The rise of patient messages sent to clinicians through patient portals has led to physician burnout and dissatisfaction, driving the adoption of artificial intelligence (AI) to alleviate this burden. As AI technology advances, it is increasingly being integrated into electronic health record (EHR) systems to automate responses to patient inquiries. This integration has the potential to transform patient communication by reducing the workload on healthcare providers and improving the efficiency and quality of patient care.
A recent study published in JAMA Network Open investigated patient preferences for the use of AI in electronic messages. The study found that participants preferred AI-drafted responses over human-drafted responses. Participants indicated higher satisfaction, perceived usefulness, and feeling cared for when messages were AI-generated. This preference for AI-drafted messages may be attributed to their tendency to be longer, more detailed, and seemingly more empathetic than human-drafted messages.
While patients preferred AI-drafted messages, their satisfaction decreased when they were informed that AI authored the message. When AI authorship was disclosed, satisfaction was lower compared to messages with human authorship or no disclosure. This suggests that patients have a preference for messages they believe are written by their clinicians.
“Our study shows us that patients have a slight preference for messages written by AI, even though they are slightly less satisfied when the disclosure informs them that AI was involved,” investigator Joanna S. Cavalier, M.D., from the Duke University School of Medicine in Durham, said in a statement.
Human disclosure vs. AI disclosure
The study also explored the impact of disclosure on patient satisfaction. Participants preferred messages with a human disclosure or no disclosure compared to an AI disclosure. There was no significant difference in preferences between human disclosure and no disclosure, which may indicate that patients assume a human author unless explicitly told otherwise.
Interestingly, there was no interaction between author and disclosure. Regardless of the actual author, patients were more satisfied when the disclosure indicated a human response or when there was no disclosure. This suggests that patients do not exhibit automation bias, which is the tendency to favor recommendations from automated systems. Instead, the findings indicate a reverse bias in medicine, where patients prefer messages they believe come from their clinicians.
The seriousness of the topic discussed in the messages did not influence patient preferences regarding disclosure. Patients did not show nuanced preferences for AI use based on the context of the message. This may be because AI is still in the early stages of adoption in healthcare, and patients have not yet developed specific preferences for its use in different situations.
How can providers adapt to artificial intelligence?
As AI continues to be implemented in healthcare, it is crucial to consider the ethical implications and patient preferences. National guidelines emphasize the importance of disclosure, stating that patients should be informed when an automated system is used. While disclosure may slightly reduce patient satisfaction, it is essential for upholding patient autonomy and empowerment.
The study also examined preferred AI disclosure statements. Participants preferred the shortest disclosure statement, which acknowledged AI assistance in composing the message. This preference for brevity aligns with the idea that overly complicated disclosures may overwhelm patients.
AI has the potential to significantly shape patient communications by improving efficiency and reducing clinician burnout. While patients express a preference for AI-drafted messages, they also exhibit higher satisfaction when messages are attributed to their clinician or when there is no disclosure. Therefore, healthcare systems must carefully consider the ethical and operational implications of AI implementation, balancing the benefits of automation with the importance of patient autonomy and trust.