Don’t Overestimate AI
“With the increasing use of artificial intelligence and machine learning in other spheres, there has been an increase in interest in exploring how they can be utilized to improve clinical outcomes,” said Suman Pal, MD, assistant professor in the division of hospital medicine at the University of New Mexico, Albuquerque, in an interview. “However, concerns remain regarding the possible harms and ways to mitigate them,” said Dr. Pal, who was not involved in the current study.
In the current study, “It was interesting to note that explanations did not significantly mitigate the decrease in clinician accuracy from systematically biased AI model predictions,” Dr. Pal said.
“For the clinician, the findings of this study caution against overreliance on AI in clinical decision-making, especially because of the risk of exacerbating existing health disparities due to systemic inequities in existing literature,” Dr. Pal told this news organization.
“Additional research is needed to explore how clinicians can be better trained in identifying both the utility and the limitations of AI and into methods of validation and continuous quality checks with integration of AI into clinical workflows,” he noted.
The study was funded by the National Heart, Lung, and Blood Institute. The researchers had no financial conflicts to disclose. Dr. Pal had no financial conflicts to disclose.
A version of this article first appeared on Medscape.com.