Patients find AI portal messages as empathetic as human ones. Really?

The sheer volume of patient inquiries through electronic health record (EHR) portals has become a significant burden for clinicians. This constant influx of messages, ranging from simple administrative questions to complex clinical concerns, contributes to burnout and diverts time from direct patient care. Health systems are actively seeking solutions to streamline this communication, and AI-powered tools are emerging as a potential answer.

These AI systems are designed to draft responses to common patient queries, freeing up clinical staff for more involved tasks. The core question, however, is whether patients will accept and trust messages generated by an algorithm as readily as those crafted by a human. This acceptance is not merely a matter of convenience; it touches upon the fundamental patient-provider relationship, which relies heavily on clear, empathetic, and trustworthy communication.

Patient Perceptions of AI-Generated Content

Patients generally exhibit a high degree of comfort with AI-generated messages in their portal. Studies exploring this phenomenon often present patients with a mix of human-written and AI-generated responses to common clinical scenarios, then ask them to rate these messages on various attributes. These attributes typically include perceived empathy, trustworthiness, clarity, helpfulness, and overall satisfaction.

A consistent observation is that patients frequently rate AI-generated messages on par with, or sometimes even slightly higher than, human-written messages for attributes such as clarity and conciseness. This may be due to AI's ability to synthesize information rapidly and present it in a structured, easy-to-understand format, devoid of the potential for human variability in writing style or fatigue.

The perception of empathy is a more complex area. While some patients can discern a difference, many report that AI-generated responses convey sufficient empathy, particularly when the AI is specifically trained on empathetic language patterns. This suggests that the language used, rather than the source, is often the primary driver of perceived empathy for routine communications. Patients are less concerned with the 'who' and more with the 'what' and 'how' of the message content.

The Role of Transparency and Oversight

Transparency about the use of AI is a critical factor influencing patient comfort. When patients are informed that a message may have been drafted or assisted by AI, their acceptance levels remain high, provided they understand that a human clinician still reviews and approves the final communication. This human oversight is essential for patient safety; it acts as a safeguard, reassuring patients that their medical concerns are ultimately handled by a qualified professional.

The types of messages where AI is most readily accepted tend to be those that are informational or administrative in nature. For instance, responses to questions about medication refills, appointment scheduling, or general health information are well-suited for AI generation. More complex clinical advice, particularly concerning new diagnoses, changes in treatment, or sensitive health issues, still requires direct human input and careful review.

The potential for AI to introduce errors or misinterpret patient queries is a valid concern. While AI models are becoming increasingly sophisticated, they are not infallible. A human clinician's final review step is therefore not merely a formality but a necessary component of patient safety and trust. This hybrid approach, where AI assists in drafting and humans provide the final approval, appears to strike the right balance for current applications.

Implications for Clinical Workflow

Integrating AI into patient portal management could significantly alleviate the administrative burden on clinicians. By automating responses to routine inquiries, clinicians can dedicate more time to complex cases, direct patient interactions, and other critical tasks. This shift could improve clinician well-being and potentially enhance the quality of care by allowing more focused attention where it is most needed.

But the implementation of AI must be carefully managed. Training for both clinicians and patients will be necessary to ensure effective adoption. Clinicians need to understand the capabilities and limitations of the AI tools, while patients need to be educated on how these systems work and what to expect. The Oxford Handbook of General Practice provides a useful reference for navigating such changes in primary care settings.

The long-term impact on the patient-provider relationship also warrants ongoing observation. While current data suggests high acceptance, the subtle aspects of human interaction, particularly in emotionally charged situations, may still be irreplaceable. The goal is not to replace human clinicians but to augment their capabilities, allowing them to operate more efficiently and focus on the uniquely human aspects of care.