Patients Are Not Fully Sold on AI in Health Care

Patients Are Not Fully Sold on AI in Health Care

Americans will accept artificial intelligence for some routine work in medicine, a new Pew Research Center survey found. But Americans get less comfortable when AI handles medical decisions without a person checking the work, and that leaves hospitals and health tech companies with two things to consider as they try to build comfort with AI. What work does AI do, and who is responsible for it?

Key Points

  • Just 6% feel comfortable letting AI examine medical scans without anyone checking its work.
  • 60% feel comfortable letting AI examine medical scans when a person checks the results, too.
  • About a quarter don’t want AI to have any part in examining medical scans.
  • 47% worry that AI could damage the relationship between patients and doctors.
  • More than half think using AI will put their personal information at greater risk.

Which Medical Jobs Americans Would Let AI Do

Many Americans are fine with AI translating for patients, and they’re also fine with AI booking visits, requesting prescription refills and putting together notes during a doctor’s appointment. They accept those uses whether a person checks the work or AI completes it without that check. But Americans aren’t comfortable giving AI that same freedom for every kind of medical task. Not many Americans are willing to let AI interpret lab results, identify a medical condition or settle an insurance claim without someone checking the work. And 60% are uneasy about doctors using AI to diagnose patients and choose treatments. Help with paperwork is one thing.

Americans can accept that help and still be unsure about letting AI make choices that affect a person’s health.

There’s a reason patients draw that line. Booking an appointment isn’t the same job as figuring out what condition a patient has, even when both jobs use the same technology. For businesses building technology for health care, the survey suggests Americans are willing to give AI a try when it helps with everyday tasks, but Americans are more reluctant when the tools start making judgments about medical care.

Houston Methodist Leaves Decisions to People

At Houston Methodist, AI is there to help doctors get through their work, not to decide things for them, said Michelle Stansbury, the hospital’s associate chief innovation officer and its vice president overseeing IT applications. That includes giving doctors information prompts and helping with other jobs. Stansbury said a person always reviews the output from AI, and Stansbury made the distinction clear: the decisions belong to people. Stansbury described how that works inside Houston Methodist, which is one hospital’s experience alongside the broader worries people expressed in the survey. Houston Methodist has put AI into some patient care tasks, and Houston Methodist is also using it for work behind the scenes that helps the hospital run.

  • Delivering patient histories more quickly so doctors have them sooner.
  • Helping doctors reply faster to questions patients submit through an online portal.
  • Taking notes during visits through ambient listening, which captures the conversation and helps make a record of it.
  • Looking for efficiencies in operating rooms so patients spend less time waiting.
  • Giving doctors information prompts that help them do their work.

Those tools leave doctors with more time for patients, Stansbury said. With ambient listening taking notes, a doctor can pay attention to the patient and the conversation instead of staying at the computer, typing everything that gets said throughout the visit.

More time for doctors to care directly for patients is the main hospital benefit Stansbury described. Stansbury also described using AI to help cut down on the time patients spend waiting. A person reviewing AI’s output addresses one concern people raised in the survey, but the broader doubts about AI in diagnosis and treatment suggest that people have plenty of reservations left.

Views Vary, and Doctors Earn the Most Trust

Some patients have already tried AI by the time they arrive for a medical visit. Stansbury said doctors are seeing more patients who turn to ChatGPT or Claude and try to figure out what condition they might have before they seek care.

People are nearly evenly divided over what AI might mean for health outcomes. Thirty-seven percent think outcomes will improve, and thirty-six percent think outcomes will worsen. About a quarter think there will be little difference.

Those expectations also vary by age. Among adults ages 18-29, forty-five percent think outcomes will worsen, up from 29% when people were asked four years earlier. Among adults age 65 or above, 42% expect an improvement, up from 34% when asked four years earlier. The youngest adults are more pessimistic than every other age group. Americans tend not to trust institutions such as government and health insurers to make patients the priority when those institutions use AI. They are divided about whether hospital leaders will make patients the priority, and doctors come out ahead on trust.

Sixty percent have at least some confidence that doctors will use AI in ways that benefit patients.

Hospitals are starting to use AI, and those different opinions are something hospitals have to weigh. Giving patients more time with doctors could strengthen a relationship that patients already trust, but worries about privacy and care feeling less personal are still there. The approach at Houston Methodist is one example of how patients could see that benefit. It does not show whether patients actually will.

Subscribe to our weekly news digest.

Join now and become a part of our fast-growing community.

Invalid Email Address
Thanks for Subscribing!
We'll be sending you our best soon!
Something went wrong, please try again later