to calculate odds ratios (ORs) of medical school matriculation for participants compared with controls. The first set of regression models was performed on the entire sample of students included in the study. The unadjusted model included program participation as the only predictor variable. In addition, a multivariate-adjusted model was estimated, which included covariates to control for the potential confounding effects of baseline characteristics. Inclusion of these covariates was based on a priori hypotheses of their potential association with medical school matriculation, even if the variables had no statistically significant association with matriculation in the bivariate analyses. The multivariate regression included variables for program participation, GPA, first-time medical school application vs reapplication, college major (science or mathematics vs all other majors), year of postbaccalaureate program participation or application, college attended (a UC campus vs all other colleges), highest level of education of either of the student's parents, and the student's race/ethnicity and sex.

A second set of regression analyses excluded students who did not have baseline data on MCAT scores. An unadjusted model and adjusted model were estimated, with the adjusted model using the same covariates as above but with the addition of MCAT scores.

In the multivariate models, missing data on race/ethnicity and parental education were handled by creating a dummy variable for the missing value. This could not be done for missing data on age because it was used as a continuous variable; we therefore omitted age as a covariate in the adjusted models. However, to determine whether the results for program participation were highly sensitive to age, we repeated the multivariate models using age as a covariate, excluding students for whom age data were missing.

The study protocol was approved by the University of California, San Francisco, institutional review board, which did not require informed consent. All statistical analyses were performed using Stata version 8.0 (StataCorp, College Station, Tex).

Baseline characteristics of participants and controls are shown in Table 1. Although participants had a significantly higher mean GPA than controls, the absolute difference was small. The mean GPA among participants was 3.11, with many students having a GPA less than 3.0. Mean cumulative preprogram MCAT scores among participants and controls were very similar. The postbaccalaureate participants were more likely to be reapplying to medical school and to have been science or mathematics majors in college. Participants were more likely than controls to have low family socioeconomic status and to be members of underrepresented minorities. For almost half of the participants, the highest level of parental education was high school or lower. Nearly two thirds of participants reported their race/ethnicity as black or Chicano/Latino.

By 2005, 3 times as many program participants as controls had matriculated into medical school (67.6% vs 22.5%; P<.001) (Table 1). When matriculation rates were examined separately for each of the 5 UC postbaccalaurate programs, the results were relatively consistent, with matriculation rates ranging from 60% to 78% across the 5 programs. Of the participants, an additional 26 (9.8%) matriculated into osteopathic medical schools; comparable data were not available for the controls.

The characteristics of students in the study sample matriculating into medical school and those not matriculating are shown in Table 2. More than 66% of all successful matriculating students had participated in a UC postbaccalaureate program. The mean GPA and cumulative MCAT score of matriculating students were significantly higher than those for nonmatriculating students. Matriculating students were younger and more likely to be reapplicants, science or mathematics majors, black or Chicano/Latino, and have parents with higher educational attainment.

The results of the regression models are shown in Table 3. In the models that included students with missing MCAT scores, the unadjusted OR of matriculation for participants relative to controls was 7.18 (95% confidence interval [CI], 5.06-10.18) (Table 3). After controlling for potential confounding variables in the multivariate model, the adjusted OR for matriculation into medical school among program participants compared with nonparticipants remained large (OR, 6.30; 95% CI, 4.08-9.72). In the models that excluded students without baseline MCAT data, the unadjusted OR for matriculation into medical school among program participants compared with nonparticipants was 6.27 (95% CI, 4.29-9.15) and the adjusted OR was 8.06 (95% CI, 4.65-13.97). The ORs did not change significantly in sensitivity analyses that excluded students with missing values for age and added a covariate for age, or that restricted the analysis to only medical school reapplicants.

GPAs and MCAT scores were strong independent predictors of medical school matriculation in the multivariate regression models (Table 3). Black and Chicano/Latino race/ethnicity was also a significant independent predictor in both regression models. This study was not powered to detect statistically significant interactions between program effects and students' race/ethnicity.

Of the 265 participants, 218 (82%) are known to have applied to medical school or were still in the application process as of 2005. The percentage of successful applications for participants was 82% (179/218). Matriculation rates ranged from 78% for the 1999-2000 cohort to 64% for the 2002-2003 cohort, with rates of application to medical school following a similar time trend across cohort years. Many students appear to delay application to medical school to the application cycle in the year following completion of the postbaccalaureate program. Of the 179 participating students who matriculated into medical school, 51 (28%) matriculated at the same UC campus where they attended the postbaccalaureate program. The Los Angeles program was the only one at which the majority of the participating students matriculated at the same campus. Of the postbaccalaureate program participants who matriculated into medical school in 2000 and 2001, 91% had graduated by 2005.

