Racial Bias in Healthcare and the use of Disparate-Impact as a Remedy

The COVID-19 pandemic put a spotlight on the increasing vaccine hesitancy that exists in the United States.  Hesitancy among the African American population stood out because of COVID-19’s disproportionate effect on their community.[i]  Distrust of the medical establishment within minority groups is partly to blame for this hesitancy.[ii]  The distrust is not unwarranted, with two historical events typically used to explain the origin: HeLa cells and the Tuskegee study.[iii] 

            The biomedical research community has used HeLa cells for over seventy years.[iv]  The cells were initially obtained in 1951 from Henrietta Lacks, an African American woman, for diagnostic purposes.[v]  Observing the cells to be “immortal,” meaning they could be grown outside the body, researchers continued growing them without Ms. Lacks’ consent.[vi]  The researchers found the cells helpful in research and began using them in experiments and sharing them with other researchers.[vii]  Since 1953 over 110,000 biomedical research studies have used HeLa cells, including those used to develop COVID-19 vaccines.[viii]  Researchers have made efforts to reconcile with her family.[ix]  But the damage from seventy years ago remains a wound within the African-American community.[x] 

            The Tuskegee Syphilis study began in 1932.[xi]  Doctors recruited African-American men with the promise of free healthcare.[xii]  But then researchers intentionally infected the participants, without their knowledge, with syphilis to observe the disease progression.[xiii]  Participants received no treatment even after penicillin was identified as a treatment in 1947.[xiv]  The study continued into the 1970s before a reporter revealed the details, and public backlash forced the end of the study.[xv]  In 1973, Congress held hearings that resulted in a $10 million settlement to surviving participants and heirs of the deceased.[xvi]  But still, fifty years later, the effects of the study linger within the African American population.[xvii] 

            The medical establishment has taken steps to resolve the damage caused by HeLa cells and the Tuskegee study with the African American community.[xviii]  But another threat has emerged that could cause more damage to the medical community’s reputation with underrepresented communities, including African Americans.[xix]  This threat is racial bias in healthcare-related artificial intelligence (AI).[xx]  Healthcare AI has been used increasingly in various health fields and is projected to save $150 billion in healthcare expenses by 2026.[xxi]  But algorithms used within AI can produce racial bias.[xxii]  In 2020 a New England Journal of Medicine study identified three examples of racial bias already existing within healthcare-related AI systems.[xxiii]  The consequences of these racial biases include a lower risk of detecting heart failure in black patients, lowered prediction of success for vaginal birth after a Cesarean for black and hispanic populations, and a higher risk of missing kidney stone diagnosis for black patients.[xxiv] 

This study raises the question of what legal remedy is available to someone affected by racial bias within the healthcare system.  The Civil Rights Act of 1964 Title VI provides two categories of protection against discrimination: intentional and disparate-impact.[xxv]  Intentional discrimination requires some evidence of discriminatory intent, which would cover incidents like the Tuskegee study.[xxvi]  But disparate-impact discrimination does not require discriminatory intent; rather, a benign, neutral practice that produces a racially disproportionate result.[xxvii]  Disparate-impact discrimination would cover incidents evolved from racial bias within healthcare-related AI.

The problem with disparate-impact discrimination is that it does not provide a private cause of action for relief.[xxviii]  Disparate-impact policy provides federal agencies the authority to issue rules, regulations, or orders.[xxix]  The passage of the Affordable Care Act (ACA) §1557 contained a provision to cover the prohibition of discriminatory practices to healthcare services- including disparate-impact discrimination.[xxx]  Initially, some thought that the ACA would provide a private action for relief from disparate-impact discrimination.[xxxi]  But in 2015, the Supreme Court held that disparate-impact discrimination relief under the ACA should mimic relief under Title VI.[xxxii] 

The ACA ruling leaves regulation as the effective route for mitigating disparate-impact discrimination.  Considering the effect of the Henrietta Lacks and Tuskegee study on the African-American communities, prevention is a far better strategy than an individual cause of action.  Legal actions such as tort remedy would leave a method for recovery, but only after the damage has occurred.  A possible framework for regulation exists in the Health Insurance Portability and Accountability Act of 1996 (HIPAA), which acts as a stop-gap for healthcare IT and privacy offenses.[xxxiii]  Algorithms, which control AI, could be regulated during the design process and ongoing monitoring after implementation to control bias.[xxxiv]  These regulatory policies could effectively prevent major incidents of healthcare-related bias from the AI systems.  Prevention of these events is critical for the medical establishment in restoring trust in communities still hurting from HeLa cells and the Tuskegee study.

[i] Amudalat Ajasa, For Some Black Americans, Vaccine Hesitancy is Just one Part of a Legacy of Mistrust, The Guardian (Aug. 20, 2021 06:30 AM), https://www.theguardian.com/society/2021/aug/20/black-americans-covid-19-vaccine-hesitancy-mistrust.

[ii] Id.

[iii] Id.

[iv] Editorial, Henrietta Lacks: Science must right a wrong, 585, Nature 7, 8 (2020). 

[v] Id. 

[vi] Id. 

[vii] Id. 

[viii] Id. 

[ix] Id at 10. 

[x] Ajasa, supra note 1.

[xi] Elizabeth Nix, Tuskegee Experiment: The Infamous Syphilis Study, History (May 16, 2017) https://www.history .com/news/the-infamous-40-year-tuskegee-study. 

[xii] Id. 

[xiii] Id. 

[xiv] Id. 

[xv] Id. 

[xvi] Id. 

[xvii] Ajasa, supra note 1.

[xviii] Id. 

[xix] Katherine J. Igoe, Algorithmic Bias in Health Care Exacerbates Social Inequities- How to Prevent it, Harvard Sch. of Pub. Health (March 12, 2021), https://www.hsph.harvard.edu/ecpe/how-to-prevent-algorithmic-bias-in-health-care/. 

[xx] Id. 

[xxi] Jen A. Miller, Artificial Intelligence is poised to transform Point-of-Care Testing, 47 Clinical laboratory News 10, 10 (2021). 

[xxii] Darshali A. Vyas, Hidden in Plain Sight-Reconsidering the Use of Race Correction in Clinical Algorithms, 383 New Eng. J. of Med. 874, 874 (2020). 

[xxiii] Id. 

[xxiv] Id. 

[xxv] Applying Title VI of the Civil Rights Act of 1964, American Bar Association (Nov. 1, 2016), https://www.americanbar.org/groups/public_interest/child_law/resources/child_law_practiceonline/child_law_practice/vol-35/november-2016/applying-title-vi-of-the-civil-rights-act-of-1964/.

[xxvi] Id. 

[xxvii] Id. 

[xxviii] Alexander v. Sandoval, 532 U.S. 275, 293 (2001). 

[xxix] Id. 

[xxx] Sidney Watson, Health Care Reform and Vulnerable Communities: Can We Afford it? Can We Afford to Live Without It?  55 How. L. J. 855, 870 (2012). 

[xxxi] Id at 880. 

[xxxii] SE. Pa. Transp. Auth. v. Gilead Sciences, Inc, 102 F. Supp. 3d 688, 701 (2015). 

[xxxiii] Jonathan P. Tomes, 20 Plus Years of HIPAA and What Have We Got?, 22 Quinnipiac Health L.J. *39, *41 (2018). 

[xxxiv] Nicolas Schmidt, An Introduction to Artificial Intelligence and Solutions to the Problem of Algorithmic Discrimination, 73 Consumer Fin L.Q. 130, 133 (2019).