Racial, Ethnic, and Gender Representation in Leadership Positions at National Cancer Institute–Designated Cancer Centers

Key Points Question What is the diversity of the leadership teams of National Cancer Institute–designated cancer centers, and how does this compare with the populations served by each center? Findings In this retrospective cross-sectional study including 63 cancer centers with 856 leadership team members, non-Hispanic White men were disproportionately represented in leadership while Black, Hispanic, and Asian leaders were underrepresented. Centers with more women leaders were more likely to have at least 1 Black or Hispanic leader; however, diverse cities were not necessarily more likely to have representatively diverse leaders. Meaning These findings suggest that establishing policy and pipeline programs to address significant racial and ethnic disparities in cancer care leadership positions is crucial for change.


Introduction
Racial disparities in cancer care access, delivery, and outcomes are well documented. 1 They are, in part, due to lack of access to high-quality, culturally competent care. 2 An increasing body of research indicates that increasing diversity among health care leaders and physicians is beneficial for the health care system and patient outcomes. 3,4 One 2017 study 5 found that older patients treated by women internists had lower mortality and readmissions compared with those treated by men internists, and a 2020 study 6 found that newborn mortality in Black infants was halved when they were cared for by a Black physician. When patients are treated by racially and ethnically concordant physicians, they are more likely to receive necessary medical care, including preventative health care. 7 Specifically within cancer care, physician diversity is essential in the provision of high quality cancer treatment to increasing racial/ethnic minority communities, such as Hispanic communities, who have made up more than half of the population growth in the US. 8 Implicit bias has known negative outcomes within oncology, 9 and improving diversity can lead to increased intercultural responsiveness and foster trust and comfort for patients. 10 Developing an oncology workforce that reflects the patients whom it serves has been a priority for both American Society of Clinical Oncology 11 and the National Cancer Institute (NCI), 12 as there are known gender and racial/ethnic gaps within the physician pipeline and workforce. 13,14 However, the leadership gap in cancer care remains largely unquantified and is a key component to understanding how institutions may prioritize equity, diversify hiring, and promote systemwide change to improve cancer disparities. This cross-sectional study was designed to evaluate the gender, racial, and ethnic makeup for the full leadership team of NCI-designated cancer centers and to compare this with the diversity of actively practicing physicians and with the city populations served by each center.

Methods
This cross-sectional study was reviewed by the institutional review board of Memorial Sloan Kettering and was found to not meet the definition of human participants research; therefore, this study did not require oversight or informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for reporting observational cross-sectional studies.
The names, photographs, degrees, academic titles, and h-index scores of leadership team members were obtained via publicly available information for each of the 63 NCI-designed cancer treatment centers. Leadership team members are defined as individuals who were identified on each cancer center's respective website as "leaders," part of the "leadership team," or listed under the center "leadership" page. The h-index is a measure of a scholar's productivity and publication impact; it is calculated as the highest number of highest cited papers. It was captured as a measure of leadership academic productivity.
Gender, race, and ethnicity were determined by facial recognition software (Kairos) with secondary manual review and additional classification from first and last name and biography for ambiguous identities. Manual review was performed by A.M. and F.C. Both authors individually reviewed each photograph and name and compared that to the facial recognition algorithm output. Any disputes or ambiguous photographs required further inspection of the individual's affiliations;

JAMA Network Open | Health Policy
unresolved disputes were marked as unknown race or ethnicity. The facial recognition algorithm used was developed from a large diverse database; the algorithm was specifically designed to reduce known bias apparent within other software. 15 City population demographic information was collected from the US Census, 16 and centers were grouped geographically (Northeast, South, West, Midwest). Cancer center rank was identified from US News and World Report 2020 Rankings. 17 Information on actively practicing physicians in the United States was gathered from the 2019 Association of American Medical Colleges workforce data. 18

Statistical Analysis
Pearson correlation and multivariate analysis were used to assess the association between racial/ ethnic representation on leadership teams and location of institution, institution ranking, median h-index of researchers on leadership team, and composition of leadership team with regards to sex and degrees earned using RStudio statistical software version 1.2.5033 (R Project for Statistical Computing). Data were analyzed in August 2020. All P values were from 2-sided tests, and the results were deemed statistically significant at P < .05.

Results
All 63 NCI cancer centers had identifiable leadership teams, with a total of 856 leadership members.
Leadership teams ranged from 1 to 42 members. We were unable to obtain photographs for 12 leaders (1.4%); of 844 remaining leaders, facial recognition software and manual review were unable to identify race/ethnicity of 7 leaders (0.8%). Of 844 leaders with photographs, we were able to identify gender for all of them, and 306 (36.3%) were women. Of 837 leaders with photographs who could be assigned race/ethnicity, 688 (82.2%) were non-Hispanic White individuals, 29 (3.5%) were Black individuals, 92 (11.0%) were Asian individuals, and 32 (3.8%) were Hispanic individuals.
Comparisons between the gender, racial, and ethnic make-up of the entire US population, active physicians, and physicians in cancer center leadership roles are shown in   correlation between city Hispanic population and representation on leadership teams (R = 0.5, P < .001) but no association between Black population and leadership (Figure 2).

Discussion
This cross-sectional study found that White men were disproportionately represented in the highest levels of cancer center leadership while Black, Hispanic, and Asian leaders did not have proportionate space at the decision-making table. Centers with more women leaders were more likely to have diverse leadership teams; however diverse cities were not necessarily more likely to have representatively diverse leaders, with Black leaders rare even in cities with large Black populations.
These findings suggest that marginalized racial/ethnic groups, such as Black, Hispanic, and Asian individuals, may face stark challenges limiting their advancement into senior leadership positions.
Our findings are similar to a recent director survey from the Association of American Cancer Institute and The Cancer Letter, 19 which found a high percentage of White men in director roles.
NCI-designated cancer centers are the anchors of the nation's cancer research effort, and center leaders are actively involved in setting the standards of care and the future direction of cancer treatment in the US, including cancer clinical trials. Thus, lack of diversity within these leadership teams is particularly concerning, as they reflect the immediate pipeline of directors, and that

Limitations
This study has some limitations, including the categorization of leadership per the US Census, which includes those with Middle Eastern and North African ethnicities as White race. This places some leaders who may have dealt with significant racial/ethnic or cultural bias in a category that may not reflect their contributions to diversity. Additionally, binary male/female gender does not appropriately categorize leaders who may identify as nonbinary or gender nonconforming. Another limitation lies with accurately identifying race and ethnicity from a photograph, name, and biography, as self-reported race/ethnicity continues to be the criterion standard for this information. Facial recognition software has known racial bias and can be ethically dubious in use. 23 We specifically sought out a software that was created to combat known bias concerns and was developed from a large, international database. 15 Also the system of facial recognition combined with manual review had a 96.2% concordance to self-report based on prior work by the study team. 24 Population demographic information was limited to the cities where each cancer center was located; this may not reflect the true demographic characteristics of patients served over the center's entire