Recent findings from the University of Minnesota School of Public Health have put the spotlight on the disparities in healthcare brought on by artificial intelligence (AI). According to the research, which was disseminated in a University report, a significant number of hospitals across the U.S. are integrating AI into their operations, yet the scrutiny these tools undergo for biases remains alarmingly inconsistent.

The study draws from the American Hospital Association Annual Survey, tapping into the experiences of 2,425 hospitals, revealing that AI-assisted predictive models are very much in use—for instance, to predict inpatient health trajectories and for administrative workflows like billing and scheduling, however, when it comes to assessing these models for biases, the statistics start to dwindle. While a reasonable 61% of hospitals evaluated their predictive models for accuracy, a mere 44% took the initiative to examine potential biases present in these systems.

A gap in resources seems to be at the heart of the varying approaches to AI. Hospitals equipped with more robust finances and technical capabilities were shown to both possess and critically evaluate their AI systems more than those dependent on off-the-shelf solutions, perhaps pointing to an intrinsic digital divide. "The growing digital divide between hospitals threatens equitable treatment and patient safety," Paige Nong, the study's lead author and School of Public Health assistant professor, told the University of Minnesota. She raised a compelling point, highlighting how wealthier institutions can afford custom-built models that they then diligently assess in-house—contrary to their less fortunate counterparts.

Future endeavors by the research team seem promising as they plan to peel back the layers on the prevalence of AI applications such as ambient scribes and chatbots in hospitals not only that, their collaboration with diverse organizations aims to yield insights that could very well influence the formulation of policies and best practices regarding AI in healthcare settings. This research received backing from the Office of the Assistant Secretary for Technology Policy at the U.S. Department of Health and Human Services, which itself signifies a recognition of the essential nature of this inquiry.