Most Patients Would Wait Up to Two Days for a Clinician to Check an AI Reply to a Medical Question. For Refills, Most Chose Speed.
A survey of 1,185 U.S. adults in JAMA Network Open found 63.1 percent would wait for a doctor or nurse to review an AI answer to a medical portal message. For scheduling and refills, 57.4 percent chose the five-minute AI reply.
Patients are willing to trade speed for a human check, but only when the question is medical. That is the main finding of a research letter published Oct. 14 in JAMA Network Open, "Public Perceptions of Timeliness for AI Use and Human Oversight," from researchers at Columbia University and Vanderbilt University Medical Center.
For practices deciding where AI-drafted portal replies can go out quickly and where they should wait for review, the survey gives a first read on what patients say they want.
What the survey asked
Participants chose between "a fast answer from the AI (within 5 minutes)" and "waiting until a health professional (doctor or nurse) reviews and edits the AI's answer (up to 48 hours)," for three kinds of portal messages. The researchers analyzed 1,185 U.S. adults recruited through Qualtrics online survey panels between June 17 and July 26, 2024, using quotas balanced on gender, education and generation and a target of 50 percent racial or ethnic minority participants.
What it found
For medical questions, 748 participants (63.1 percent) chose to wait for the reviewed answer. For logistical messages such as scheduling and prescription refills, 680 (57.4 percent) preferred the fast AI answer. For general messages, the split was close to even, with 600 (50.6 percent) choosing speed.
Mental health support followed the same pattern. Asked how likely they were to use each option, 63.9 percent said a licensed human therapist available in three months, 50.2 percent a therapist-monitored AI chatbot available within a month, and 44.3 percent an AI chatbot alone, available immediately.
Even the slower option was an AI draft. Patients were choosing a human check, not a human author.
Who was more open to AI
Black participants, 39.3 percent of the sample, were more likely than White participants to accept a fast AI answer to a medical question (prevalence ratio 1.40). Women were less likely than men to do so (0.76). Participants with higher health literacy were more open to AI for general and medical messages.
Limits
The authors list several: the design confounds response time with who reviews the answer, the scenarios were hypothetical, the sample was a nonprobability online panel, answer options were not randomized, and the data were collected in mid-2024, before more recent changes in how people use AI. The study measured stated preferences, not what patients did with real messages.
What practices can take from it
If you are piloting AI-drafted replies, the data support sorting messages by type: routine logistics are where patients most readily accept speed, and clinical questions are where they expect a clinician's eyes. Tell patients which path their message is on. Our guide to AI inbox triage for patient messages covers the routing, and patients who arrive with chatbot answers covers the other side of the conversation.
Reported, fact-checked and edited by The Script Pad's editorial staff.
This article is for professional education and does not replace clinical judgment. Treatment decisions should be based on the individual patient and current guidelines.
