Patients prefer AI-generated portal messages — until they learn the source

© Rudzhan - stock.adobe.com

Patients generally prefer electronic clinician messages generated by artificial intelligence (AI) over those written by humans, but report greater satisfaction when the responses are attributed to a clinician rather than AI, according to a new study published in JAMA Network Open.

The study, conducted within the Duke University Health System’s patient advisory committee, surveyed 1,455 patients to gauge their preferences regarding AI-authored responses to patient portal messages. Researchers examined satisfaction, perceived usefulness, and how cared-for patients felt based on whether a response was written by AI or a human and whether AI involvement was disclosed.

AI outperforms humans — with a catch

The results indicate that participants rated AI-generated responses slightly higher in satisfaction, usefulness, and the perception of being cared for, compared to human-written responses. However, when participants were explicitly told that a response was AI-generated, there was a slight decline in satisfaction.

In contrast, patients reported greater satisfaction when a response was disclosed as human-written or when no disclosure was made at all.

Regardless of the message’s author or level of disclosure, over 75% of respondents reported satisfaction with the responses they received.

The study’s findings suggest a nuanced view of AI in health care communications: while AI-generated responses are generally well-received, transparency about their use remains a significant hurdle and an ethical dilemma.

Balancing ethics and patient autonomy

The ethical implications of AI disclosure remain a subject of debate. Current AI governance guidelines, including the White House’s Blueprint for an AI Bill of Rights, advocate for transparency in AI-assisted interactions.

However, it can be argued that while transparency is important, excessive technical disclosures may not benefit patients and could potentially lead to unnecessary concerns.

One proposed approach is to use brief and clear disclosure statements, like, “This message was written by Dr. T with the support of automated tools,” which was the most preferred (33%) phrasing among study participants.

With increased usage of patient portals for clinical communication, AI-generated responses could help reduce physician workload and burnout, particularly in handling administrative and routine inquiries. However, practices should carefully consider how they communicate AI’s role in patient interactions to maintain trust and uphold ethical standards.

Researchers suggest further investigation into how AI preferences evolve over time and whether patient comfort with AI-driven communication is at all dependent on varying clinical scenarios. Regardless, the study is representative of the balance health care providers must maintain: leveraging AI to enhance efficiency while ensuring patient autonomy and trust remain at the forefront of digital health advancements.