Assessment of the Medical Schools From Which Radiation Oncology Residents Graduate and Implications for Diversifying the Workforce

Abstract

Purpose:

To identify factors predictive of a medical school graduating a high volume of future radiation oncology (RO) residents to better understand potential pathways to effectively recruit women and underrepresented minority (URM) students into RO.

Methods and Materials:

Demographics for US allopathic medical schools and affiliated RO departments were collected from web resources and correlated with the percentage of graduates from each school currently enrolled in RO residency in 2019, and the probability of at least 1 female or URM student from each school pursuing RO.

Results:

The median percentage of students per medical school who pursued RO residency was 0.37% (interquartile range, 0.16%-0.66%). A total of 79.7% of schools graduated at least 1 RO resident, whereas 51.7% graduated at least 1 female RO resident and 14.0% graduated at least 1 URM RO resident. The 30 schools graduating the highest percentage of RO residents accounted for 52.1% of current RO residents, only 4 of which were in the top quartile for URM enrollment. Medical students were significantly more likely to pursue RO when there was an affiliated RO department (0.42% vs 0.18%, P < .001) or RO residency program (0.51% vs 0.18%, P < .001), more total RO faculty (r_s = 0.521, P < .001), female RO faculty (r_s = 0.481, P < .001), and URM RO faculty (r_s = 0.197, P < .001). The probability of at least 1 female student pursuing RO was also correlated with the number of female faculty in the affiliated RO department (r_pb = 0.348, P = .001), and a similar correlation was observed between URM students and URM faculty (r_pb = 0.312, P = .011).

Conclusions:

Most RO residents graduate from medical schools with larger affiliated RO departments but fewer URM students. To promote greater RO diversity, outreach should be considered among schools with greater URM enrollment but fewer affiliated radiation oncologists, and among female and URM students in schools that graduate many RO residents.

Introduction

The radiation oncology (RO) workforce has significant underrepresentation of women, African Americans, and Latinos compared with both the US population and medical school graduates.¹ For instance, women are 48% of medical school graduates but only 27% of the RO workforce; underrepresented minorities (URM) compose 15% of all US residents, but only 9% of RO residents.¹ There are likely a variety of reasons for this, but for many, ultimate pursuit of an RO residency position begins with initial exposure to the specialty during medical school, after which interest can be cultivated through effective mentorship, along with clinical and research opportunities.^2,3 Unfortunately, at many medical schools the probability of any student being exposed to RO is relatively low,^4–7 particularly at schools without or with smaller affiliated RO departments.¹ Furthermore, there is little data on which medical schools RO residents commonly graduate from, let alone for the female and URM subgroups of students. The objective of this study was to identify factors predictive of a medical school graduating a high volume of future RO residents to better understand potential pathways to more effectively diversify the RO physician workforce.

Methods and Materials

Primary data were obtained through publicly available online resources. The American Society for Radiation Oncology (ASTRO) directory was used to search all RO physicians who self-identified as a “resident” in 2019 and to listed the medical school from which they graduated.⁸ This included 498 residents, of whom 36 who attended osteopathic or international medical schools were excluded.

At the time of data collection, it was estimated that 725 RO residency positions were offered,⁹ indicating that this study cohort represents 64% of all residents. Because self-reported race, ethnicity, and gender are not publicly available or shared by ASTRO, the Accreditation Council for Graduate Medical Education, or other such sources, the gender and race/ethnicity of each resident was assigned “socially” based on their name and photograph in the ASTRO directory and, as necessary, confirmed using individual residency program websites or Google images searches. In most cases these assignments appeared relatively straightforward and were handled by a single investigator; a second investigator was involved in the event of ambiguity. Socially assigned race has been associated with a 3.4% rate of discordantly attributing nonminority status to an individual who self-identifies as a minority and a 1.5% rate of discordantly attributing minority status to an individual who self-identifies as a nonminority, with the majority of discrepancies involving Latinos.¹⁰ Fewer data are available on the accuracy of socially assigned gender, although previous studies have based gender assignments on an individual’s name.¹¹ The proportion of transgender adults in the United States has been reported to be 0.6%, and subjectively the authors believed there was less ambiguity in their assignment of gender than there was for race.

