Duke University School of Medicine Letter

September 8, 2026

VIA E-mail to: mary.klotman@duke.edu

Mary E. Klotman, MD
Executive Vice President for Health Affairs, Duke University
Dean, Duke University School of Medicine
Chief Academic Officer, Duke Health
DUMC 2927
40 Duke Medicine Circle
124 Davison Building
Durham, NC 27710

Re: OCR Transaction No.: 25-623495

Dear Dr. Klotman:

On July 28, 2025, the U.S. Department of Health and Human Services (“HHS”), Office for Civil Rights (“OCR”), initiated an investigation in response to a complaint alleging that Duke University School of Medicine (“Duke SOM”) and the Duke University Health System (“DUHS”) (collectively “Duke”) may have violated Title VI of the Civil Rights Act of 1964 (“Title VI”) and Section 1557 of the Patient Protection and Affordable Care Act (“Section 1557”).¹ The complaint asserted, among other allegations, that Duke SOM’s admission practices did not comply with Title VI as interpreted by the Supreme Court’s decision in Students for Fair Admissions, Inc. v. President & Fellows of Harvard College, 600 U.S. 181 (2023) (“SFFA”).²  Based on Duke SOM’s admitted pre-SFFA race-conscious admissions practice, contemporaneous post-SFFA admissions communications, admissions materials preserving race-revealing channels, and applicant-level admissions data, OCR finds that Duke SOM used, and continues to use, race and race-revealing proxies in medical school admissions decisions after SFFA in violation of Title VI.

1. Legal Standard

Title VI and Section 1557 prohibit recipients of federal funds from discriminating based on race, color, or national origin.³ In SFFA, the Supreme Court held that racially discriminatory practices in higher education can rarely be squared with civil rights laws, no matter what a university’s justification for the practices may be.⁴  Applying strict scrutiny, the Court held that universities “may never use race as a stereotype….”⁵ The Court rejected the notion that considering race in admissions could survive strict scrutiny because of an interest in “producing new knowledge stemming from diverse outlooks.”⁶ The Court made clear that proxies for race also may not be used to discriminate, noting that “universities may not simply establish through application essays or other means the regime we hold unlawful today.”⁷ OCR considers applicant-level and aggregate admissions data, along with other information, as evidence bearing on intent, causation, and pretext, consistent with HHS’s determination under Title VI that disparate outcome data may help establish and prove intentional discrimination.

Title VI authorizes HHS, through OCR, to conduct periodic compliance reviews and investigations of the recipients of federal funding.⁸  Enforcement under Title VI requires funding agencies to advise recipients of their failure to comply and to determine whether compliance can be obtained by voluntary means.⁹  If OCR determines that the noncompliance with Title VI cannot be corrected by voluntary means, OCR may seek to compel compliance through an enforcement action.¹⁰ OCR also applies Section 1557 to the extent that Duke SOM’s admissions process is part of a covered health program or activity receiving HHS federal financial assistance.

2. Procedural Background

OCR confirmed that Duke receives federal financial assistance from HHS. Specifically, during fiscal years 2021 through 2026, Duke received a total of $3,446,718,969 in Federal financial assistance from HHS.¹¹ Additionally, Duke operates health programs and activities that receive federal financial assistance.

When OCR opened its investigation on July 28, 2025, it requested Duke provide narrative responses and a corresponding document production responsive to its initial request for data. Duke provided responses to this initial data request between August 2025 and March 2026. Thereafter, OCR requested supplemental data from Duke on March 20, 2026, and requested Duke provide a complete response by June 26, 2026. Duke provided responses to the supplemental data request between April 2026 and June 26, 2026. Within its productions, Duke produced applicant-level data for admissions cycles for the incoming Duke SOM classes of 2020 through 2025.  Among its requests, OCR asked Duke SOM to produce an explanation for any deviation in test scores and grade point averages (GPA) between racial groups.¹² Duke SOM failed to produce anything in response to those requests, as well as failed to respond fully to a number of other requests. OCR has carefully reviewed and analyzed the evidence Duke produced which forms the basis of OCR’s findings.

