Written by Rayyan Bhuiyan
Edited by Anling Chen
Today, algorithms are trusted to make various decisions for us by crunching data and using it to make predictions. By doing so, these neat computations can fill TikTok feeds with addictive content, suggest efficient routes on Google Maps, and produce output for artificial intelligence tools like ChatGPT. Decision-making is frequently influenced or done by set equations, thus healthcare professionals as well utilize algorithmic tools to diagnose cancer faster or assess specific organ functions. While algorithms may have their benefits like quicker turn-around times during the diagnostic process, it is important to consider how certain race-based factors in formulating these algorithms contribute to racial inequalities in healthcare.
Various socioeconomic considerations must be taken into account when explaining racial inequalities in healthcare, such as racism against minority patients, economic instability, health literacy gaps, and more (Maness et al., 2020). It is also important to recognize how race can be harmfully utilized in calculations, as it is being considered in algorithms relating to kidney, lungs, and other organ functions. For instance, popular equations like the estimated Glomerular Filtration Rate (eGFR) calculation for assessing kidney function contribute to inequality by overestimating the health of Black patients.
eGFR equations are used to categorize a patient’s kidney function, with a higher score implying a healthy kidney and a lower score signifying near kidney failure (Uppal et al., 2022). While there are multiple formulas, the two most popular ones utilize a biomarker known as serum creatinine, while also considering gender, age, and a multiplier of about 1.2 if the patient is Black (National Institute of Diabetes and Digestive and Kidney Diseases, 2023). This race multiplier is included due to a bias in a pilot study from 1999 that created the GFR equation during which physicians suggested “on average, black persons have higher muscle mass than white persons,” when they saw higher serum creatinine concentrations in Black individuals (Levey, 1999). However, the study failed to measure muscle mass or consider other factors like diet or athletic ability that may lead to a high creatinine level.
Thus, these race factors may have been fueled by racial discrimination, which is important as a patient’s score determines the health interventions they can receive, from getting a referral to receiving organ transplants (Tsai, 2021). Inflated scores have dire consequences, evident when evaluating the kidney health of the American Black population. Black people are four times more likely than their White counterparts to end up with kidney failure; despite making up only 13% of America, 35% of dialysis-receiving patients are Black (Uppal et al., 2022). These statistics are alarming, as they likely have to do with non-medical biases present in kidney function metric calculations. Unfortunately, the kidney function equation is not the only biased equation in the healthcare system.
Assessments used for estimating lung function and the likelihood of death in the hospital use similar parameters and assumptions that are skewed by human bias. The Pulmonary Function Testing equation (PFT) assumes that Black individuals have lower lung capacity, leading to Black individuals being marked with higher lung function, delaying necessary treatment for what could be unhealthy lungs (Einstein et al., 2023). Algorithms used to estimate the risk of death for patients admitted to hospitals and the risk of complications from cardiac surgery also involve preconceived assumptions about Black people, which may prevent them from getting the optimal treatment (Peterson et al., 2010).
On the other hand, algorithmic-based assessments also have their benefits in the healthcare world, like their speed and accuracy in diagnosing certain medical conditions. Dr. Daniel Orringer at NYU Langone Health developed Deep Gioma, a device that uses A.I. to reduce the diagnosis time of Glioblastoma, a type of vicious brain tumor, from four weeks to three minutes (NBC News, 2023). This is groundbreaking, as Glioblastoma tumors are extremely deadly and patients survive for a maximum of eight months following their diagnosis. Waiting four weeks to have a potentially lethal tumor analyzed and diagnosed allows it to grow in mass exponentially; with faster diagnoses that A.I. developed from algorithms, hopes are that cancer treatment outcomes improve dramatically (ESMO, 2023).
Overall, it is important to consider the tools we use and with what intentions they were formulated. By using biased algorithms, Black patients are less likely to receive proper treatment, as seen with kidney, lung, and other types of calculations used in diagnosis. Most physicians are unaware that such biases are included in the algorithms, letting this problem go unchecked and fester. While algorithms have the potential to revolutionize healthcare by improving diagnosis abilities, we must also be aware that they can be used to perpetuate racial inequalities, depending on how the algorithm was developed. Hopefully, with increased awareness and further research, we can aim to use only unbiased equations, resulting in more equitable treatment of patients.
References
Einstein, L., Elmaleh-Sachs, A., Maru, D., Morse, M., Mph, M., Khazanchi, R., & Mph, A. (2023). Evaluating the Elimination of Race from PFT Equations. https://www.nyc.gov/assets/doh/downloads/pdf/cmo/cerca-pft-evaluation.pdf
ESMO. (2023, October 16). The Potential of AI to Improve Cancer Care is Only Going to Grow. Www.esmo.org. https://www.esmo.org/newsroom/press-releases/the-potential-of-ai-to-improve-cancer-care-is-only-going-to-grow
Levey, A. S. (1999). A More Accurate Method To Estimate Glomerular Filtration Rate from Serum Creatinine: A New Prediction Equation. Annals of Internal Medicine, 130(6), 461. https://doi.org/10.7326/0003-4819-130-6-199903160-00002
Maness, S. B., Merrell, L., Thompson, E. L., Griner, S. B., Kline, N., & Wheldon, C. (2020). Social Determinants of Health and Health Disparities: COVID-19 Exposures and Mortality Among African American People in the United States. Public Health Reports, 136(1), 18–22. https://doi.org/10.1177/0033354920969169
National Institute of Diabetes and Digestive and Kidney Diseases. (2023). Estimating Glomerular Filtration Rate | NIDDK. National Institute of Diabetes and Digestive and Kidney Diseases. https://www.niddk.nih.gov/health-information/professionals/clinical-tools-patient-management/kidney-disease/laboratory-evaluation/glomerular-filtration-rate/estimating
NBC News. (2023, May). Live brain surgery: see how doctors are using A.I. in the O.R. Www.youtube.com. https://www.youtube.com/watch?v=Bcgms_xkpSg&t
Peterson, P. N., Rumsfeld, J. S., Liang, L., Albert, N. M., Hernandez, A. F., Peterson, E. D., Fonarow, G. C., & Masoudi, F. A. (2010). A Validated Risk Score for In-Hospital Mortality in Patients With Heart Failure From the American Heart Association Get With the Guidelines Program. Circulation: Cardiovascular Quality and Outcomes, 3(1), 25–32. https://doi.org/10.1161/circoutcomes.109.854877
Tsai, J. (2021, June 27). Jordan Crowley Would Be in Line for a Kidney—if He Were Deemed White Enough. Slate Magazine. https://slate.com/technology/2021/06/kidney-transplant-dialysis-race-adjustment.html
Uppal, P., Benjamin Ira Golden, Panicker, A., Khan, O. A., & Burday, M. J. (2022). The Case Against Race-Based GFR. Delaware Journal of Public Health, 8(3), 86–89. https://doi.org/10.32481/djph.2022.08.014