ARITIFICIAL INTELLIGENCE AND HEALTH EQUITY
Artificial Intelligence (AI) is everywhere. It answers the customer service phone lines of companies. It’s the online chatbot anxiously awaiting to assist you. It’s behind the algorithm in your browser searches that determines the results you see. It can do your search for reference material to write a paper. It can summarize a report or paper, so you don’t have to read it in its entirety. Heck, it can write the paper and cite the sources. AI also has many applications in healthcare. Like other new technologies, it is being touted as a solution to improving healthcare and addressing health equity.
Simply stated, AI is the ability of computers to perform tasks commonly associated with intelligent beings like learning and problem solving.1 Given the volume of data available to providers today, diagnosis and treatment have become even more complex. AI uses the data available in the electronic health record (EHR) to identify patterns and predict outcomes. This information can be used in clinical decision making by the health care provider.2 However, AI’s performance is based on the accuracy of the information or data provided for its analysis.
Healthcare and Implicit Bias
Health record data like health outcomes, are affected by unconscious or implicit bias. Implicit bias is a negative attitude, of which you are not aware (unconscious) against a specific group.3 Groups commonly affected include but are not limited to gender, race and ethnicity, sexual orientation and gender identity, age, and socio-economic status. For example, holding your purse tighter when a male of a certain race walks by, or assuming certain groups of people are more likely to be criminals or inferior are examples of implicit bias.
Research has shown that healthcare providers (i.e. doctors, nurses, medical students) have implicit bias.4 This implicit bias has led to the under treatment of pain in some racial groups,5 and disparities in providing evidenced based treatment for heart disease in other groups.6 No surprise that health care providers have implicit bias because we all have it. The development of implicit bias is the result of how our biological personal computer, the brain, processes information. Through our senses of smell, vision, taste, touch, and hearing, we process billions of stimuli or data unconsciously and store it. This data comes from everyday experiences with people, movies, media, etc. Our brain develops patterns and the ability to quickly access and use this information without thought (unconsciously). This allows us to perform many tasks quickly.7 This is useful for avoiding a hot stove. However, when the patterns are negative stereotypes (e.g. assuming some groups are drug seeking and not really in pain), then their application may lead to harm.
AI and Bias
It is not only the bias within the electronic health record which causes concern. It is also the lack of diversity of those who are creating, developing, reviewing, and making final decisions on the AI tools used. A small homogeneous group of people typically have a narrow frame of reference. They will be less likely to recognize the inaccuracies secondary to bias. Do the rich have the same perception of cost as those with lower socio-economic status? Do males recognize female gender bias?
Moreover, once the AI tools are developed, the end users (e.g. providers) cannot see the process that led to the results. Therefore, errors will not be obvious, and it will take longer to recognize them. Consequently, the development of AI tools without diverse input could move us further away from health equity. To paraphrase Burr in Hamilton, we need adequate representation in the room where it happens, when the sausage is being made,8 if we are to assure that AI has a positive impact on health care outcomes.
References
[1] Copeland, B.J.. “artificial intelligence”. Encyclopedia Britannica, 27 Dec. 2024, https://www.britannica.com/technology/artificial-intelligence. Accessed 30 December 2024.
[2] Bajwa J, Munir U, Nori A, Williams B. Artificial intelligence in healthcare: transforming the practice of medicine. Future Healthc J. 2021 Jul;8(2):e188-e194. doi: 10.7861/fhj.2021-0095. PMID: 34286183; PMCID: PMC8285156.
[3] Adapted from the American Psychological Association and Wikipedia (https://www.apa.org/topics/implicit-bias). Accessed 30 December 2024.
[4] Hall WJ, Chapman MV, Lee KM, et al. Implicit Racial/Ethnic Bias Among Health Care Professionals and Its Influence on Health Care Outcomes: A Systematic Review. American Journal of Public Health. 2015. Dec; 105(12):e60-e76.
[5] Hoffman, KM, Trawalter S, Axt JR, MN Oliver. Racial bias in pain assessment and treatment recommendations, and false beliefs about biological differences between blacks and whites. Proc. Natl. Acad. Sci. U.S.A. 113(16)4296-4301, https://doi.org/10.1073/pnas.1516047113 (2016).
[6] Nelson AR. Unequal treatment: report of the Institute of Medicine on racial and ethnic disparities in healthcare. Ann Thorac Surg. 2003 Oct;76(4):S1377-81. doi: 10.1016/s0003-4975(03)01205-0. PMID: 14530068.
[7] Kahneman, D. (2011). Thinking, Fast and Slow. P 20-24.
[8] Miranda, L., Lacamoire, A., & Chernow, R. (2016). Hamilton: an American musical. Vocal selections. [Los Angeles, California], Warner/Chappell.