3. Intent to Discriminate in Medical School Admissions

Background: Duke SOM admissions process before and after SFFA

Duke SOM employs a three-stage admissions process:¹³

1. Application screening to determine interview eligibility – For candidates that meet Duke SOM’s minimum qualifications (GPA of 3.0 or greater, Medical College Admission Test (MCAT) score of 500 or greater, and four letters of recommendation), members of the screener committee review application materials for clinical experiences, research experiences, community services and leadership potential, quality of essays, letters of recommendation, and considerations with regard to patient care.  Each application receives two preliminary screens. A third screener then makes a final determination on whether the applicant will proceed to an interview based on the results of the two preliminary screens. Application screening is the most selective phase of the admissions process. Duke SOM asserts that race and ethnicity data are blinded to individuals on the admissions committee, meaning application materials do not explicitly disclose an applicant’s race or ethnicity, and committee members do not have access to this information.
2. Interview evaluation – Duke SOM utilizes a virtual multiple mini-interview process whereby candidates rotate through stations and interact with multiple interviewers. Candidates are assessed for professionalism, critical thinking, ethics and decision-making, communication, resilience and adaptability, teamwork and leadership. At this phase (if not before), an applicant’s race and/or ethnicity may become apparent to the admissions committee members involved in the interview process.
3. Committee review and final admissions decisions – Candidates are then presented to the admissions committee highlighting academic credentials, experiences, essays, letters of recommendation, and interview performance. A selection subcommittee reviews the recommendations of the admissions committee and makes the final decision to admit, deny, or waitlist a candidate.

Prior to SFFA, an applicant’s race and ethnicity were included in the application materials, if provided by an applicant, and available to admission committee members during their review.  Race was one factor among many that Duke SOM considered as part of the admissions process.¹⁴ 

Duke SOM asserts that race and ethnicity play no part in its admissions decisions since SFFA.¹⁵ Duke SOM maintains that, despite SFFA, it is committed to expanding the pool of future physicians to include those who bring a wide range of backgrounds and experiences, regardless of race, color, or national origin.¹⁶  In support, Duke SOM asserts that applicants’ race and ethnicity on an application are blinded to those individuals on the admissions committee, and all committee members are instructed not to consider race or ethnicity, even if that information becomes inadvertently available, and states that committee members sign an attestation to that effect.¹⁷

Evidence of Duke SOM’s Intent to Discriminate Based on Race After SFFA

Duke SOM discriminates against white and Asian applicants by using race and race proxies in its admissions practices, all in violation of Title VI, Section 1557, and the SFFA decision.

In anticipation of the impending SFFA decision, the President of Duke University issued a letter to faculty, staff and students assuring them “that our work to advance racial and social equity—at Duke, in our local community and beyond—will not be deterred.”¹⁸ He went on to say that “[m]any of us are anticipating with apprehension potential changes in our federal law, and noting shifts in the local landscape that could adversely affect our community.”¹⁹ On June 29, 2023, the day the decision came down, the Duke Office of the President advised senior administrators against “making statements interpreting the decision or speculating on our practices.”²⁰ Regardless, the Duke SOM Dean of Admissions remarked to the Assistant Dean of Admissions, that, by his lights, SFFA is “not overturning prior cases authorizing race based affirmative action and suggested that how race has affected an applicant’s life can still be part of how their applicant is considered.”²¹ Another Duke SOM admissions committee member wrote, “We know in our hearts the right thing to do is to compensate for prior injustices but 6 of the 9 members of SCOTUS felt that enough has been done in the realm of education. I do not agree with flouting the law of race-blind admissions. But we were given an out and should encourage applicants from underrepresented groups to talk about challenges they faced.”²² This committee member thus indicated that Duke SOM’s admissions committee members could and should discern race from personal statements and essays, and reward and penalize them accordingly.

In the wake of the SFFA decision, Duke SOM application screeners were encouraged to “shift focus from [Diversity, Equity and Inclusion] DEI to inclusion” and to “focus on personal statements/(revised) essay questions.”²³ The Duke SOM Assistant Dean of Admissions remarked that the admissions “[c]ommittee members should have some DEI training before being considered—need to determine what each person has and if it meets the criteria.”²⁴ Duke SOM Screener Guidelines for the 2023-2024 admissions cycle²⁵ note that “DukeMed has blinded all Screeners from being able to see the applicants’ reported race or ethnicity,” however there is nothing keeping them from responding to race and ethnicity from essays, personal statements, and experiences, including “Social Justice Advocacy.”²⁶

By encouraging applicants to reveal their race through personal statements, essays and “Social Justice Advocacy,” the Duke SOM admissions committee was able to arrive at results of mean MCAT scores in the 88th, 86th and 87th percentiles for black admittees in the 2022-2023 (pre-SFFA), 2023-2024, and 2024-2025 admissions cycles, respectively. In contrast, the mean MCAT score for whites remained stiffly at the 97th percentile for each of those cycles, while Asians hovered between the 98th and 99th percentiles. Latino and Hispanic admittees ranked in the 94th, 92nd, and 89th percentiles for those cycles.

