Hospitals are more likely to assess artificial intelligence and predictive models for accuracy than for potential biases, according to a study published Jan. 6 in Health Affairs.
Researchers at the University of Minnesota School of Public Health analyzed data from 2,425 hospitals nationwide that participated in the 2023 American Hospital Association Annual Survey.
Three key study findings:
- About 65% of hospitals reported using AI-based predictive models, the majority of which were developed by their EHR vendor.
- Of the hospitals using AI models, 61% reported evaluating them for accuracy, while just 44% reported evaluating them for bias.
- Hospitals with higher operating margins, in-house model development capabilities or membership in a health system were more likely to evaluate models for bias.
"The growing digital divide between hospitals threatens equitable treatment and patient safety," lead author Paige Nong, PhD, an assistant professor at University of Minnesota School of Public Health, said in a news release. "Many better-funded hospitals can design models tailored to their own patients, then conduct in-house evaluations of them. In contrast, hospitals with fewer resources are buying these products 'off the shelf,' which may not reflect the needs of local patients."
The researchers recommended policies and programs that provide technical support and resources to help hospitals effectively evaluate and govern AI models to ensure safe and equitable use across the industry.