The demographic composition of all 143 allopathic US medical schools (excluding US territories) in 2018 was collected from the Association of American Medical Colleges website.¹² This included percentages of female, African American, and Latino students at each school by total medical school enrollment. Other URM groups (eg, American Indians Alaska Natives, Native Hawaiians, and Pacific Islanders) were not specifically analyzed owing to their smaller numbers and the greater difficulty of socially assigning race.

The average Medical College Admission Test (MCAT) score and grade point average (GPA) of matriculating students was obtained from each medical school admissions website, if available. MCAT scores were converted to percentile for each school for analysis, using the Association of American Medical Colleges conversion table.¹³ Additional web searches (and telephone calls to clerical staff in medical schools, cancer centers, and medical or RO departments if needed) were used to determine which medical schools had an affiliated RO department, which had an RO residency program, the total number of on-site physician RO faculty in each department, and the number of female, African American, and Latino RO faculty members in each department. Gender and race/ethnicity of RO faculty was socially assigned using the same method as for residents.

Additional medical school demographic features evaluated included whether the school was public or private, whether the school was affiliated with a National Cancer Institute (NCI) designated cancer center, and the school’s total National Institute of Health research funding.

Statistical analysis

Descriptive statistics (median and interquartile range [IQR]) were used to report demographic features of medical schools and RO departments. These represented the independent (predictive) variables for the study. Associations between the percentage of total students enrolled at each medical school who pursued an RO residency and continuous independent variables were assessed using Spearman rank-order correlation and linear regression analysis. Associations with binary independent variables were assessed using the Mann-Whitney test. Statistically significant variables from these univariate analyses were included in a multiple logistic regression model, using a stepwise variable selection procedure, given the multi-collinearity between several independent variables.

Predictors of at least 1 female student or at least one URM student pursuing an RO residency from each school were also assessed. These dependent variables were primarily analyzed as binary categorical variables due to the relatively small numbers of female and URM RO residents. Associations between these outcomes and continuous independent variables were assessed using point-biserial correlations and logistic models, whereas associations with binary independent variables were assessed using the χ² test. Multiple logistic regression is not reported for these subgroups due to the relatively small numbers of female and URM RO residents. This study was approved by the local institutional review board.

Results

There were 143 allopathic US medical schools assessed, with a median total enrollment per school of 627 students (IQR, 450-787 students). The median percentage of female, African American, and Latino student enrollment per medical school was 51.2% (IQR, 48.2%-55.4%), 5.7% (IQR, 3.7%-8.1%), and 4.1% (IQR, 2.4%-6.4%), respectively. A total of 113 medical schools (79%) had an on-site affiliated RO department, and 83 medical schools (58%) had an affiliated RO residency program. These RO departments comprised 1213 RO faculty members. There was a median of 8.5 RO faculty members (IQR, 4-15) per RO department, with a median of 2 female faculty (IQR, 1-4) and 0 African American or Latino faculty per department (IQR, 0-1). In total, 330 (29.0%) US RO academic faculty were women and 53 (4.8%) were African American or Latino. Departments with a larger overall number of faculty were also more likely to have a larger number of female faculty (r_s = 0.842, P < .001) and URM faculty (r_s = 0.368, P < .001). However, there was no association between the percentage of female or URM faculty in a department and the size of the department (P = .280 and .494, respectively).

The median number and percentage of students per medical school who pursued an RO residency was 3 (IQR, 1-5) and 0.37% (IQR, 0.16%-0.66%), respectively. The median percentage of all RO residents who graduated from each medical school was 0.65% (IQR, 0.22%-1.08%). RO residents comprised 141 (30.5%) females and 31 (6.8%) African Americans or Latinos. Over the 4 years of residents evaluated, a total of 114 medical schools (79.7%) graduated at least 1 RO resident, 74 medical schools (51.7%) graduated at least 1 female RO resident, and 20 medical schools (14.0%) graduated at least 1 URM RO resident. Only 32 RO residents (6.8%) graduated from the 30 schools without an affiliated RO department. Conversely, the 30 schools that graduated the highest percentage of RO residents accounted for 52.1% of all RO residents. Table E1 shows the 30 medical schools that graduated the highest percentage of RO residents, female RO residents, and URM RO residents.