Among its data requests, OCR asked Duke SOM to produce an explanation for any deviation in test scores and GPAs between racial groups. Duke SOM failed to provide a response to those requests. Duke SOM’s failure to provide a responsive explanation for the GPA and MCAT deviations evident in the admissions data strengthens OCR’s inference when considered with the admissions communications and applicant-level data.

Duke SOM continues to use race and race proxies in admissions. Duke SOM internal documents and communications are unanimous in their commitment to DEI practices and policies that discriminate against whites and Asians. The documents and communications show Duke SOM admissions committee members and staff using subjective standards like “overcoming hardship” and “Social Justice Advocacy” that allow them to identify applicants by race and reward or penalize them accordingly, in a remarkably consistent fashion, year after year, despite the requirements of SFFA.

Applicant-Level and Aggregate Data Corroborating Intent and Causation

OCR assessed applicant-level admissions data for the 2024-2025 admissions cycle (class starting in August 2025) to determine whether race remained associated with admission after accounting for available academic, application, and admissions-stage variables. OCR then considered aggregate GPA, MCAT, and admission-rate data as contextual evidence. OCR performed a data analysis of the applicant-level admissions data Duke SOM produced for the 2024-2025 admissions cycle.  OCR’s analysis of this data shows significant disparities in MCAT scores and GPAs between different racial groups of admitted students. The analysis shows that Duke SOM admission rates for black and Hispanic applicants were two or three times higher than those of white or Asian applicants. For this admissions cycle, 3.3% of white applicants and 3.4% of Asian applicants were admitted, while 10.0% of black applicants and 6.7% of Hispanic applicants were admitted.  However, the average GPA and MCAT scores for black and Hispanic admitted applicants were lower than the GPA and MCAT scores of the white and Asian admitted applicants.

The following chart shows the mean data on admittees for the 2022-2023 admissions cycle (the last admissions cycle before SFFA):

The 2022-2023 admittees’ mean MCAT scores show significant differences across racial lines.  Blacks had the lowest mean scores at the 88th percentile, and Hispanics were at the next lowest at the 94th percentile, while whites and Asians were at the 97th and 98th percentiles, respectively. This is a significant 10-percentage point difference between blacks and Asians who were admitted.  The GPAs also show a disparity, with blacks having the lowest mean GPA, 0.13 grade points below the highest group, Asians.

The mean data on admittees for the 2023-2024 admissions cycle for the class starting in August 2024, a year after SFFA was decided, also included significant racial disparities:

As with the 2022-2023 admissions cycle, the mean MCAT scores for the admittees in the 2023-2024 admissions cycle exhibit significant differences across racial lines. Blacks and Hispanics had the lowest mean scores at the 86th and 92nd percentile, respectively, while whites were at the 97th percentile, and Asians were at the 99th percentile. The GPAs also continue to show a disparity, with blacks’ mean GPA 0.16 grade points below Asians.

For the 2024-2025 admissions cycle, the disparities between the academic qualifications of admitted students in different racial groups continued:

Black and Hispanic mean MCAT scores were at the 87th and 89th percentile, respectively, while whites were at the 97th percentile and Asians were at the 98th percentile. The GPAs also show a disparity, with Hispanics’ mean GPA falling 0.16 grade points below the highest group, Asians.

OCR’s applicant-level analysis found that, after accounting for available academic, application, and admissions-stage variables, race remained associated with statistically significant advantages or disadvantages at key stages of the medical school admissions process. The statistical significance of those race-associated effects strengthens OCR’s inference when considered with the admissions communications, race-revealing channels, and Duke SOM’s failure to provide a responsive explanation for the GPA and MCAT deviations evident in the admissions data. OCR’s statistical review corroborates OCR’s finding that Duke SOM’s race-revealing admissions channels continue to be associated with admissions outcomes after SFFA.

