A majority of patients expressed a desire to be notified when asked about the use of AI in health care, according to a research letter published in JAMA Network Open.

Results like these regarding patient opinions "can support policymakers and health systems in priority setting" and suggest that AI notification is a necessity in the future, the researchers pointed out.

Patients have shown a more positive, although cautious, outlook regarding AI use in health care. In a recent survey, most respondents expressed comfort with AI being used to improve health care delivery and recognized the technology's possible value, but some felt its use could be 'scary' and many recognized concerns about data privacy.

Patient notification "is central to data privacy laws and informed consent," Jodyn Platt, MPH, PhD, an associate professor of learning health sciences at the University of Michigan, and colleagues wrote.

"As [AI] applications increase across industries, policy frameworks highlight transparency via notification as a core component of appropriate AI use," they added. "However, requirements and policies for such notifications are not well established, and health systems vary in their commitment to transparency."

In the analysis, researchers surveyed 2,021 individuals on their desire for notification of AI use. Specifically, they asked patients whether it was "not true at all," "somewhat true," "fairly true" or "very true" that it is important that they are notified.

Platt and colleagues weighted responses, both overall and by demographic groups, to the question using poststratification survey weights from the Current Population Survey.

Overall, 62.7% (95% CI, 60%-65.8%) of respondents said it was very true they wanted to be notified, whereas just 4.8% (95% CI, 3.7%-6.2%) did not find notification to be important.

The researchers reported a weighted mean score on the desire for notification of 3.39 (95% CI, 3.33-3.44), with the following maximum differences in mean score:

The researchers pointed out that the desire to be notified was higher for AI use than that for health information or biospecimens.

They also acknowledged multiple study limitations. These included unmeasured error and the analysis' cross-sectional design, the latter of which warrants "future longitudinal studies to examine change over time."

Platt and colleagues highlighted several differences observed among demographic groups, "particularly in the ethical context of historical, structural and systemic inequity."

For example, women and white respondents expressed a higher desire for AI notification vs. men and Black respondents, respectively, "suggesting that notification may be necessary but not sufficient," they wrote.

The researchers concluded that the results underline "that notification about AI will be necessary for ethical AI and should be a priority for organizations and policymakers."

"With this signal about the public's preference for notification, the question for health systems and policymakers is not whether to notify patients but when and how," they wrote. "As health systems begin to establish governance for AI tools, multiple approaches to notification will be needed."

They added that such efforts "should support comprehensive, evidence-based programs that promote transparency about AI in health care to ensure trustworthiness of health systems" among patients and experts.