How to address inequity in healthcare AI? Hire a diverse data team
Even as artificial intelligence has become more thoroughly integrated into healthcare, experts have cautioned about its potential downsides – including the potentially hazardous risk of bias and discrimination encoded into algorithms.
In fact, said Dr. Tania M. Martin-Mercado, digital advisor in healthcare and life sciences at Microsoft, the implication that automation will miraculously improve care delivery is one that hasn’t been fully explored.
“This is an over-reaching generalization,” said Martin-Mercado, who will be presenting on the subject at HIMSS22 in March.
Further, she continued, the assumption “does not take into consideration the multitude of factors that contribute to the very bias and inequity that is sought after by implementing AI in healthcare.”
Algorithms have the potential to increase inequities, she explained, when they don’t take into account considerations such as socioeconomic status, gender, disabilities, sexual orientation and other factors that contribute to disparities in outcomes.
She says one way to mitigate such risks is by ensuring the professionals working with that information represent a wide variety of backgrounds.
“In other words, homogenous data teams should be avoided,” she said.
She also stresses the importance of acknowledging how bias exists in all of us.
“Removing the emotional attachment to the awareness of one’s bias allows for a colleague or coworker to address that bias in a safe way,” she said.
This is important, she explained, because responding with denial or defensiveness does not address the problem.
“In order to see the change and improve health equity, we must be able to be accountable for the bias that we all have and start there,” she continued.
At HIMSS, she hopes panel attendees will internalize this message, coming away inspired to act when they encounter bias in healthcare and data.
Another equally important lesson, she said, is to ensure that a diverse group of professionals are entrusted with looking at and working with data.
“This will ultimately reduce damage to patients,” she said.
Martin-Mercado will explain more in her HIMSS22 session, “How Implicit Bias Affects AI in Healthcare.” It’s scheduled for Wednesday, March 16, from 1-2 p.m., in Orange County Convention Center W300.
Kat Jercich is senior editor of Healthcare IT News.
Twitter: @kjercich
Email: [email protected]
Healthcare IT News is a HIMSS Media publication.
Source: Read Full Article