Tables 1 (for continuous variables) and 2 (for categorical variables) correlate medical school and RO department factors with the percentage of students pursuing RO from a given medical school and the probability of at least 1 female or URM student pursuing an RO residency from a given medical school. The percentage of students pursuing RO residency from a given school was correlated with average matriculating student MCAT and GPA, having an affiliated RO department and residency program, the total number of RO faculty, the total number of female and URM faculty, total medical school National Institute of Health research funding, private (vs public) medical school, and affiliation with an NCI-designated comprehensive cancer center. The final model on multiple logistic regression analysis included 3 independent variables, including the average matriculating student MCAT score (0.008 ± 0.003, P = .014) and GPA (0.940 ± 0.431, P = .031) and total number of faculty in an affiliated RO department (0.023 ± 0.004, P < .001).

Table 1.

Correlations and regression analyses between continuous medical school (n = 143) and radiation oncology department (n = 113) variables and medical students pursuing radiation oncology residency (n = 471)

	Percentage of all students pursuing RO residency from each school			At least one female student pursuing RO residency from each school			At least one URM student pursuing RO residency from each school
	Correlation coefficient, rs	Effect ± standard error	P value	Correlation coefficient, rpb	Effect ± standard error	P value	Correlation coefficient, rpb	Effect ± standard error	P value
Medical school’s percentage of URM students	0.121	−0.003 ± 0.004	.435	−0.115	−0.023 ± 0.018	.190	0.046	0.011 ± 0.020	.584
Medical school’s percentage of female students	0.047	0.006 ± 0.004	.123	−0.068	−0.014 ± 0.017	.422	−0.013	−0.004 ± 0.023	.875
Average medical school’s matriculating student MCAT percentile	0.508	0.021 ± 0.003	< .001 *	0.401	0.086 ± 0.020	< .001 *	0.161	0.048 ± 0.026	.060
Average medical school’s matriculating student GPA	0.447	2.533 ± 0.425	< .001 *	0.315	8.802 ± 2.504	< .001 *	0.145	6.013 ± 3.513	.087
Medical school’s total NIH research funding ($)	0.497	0.002 ± 0.0003	< .001 *	0.336	0.007 ± 0.002	.001 *	0.205	0.004 ± 0.002	.030 *
Percentage of medical school’s graduates that pursued RO	—	—	—	0.551	4.886 ± 0.897	< .001 *	0.413	0.620 ± 0.020	< .001 *
Total number of faculty in RO department	0.521	0.033 ± 0.004	< .001 *	0.400	0.123 ± 0.029	.001 *	0.331	0.093 ± 0.026	< .001 *
Number of URM faculty in RO department	0.197	0.105 ± 0.048	.030 *	0.156	0.431 ± 0.271	.112	0.312	0.732 ± 0.288	.011 *
Percentage of URM faculty in RO department	0.115	−0.001 ± 0.004	.796	−0.068	−0.011 ± 0.016	.482	0.062	−0.003 ± 0.003	.379
Number of female faculty in RO department	0.481	0.090 ± 0.013	.001 *	0.348	0.317 ± 0.094	.001 *	0.294	0.223 ± 0.078	.004 *
Percentage of female faculty in RO department	0.236	0.359 ± 0.204	.081	0.081	0.008 ± 0.009	.395	0.007	0.001 ± 0.012	.940

Abbreviations: GPA = grade point average; MCAT = Medical College Admissions Test; NCI = National Cancer Institute; NIH = National Institutes of Health; RO = radiation oncology; URM = underrepresented minority.

^*Statistically significant.

Table 2.

Associations between categorical medical school (n = 143) and radiation oncology department (n = 113) variables and medical students pursuing radiation oncology residency (n = 471)