4. Allegations of Discrimination in Duke’s Scholarships and Financial Aid, Pipeline and Enrichment Programs, Curriculum and Training, and Employment Practices

The complaint received by OCR included additional allegations of discrimination on the basis of race in violation of Title VI and Section 1557 in Duke’s recruitment, scholarships, pipeline and enrichment programs, training and employment practices.³⁰  Duke represents that many of the programs and initiatives identified in the complaint ended in 2023 or earlier. However, Duke acknowledged offering one merit-based scholarship that included race as a consideration but asserted that the scholarship no longer considers race as of 2024. Duke also represents that implicit bias training for admissions committee members was last offered in 2024.

Additionally, Duke maintained an Equity, Diversity, and Inclusion (EDI) website, which promoted programs to expand the applicant pool for medical residency, and fellowship training and faculty positions.  Duke represents that the EDI website was removed in April 2025. One such pipeline program, Duke’s Building Overtures and Opportunities in Science and Technology program, remains active, but no language restricting participation based on race, color, or national origin is evident.

Duke also represents that its Multicultural Resource Center closed in 2025 and that its Minority Recruitment and Retention Committee ceased activities in 2025.

At present, based on Duke’s representations during the course of this investigation, OCR has determined not to further investigate the complaint’s allegations regarding Duke’s recruitment, scholarships, pipeline and enrichment programs, training, or employment practices.

5. Findings

OCR finds that Duke SOM used, and continues to use, race and race-revealing proxies in medical school admissions after SFFA. OCR bases this finding on the totality of the record, including Duke SOM’s admitted pre-SFFA race-conscious admissions practice, contemporaneous admissions communications, admissions materials preserving race-revealing channels, Duke SOM’s failure to provide a responsive non-race explanation for identified GPA and MCAT deviations, and applicant-level data corroborating race-associated admissions effects. In its highly selective medical school admissions process, Duke SOM’s use of race or race-revealing proxies advantaged some applicants at the expense of others and subjected applicants to differential treatment on the basis of race.

6. Procedure for Effecting Compliance

HHS regulations permit referral of matters to the United States Department of Justice if compliance under Title VI cannot be obtained voluntarily. 45 C.F.R. § 80.8(a). The regulations provide:

If there appears to be a failure or threatened failure to comply with this regulation, and if the noncompliance or threatened noncompliance cannot be corrected by informal means, compliance with this part may be effected by …  any other means authorized by law.  Such other means may include, but are not limited to … a reference to the Department of Justice with a recommendation that appropriate proceedings be brought to enforce any rights of the United States under any law of the United States (including other titles of the Act), or any assurance or other contractual undertaking … .

Further, “[n]o action to effect compliance by any other means authorized by law shall be taken until:

(1)    The responsible Department official has determined that compliance cannot be secured by voluntary means,

(2)    The recipient or other person has been notified of its failure to comply and of the action to be taken to effect compliance, and

(3)    The expiration of at least 10 days from the mailing of such notice to the recipient or other person. During this period of at least 10 days additional efforts shall be made to persuade the recipient or other person to comply with the regulation and to take such corrective action as may be appropriate.”

45 C.F.R. § 80.8(d).

This Notice of Violation constitutes official notice of Duke SOM’s failure to comply with Title VI, as required by subsection (2), and Section 1557. Consistent with sections 80.8(a) and (d), OCR seeks to secure compliance through voluntary means. If DUHS and Duke SOM are willing to come, voluntarily, into compliance with Title VI through execution and implementation of a voluntary resolution agreement, please contact Marisa M. Smith, Deputy Director for Enforcement.

If OCR does not hear from DUHS and Duke SOM within 10 business days or if DUHS and Duke SOM will not execute a voluntary resolution agreement satisfactory to OCR, we will have reason to determine, under 45 C.F.R. § 80.3(d)(1), that compliance cannot be secured by voluntary means. In such case, we will take all appropriate actions to secure DUHS’s and Duke SOM’s compliance with Title VI, including by promptly referring this matter to the Department of Justice.

If you have any questions about this letter, please contact Marisa M. Smith, Deputy Director for Enforcement at marisa.smith@hhs.gov or (214) 767-6973.  Thank you in advance for your attention and cooperation.

