Patients Prioritize Tone Over AI Origin in Portal Messages

Patients' comfort levels with AI-drafted patient portal messages are primarily influenced by the message's tone, length, and the inclusion of specific details, rather than its origin, according to a small study published this week. Investigators found that patients reported "high comfort" with AI-generated messages as long as these qualitative aspects were satisfactory. The research aimed to understand patient perceptions of AI integration into healthcare communication, a growing area of interest for providers seeking to improve efficiency.

The study, conducted by researchers at an unnamed institution, involved a cohort of patients who were presented with various sample portal messages. These messages were either drafted by human clinicians or generated by an AI system. Participants were then asked to rate their comfort, satisfaction, and perceived quality of the communication. The findings suggest that the perceived empathy, clarity, and comprehensiveness of the message outweigh the knowledge of whether a human or an AI was the author.

This emphasis on message quality over authorship indicates a potential pathway for AI adoption in healthcare settings. Providers may be able to leverage AI tools to draft initial messages, such as appointment reminders, follow-up instructions, or responses to common queries, without significant patient apprehension. However, the study also implicitly highlights the importance of robust AI training and oversight to ensure messages consistently meet patient expectations for tone and detail. The researchers noted that further investigation is needed to explore how different patient demographics and varying levels of AI sophistication might affect these perceptions.

While the study was small, its conclusions offer valuable insights for healthcare organizations considering the implementation of AI in patient communication platforms. The focus remains on delivering clear, empathetic, and informative messages, a goal that AI can potentially help achieve if designed and deployed thoughtfully. The investigators did not name the specific AI model or platform used in their research, nor did they disclose the exact number of participants or the institution where the study was conducted, citing ongoing research and publication plans.