	Percentage of all students pursuing RO residency from each school			At least one female student pursuing RO residency from each school			At least one URM student pursuing RO residency from each school
	Median (IQR)		P value	Percentage	Odds ratio	P value	Percentage	Odds ratio	P value
Affiliated RO department
No (n = 30)	0.18%	(0.00%-0.40%)	< .001 *	30.0%	3.16	.007 *	6.7%	2.63	.194
Yes (n = 113)	0.42%	(0.21%-0.74%)		57.5%			15.9%
Affiliated RO residency
No (n = 60)	0.18%	(0.00%-0.37%)	< .001 *	28.3%	5.56	< .001 *	5.0%	4.90	.008 *
Yes (n = 83)	0.51%	(0.35%-0.93%)		68.7%			20.5%
Type of medical school
Public (n = 51)	0.36%	(0.13%-0.36%)	.006 *	46.2%	1.90	.072	10.8%	2.06	.128
Private (n = 92)	0.47%	(0.21%-0.89%)		62.0%			20.0%
Affiliated NCI Cancer Center
No (n = 84)	0.30%	(0.00%-0.45%)	< .001 *	34.5%	6.11	< .001 *	8.3%	3.12	.020 *
Yes (n = 59)	0.55%	(0.34%-0.96%)		76.3%			22.0%

Abbreviations: NCI = National Cancer Institute; RO = radiation oncology; URM = underrepresented minority.

^*Statistically significant.

The probability of at least 1 female student pursuing RO from a given school was most closely correlated with the percentage of all students pursuing RO from that school, as well as essentially the same other factors relevant to all students. The top quartile of RO departments with the most female faculty were affiliated with schools that graduated 49% of all female RO residents. The probability of at least 1 URM student pursuing RO from a given school was most closely correlated with the percentage of all students pursuing RO from that school, as well as the number of faculty in the RO department, the number of URM and female faculty, and affiliation with an RO residency program and NCI-designated comprehensive cancer center, although there was less of a correlation with the average MCAT or GPA of matriculating students. Table E2 shows that when evaluating female and URM student recruitment as a continuous variable (percentage who pursued RO) there was no difference in which variables were shown to be predictive of pursuit of RO residency, with minor variations in the strength of the correlations. The percentage of URM students in a medical school did not correlate with recruitment of URM students into RO. However, the 12 schools with greater than 20% enrollment of URM students only graduated 3.7% of RO residents, and the 40 schools with the highest URM student enrollment included only 5 schools that were also in the top quartile for graduating RO residents.

Discussion

This study has shown that the majority of medical students who pursue RO graduate from a minority of medical schools, the majority of which do not have high enrollment of minority students. The most strongly associated and independent predictors for a medical school graduating a higher volume of future RO residents included having a larger number of faculty in an affiliated RO department and a higher average MCAT score and GPA among matriculating students. However, other factors related to the prestige/competitiveness of the school and students (eg, attendance at a private medical school, greater institutional research funding, affiliation with an NCI-designated comprehensive cancer center) and a higher probability of exposure to RO and effective mentorship (eg, affiliation with an on-site RO department and residency program, with larger numbers of female and URM faculty) were also relevant. Although the prestige-related factors may be less amenable to change through the actions of any individual radiation oncologist or department, given the relative lack of progress in increasing diversity in RO over the past 2 decades,^2,3 our findings suggest that it may be necessary to develop better approaches to stimulate awareness of RO among a broader population of students, which may begin with the way we initially expose students to the specialty and cultivate their interest through effective mentorship.

An association between exposure to, and ultimate pursuit of, several medical specialties has been demonstrated in prior studies. For instance, among all female graduate medical education trainees, any medical specialty that was not part of the third-year core clerkships or had a lower percentage of female faculty members was less likely to attract female medical students.¹⁴ Similarly, pipeline programs geared toward lower income or URM high school and college students have been shown to increase interest in the health professions.^15,16 Among medical students, a strategic pipeline program specifically designed to recruit women and URMs to procedure-based specialties was found to have a 75% retention rate.¹⁷ Currently, URM medical students most often choose primary care specialties, such as family medicine, obstetrics/gynecology, pediatrics, and psychiatry. Regardless of whether these specialties are chosen based on career goals or other factors, the aforementioned studies and our findings support future research evaluating the direct effect of exposure-driven interventions geared toward diversifying the RO workforce.

