AI-driven Healthcare: Bridging Gaps or Widening Disparities?

By: Emily Zhang, Human Biology, Health and Society Class of 2027

Medical racism is not an issue of the past. Despite efforts to combat race-based medicine, racism continues to be deeply ingrained within the US healthcare system both implicitly and explicitly [1]. Racialized biases in healthcare are highly consequential especially for the  minority population, leading to poorer health outcomes and quality of life. Now, as the healthcare industry becomes increasingly digitized, a new and pressing threat to health equity emerges: artificial intelligence. 

Artificial intelligence (AI) is at the forefront of the digital healthcare revolution. Through its diagnostic potential, ability to perform administrative tasks, and potential  in creating life-saving drugs, there is no doubt that AI is well underway in addressing healthcare challenges worldwide. But despite its massive potential, research suggests that these systems can perpetuate long-standing racial disparities and biases [2]. But, how exactly can these automated systems harbor bias? 

The answer lies within the datasets that these AI are trained with. AI relies on machine learning, which is a technique that trains algorithms to identify patterns within data to create accurate associations and predictions [2]. The success of machine learning is dependent on high-volume and high-quality data. However, the data used to train AI systems can be flawed by lacking diversity, whether by sex, race, or other factors. As Mattie of Harvard University explains, “bias can creep into the process anywhere in creating algorithms: from the very beginning with study design and data collection, data entry and cleaning, algorithm and model choice, and implementation and dissemination of the results” [3]. 

Algorithmic bias also stems from the reality that racialized minorities have been historically excluded from medical research. Whether it be due to a fear of exploitation, distrust of the medical professionals, or lack of time or resources, participants of color are not equally represented in clinical trials [4]. When trained on these datasets that are not representative of the full population, AI can harm these underrepresented groups predominantly those of color.  “When you learn from the past, you replicate the past. You further entrench the past,” says Dr. Mark Sendak, a lead data scientist at the Duke Institute for Health Innovation [5]

The implications of AI bias are direct and harmful. In a landmark study published by Science in 2019, Obermeyer et al. found that a widely used algorithm that affects 100 million patients has significant bias. In this model, health risk was predicted using past healthcare spending. Due to Black patients having unequal access to care, they often spend less on healthcare despite their actual health conditions. Therefore, this model consistently underestimated the health needs of Black patients [6]. Another study published in JAMA Dermatology Network identifies disparities in how skin cancer is diagnosed across different skin colors as most models are trained with light-skinned subjects [7]. While developers of these algorithms hadn’t intended to include these biases, these flaws can affect the diagnoses and recommendations physicians provide and translate to lesser quality of care for minorities. 

As of 2024, the first drugs designed with AI begin to reach clinical trials [8].  AI-driven technologies have also been shown to be capable of tackling administrative tasks and interpreting images— radiographs, histology, and optic fundi [9]. It has even been shown that two large language models including Chat GPT can pass the USMLE, the three-step exam for medical licensure [10]. While AI won’t necessarily replace the need for physicians, there is no question that its presence in healthcare will continue to rapidly expand. However, the implementation of AI requires multidisciplinary efforts from data scientists, physicians, and legal professionals to make combating bias a priority. Leveraging the capabilities of AI in the right ways is crucial, ultimately seeking to promote greater equity instead of worsening divides. 

References

1. Yearby, R., Clark, B., & Figueroa, J. F. (2022). Structural racism in historical and modern US health care policy. Health Affairs, 41(2), 187–194. https://doi.org/10.1377/hlthaff.2021.01466

2. Vokinger, K. N., Feuerriegel, S., & Kesselheim, A. S. (2021). Mitigating bias in machine learning for medicine. Communications Medicine, 1(1). https://doi.org/10.1038/s43856-021-00028-w

3. Igoe, K. (2021, March 12). Algorithmic Bias in Health Care Exacerbates Social Inequities — How to Prevent It | Executive and Continuing Professional Education | Harvard T.H. Chan School of Public Health. Www.hsph.harvard.edu. https://www.hsph.harvard.edu/ecpe/how-to-prevent-algorithmic-bias-in-health-care/

4. THE EDITORS. (2018). Clinical Trials Have Far Too Little Racial and Ethnic Diversity. Scientific American, 319(3). https://doi.org/10.1038/scientificamerican0918-10

5. Levi, R., & Gorenstein, D. (2023, June 6). AI in medicine needs to be carefully deployed to counter bias – and not entrench it. NPR. https://www.npr.org/sections/health-shots/2023/06/06/1180314219/artificial-intelligence-racial-bias-health-care

6. Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342

7. Adamson, A. S., & Smith, A. (2018). Machine Learning and Health Care Disparities in Dermatology. JAMA Dermatology, 154(11), 1247. https://doi.org/10.1001/jamadermatol.2018.2348

8. Heaven, W. D. (2023, February 15). AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work. MIT Technology Review. https://www.technologyreview.com/2023/02/15/1067904/ai-automation-drug-development/

9. Beam, A. L., Drazen, J. M., Kohane, I. S., Leong, T.-Y., Manrai, A. K., & Rubin, E. J. (2023). Artificial Intelligence in Medicine. New England Journal of Medicine, 388(13), 1220–1221. https://doi.org/10.1056/nejme2206291

10. DePeau-Wilson, M. (2023, January 19). AI Passes U.S. Medical Licensing Exam. Www.medpagetoday.com. https://www.medpagetoday.com/special-reports/exclusives/102705