Only aggregate data were available on the proportion of controls who had submitted medical school applications through AMCAS. Of these, 65% (257/396) had applied by academic year 2005. The percentage of successful applications for controls was 35% (89/257) compared with 82% for participants (P<.001).

To our knowledge, this is the first controlled outcomes study focusing exclusively on postbaccalaureate premedical programs, adjusting for potential underlying differences between program participants and nonparticipating controls. Participation in UC postbaccalaureate programs is associated with an increased likelihood of matriculating into medical school. Cohorts of students attending UC postbaccalaureate programs in the study years had significantly higher odds of matriculating into medical school than did controls, even after adjusting for differences in underlying factors, such as college GPA, that are strong predictors of successful admission.

The University of California postbaccalaureate premedical programs target disadvantaged students. Nearly 50% of participating students came from homes in which the highest level of parental education was no greater than high school, and 66% of participants self-identified as black or Latino. Moreover, the mean baseline MCAT score (21.9) and GPA (3.11) of the postbaccalaureate students reflect an academic disadvantage for successful application to medical school. Based on national AMCAS data, the mean MCAT score of all US medical school matriculants in 2005 was 30.2, and the mean GPA was 3.63.¹⁷

Our findings suggest that the UC postbaccalaureate programs studied are effective in reaching students from groups traditionally underrepresented in medical school and in enhancing the ability of these students to successfully apply to medical school. Consistent with other studies,¹⁶ almost all postbaccalaureate program participants in our study who matriculated into medical school graduated from medical school on time. The cost per student for the 1-year enrollment in the postbaccalureate program at the San Francisco campus has been estimated to be $14 500, including staff salaries, supplies, student stipends, and summer housing allowance (Valerie Margol, MA, School of Medicine, University of California, San Francisco, written communication, April 2005). These costs do not include student expenses of $3500 for 2 semesters of tuition and fees to enroll in undergraduate courses at an affiliated state university campus and in-kind institutional contributions such as guest faculty time and classroom space.

Our study has several limitations. The study design did not permit us to detect whether some students in the control group actually attended postbaccalaureate programs that were not sponsored by the University of California. If this occurred, some of the controls would have been incorrectly classified. If non-UC programs also increase the probability of medical school matriculation, then our findings would underestimate the effect of the UC postbaccalaureate programs. The study also likely underestimates the ultimate “yield” of the postbaccalaureate programs; based on the time trends across cohorts, longer follow-up of the most recent participant cohorts would be expected to find a higher cumulative matriculation rate.

We also used a narrow definition of a “successful” outcome, classifying success as matriculation at a US allopathic medical school. The lack of a centralized data repository for osteopathic schools comparable with the AAMC repository limits our ability to include matriculation at an osteopathic school as a successful outcome. Tracking data performed by the programs on participants show that 26 (9.8%) of the UC postbaccalaureate participants matriculated into osteopathic medical schools. However, no similar data are available on the controls. No information on matriculation into non-US medical schools is available for either participants or controls. It is not clear whether this limitation introduces a bias into our results, because we do not know whether nonparticipants would be more likely than participants to matriculate into an osteopathic school or non-US medical school.

We did not measure whether students successfully pursued health professional fields other than medicine. Entering a training program in dentistry, public health, or other field certainly constitutes success for these students, but we could not systematically measure matriculation into these schools. In addition, our study only investigated UC programs and therefore may not be generalizable to similar programs at other institutions. We could not determine which specific components of the postbaccalaureate program might be most responsible for the outcomes observed.

Finally, even rigorous controlled observational studies may be susceptible to unmeasured confounding that may bias findings. We attempted to minimize this effect by measuring and adjusting for key confounders, such as GPA and MCAT score. However, it is plausible that participants and controls differed in personal characteristics such as motivation that we did not measure and that may also be associated with the likelihood of matriculating into medical school. To fully eliminate unmeasured confounding would require a randomized controlled trial. Due to ethical and other considerations, it is unlikely that a postbaccalaureate program for minority and disadvantaged students will be subjected to a randomized experiment. We believe that our study is as scientifically rigorous as is feasible, short of a randomized trial.

In conclusion, our study offers strong evidence that participation in postbaccalaureate premedical programs for minority and disadvantaged students is an effective intervention to increase the number of such students entering medical school. Among the continuum of educational pipeline programs, postbaccalaureate interventions are relatively high yield because they require only a single year of intervention, target students who have an explicit commitment to a career in medicine, and have a short timeline for achieving their payoff. The continued support and expansion of postbaccalaureate premedical programs is an important strategy for increasing the diversity of the physician workforce. Reductions enacted in the 2006 fiscal year in federal funding for Health Career Opportunities Programs, Health Careers Centers of Excellence, and related pipeline programs may threaten the continued existence of many postbaccalaureate programs that have traditionally received support from these federal programs.