Interestingly, the absolute number of female and URM faculty in an RO department was moderately correlated not only with larger numbers of female and URM students pursuing RO, but also all students pursuing RO from a given medical school. This finding could simply be related to the fact that more female and URM faculty tend to work in larger RO departments; however, it may also stem from the nature of mentoring relationships. A more diverse pool of students may become more interested in RO if they are able to connect with a critical mass of mentors and role models from a similar background, which could explain our findings that approximately one-half of all female residents came from schools with RO departments in the top quartile for female faculty and that there was a closer correlation between a large number of URM faculty and URM students pursuing RO than there was between the number of URM faculty and all students, or female students, pursuing RO. Furthermore, several studies have demonstrated that racial/ethnic or gender concordance of the student—mentor relationship leads to greater productivity,^14,18 something necessary (particularly in terms of research productivity) to develop a strong RO residency candidate. Surgical training programs that promote gender and racial diversity have also been shown to be more successful in recruiting female and URM students into residency positions.¹⁹

The fact that the percentage of female and URM in an RO department (which more closely defines diversity than the absolute numbers) did not affect the outcomes of this study may be related to the fact that not all academic faculty, regardless of gender or ethnicity, choose to devote appreciable time to mentoring medical students. As such, a given RO department that has a larger number of faculty may be more likely to have stronger student mentors to facilitate more effective recruitment than a department with a smaller number of faculty, even if the latter happens to be more diverse in some cases. For all of these reasons, promoting more female and URM academic faculty members, especially those with an interest in medical student education, may also help diversify the RO workforce.²⁰

A key difference in the demographics of female and URM students is that female students compose approximately one-half of nearly all US medical school classes, but the proportion of URM students varies considerably between schools. As such, optimizing recruitment of female students is likely to be most readily achieved at institutions that already recruit many students into RO; recruitment of more URM students into RO may also be enhanced by targeted education and mentoring outreach initiatives for schools with fewer affiliated RO faculty and larger URM student enrollment. Even if some of these schools may enroll students using a more holistic approach with less emphasis on examination scores, a large component of producing a competitive RO residency applicant comes from letters of recommendation, commitment to the specialty, and research experiences,²¹ all of which can be achieved through effective mentorship that begins as early during medical school as possible. This type of endeavor could be facilitated by national organizations like ASTRO, or even some larger or more diverse individual institutions.

For many radiation oncologists, finding ways to integrate a small, highly specialized field like RO into a medical school curriculum may be challenging, especially considering how little RO is tested on United States Medical Licensing examinations, but there are a variety of ways in which faculty at some institutions successfully approach it.^22–25 How to bring RO exposure and mentorship to schools with small or nonexistent affiliated RO departments is an even greater challenge; however, the recent change of United States Medical Licensing examinations Step 1 to pass/fail grading represents an opportunity for radiation oncologists to push curriculum committees for dedicated time in the preclinical curriculum.

Finally, given that nearly three-fourths of radiation oncologists ultimately practice in the region of their medical school or residency,²⁶ targeting students at medical schools without an affiliated RO department or residency program but with greater contact with underserved patients may help improve patient access to RO and reduce health care disparities. Outreach to students in geographic regions with fewer radiation oncologists is also likely to mitigate the perception of a challenging RO job market in the US regions where there is currently overselection of students from certain medical schools, irrespective of student gender and race/ethnicity.

The major limitation of our study was reliance upon the ASTRO directory to collect information about RO residents, some of whom were either unlisted or did not report information about the medical school they attended. Nevertheless, approximately two-thirds of residents over 4 years of training were evaluable. However, this limitation on the power of the study would perhaps have had its greatest impact on analysis of the URM residents because there are so few of them in RO. We also could only collect information on RO faculty members at the present time from department webpages, which would not perfectly reflect faculty composition at the time that each of the residents in the ASTRO directory were medical students. Finally, the social assignment of gender and race/ethnicity for faculty and residents manually based on photographs and names may differ from self-reporting, although ultimately the percentage of female and URM residents determined using this method was similar to compiled numbers from the national graduate medical education census in which residents self-identified their race.²⁷ We expect that attributing nonminority status to an individual who self-identifies as a minority was more likely using our method given the greater ambiguity that may arise from individuals of mixed race/ethnicity, which may have also limited the power of our URM analysis.

Conclusions

Most RO residents graduate from medical schools with larger RO departments but fewer URM students. To promote greater diversity in the RO workforce, education and mentoring outreach initiatives should be considered among schools with greater URM enrollment but fewer affiliated radiation oncologists and among female and URM students in schools that graduate many RO residents and already have strong recruitment approaches in place.