Sincerely,

/s/
Paula M. Stannard
Director, Office for Civil Rights

cc:    Marc Berger, Esq.
Latham & Watkins LLP
marc.berger@lw.com

Footnotes

¹  OCR enforces Federal civil rights laws that prohibit discrimination in the delivery of health and human services, by Department-funded programs and activities, based on race, color, national origin, disability, age, sex, religion, and the exercise of conscience. OCR also enforces the Health Insurance Portability and Accountability Act (HIPAA) Privacy, Security, and Breach Notification Rules and 42 C.F.R. Part 2 (Confidentiality of Substance Use Disorder Patient Records).

²  The complaint also alleged unlawful discrimination on the basis of race in violation of Title VI and Section 1557 in Duke’s programs and activities specific to scholarships and financial aid, pipeline and enrichment programs, curriculum and training, and employment practices.

³  42 U.S.C. § 2000d, and its implementing regulations at 45 C.F.R. Part 80. See also 42 U.S.C. § 18116, and its implementing regulations at 45 C.F.R. Part 92.

⁴  600 U.S. at 214-15.

⁵  Id. at 213; see id. at 218 (emphasizing "the twin commands of the Equal Protection Clause," i.e. "that race may never be used as a 'negative' and that it may not operate as a stereotype").

⁶  Id. at 214-15.

⁷  Id. at 230.

⁸  45 C.F.R. § 80.7(c) and 45 C.F.R. § 80.8. See also 45 C.F.R. § 92.303 (incorporating the procedural provisions of Title VI with respect to administrative enforcement actions against health programs and activities of recipients concerning discrimination on the bases of race, color and national origin, among others).

⁹  42 U.S.C. § 2000d-1.

¹⁰  45 C.F.R. § 80.8 and 45 C.F.R. § 80.8. See also 45 C.F.R. § 92.303.

¹¹  https://taggs.hhs.gov/Detail/RecipDetail?arg_EntityId=AN4%2B6Hbo1NsMNcj4TaMJKQ%3D%3D (last visited July 29, 2026).

¹²  OCR’s Second Data Request (March 20, 2026), Data Request 19(c) and (e).

¹³  Duke’s Fifth Response to OCR’s Initial Data Request (January 30, 2026).

¹⁴  Id.

¹⁵  Id.

¹⁶  Id.

¹⁷   Id.

¹⁸  Email from Duke President Vincent E. Price, dated June 8, 2023, DUHS-OCR00002485.

¹⁹  Id.

²⁰  Email from Margaret Epps, Chief of Staff to the Duke President, dated June 29, 2023, DUHS-OCR00002751.

²¹  Email from Andrea Liu, Assistant Dean for Admissions, dated June 29, 2023, DUHS-OCR00002501.

²²   Email from Bradley Collins, Associate Professor of Surgery and Admissions Committee Member, to Linton Lee, M.D., Associate Dean for Admissions, dated January 31, 2024, DUHS-OCR00003776.

²³  Identifying and Developing the Next Generation of Thought Leaders in Medicine, Research, and Patient; Care Duke Screener Rubric (2024-25 Admissions Season), DUHS-OCR00001585.

²⁴  Email from Andrea Liu, Assistant Dean of Admissions, to Admissions Committee Member Dr. Taylor Broome, dated April 22, 2024, DUHS-OCR00003811.

²⁵   The 2023-2024 admission cycle began in July 2023, for the medical school class that entered Duke in the fall of 2024. Accordingly, this admission cycle was completely post-SFFA. It appears that an admissions committee is formed in the summer, with standard applications received between July and October, and the Duke secondary application due by mid-November; interviews are conducted between September and February, with final decisions made in February and classes starting in August.

²⁶   Attachment to email from Andrea Liu, Assistant Dean of Admissions, Duke Screener Guidelines (2023 Admissions Cycle), Emphasis added, DUHS-OCR00003838.

²⁷  https://students-residents.aamc.org/media/13381/download.

²⁸  https://students-residents.aamc.org/media/14536/download.

²⁹  https://students-residents.aamc.org/media/15781/download.

³⁰  OCR determined the employment practices cited to in the complaint pertained to a specific individual and that that individual had independently filed a complaint with the Equal Employment Opportunity Commission; therefore, OCR declines to further review this allegation.