By researching health disparities, I am learning that there are disparities in access to health insurance and healthcare, quality of care received based on geographic regions, and real biological differences. For example, University of Michigan Medicine researchers have been studying racial disparities in the healthcare system to gain a better understanding of how complex factors create differing health outcomes for Black and white Americans. With this research I am learning how the use of artificial intelligence (AI ) in healthcare helps combat some of these disparities.
Researchers note that Black patients tend to report suffering more intense pain than white patients but are less likely to undergo knee replacement surgery. Interestingly, U-Michigan researchers are training a new algorithm to read knee X-rays for patients with arthritis found that their AI program performed more efficiently in diagnosing Black patients’ reported pain than human radiologists.
Yet, my research revealed that some AI tools have increased disparities rather than combating them. For example, an algorithm previously used by a health insurance company to predict future healthcare costs and recommend patients for care actually reduced care provided for Black patients from 50% to 20%. As AI and ML algorithms scale, they take more repetitive tasks, so inequities included in these algorithms can increase as well. As these algorithms used cost as an endpoint, the algorithm was faulty, even though it was intended to be theoretically race-blind. In contrast, the algorithm researchers trained on knee X-rays was “trained on patient reports of pain” so as Black patients reported higher levels of pain with clinically equivalent images, AI-assisted Xray readings were improved and resulted in more accurate diagnosis of disease patterns and in turn, led to better care for Black patients.
References:
Adnan Asar “AI Could Reduce Racial Disparities In Healthcare” Forbes Technology Council 1 Oct 2021 online.
Jordyn Imhoff “Health Inequality Actually Is a “Black and White Issue” University of Michigan Health 3 June 2020 online.
Wired “New Algorithms Could Reduce Racial Disparities in Health Care: Machine learning programs trained with patients’ own reports find problems that doctors miss—especially in Black people” 25 Jan 2021 online.
