Balancing excitement and caution with AI in healthcare

Steven Ullmann notes that there are grounds for optimism and skepticism with the growing use of artificial intelligence in healthcare.

Steven Ullmann, director of the University of Miami's Center for Health Management and Policy

Ullmann is the director of the University of Miami’s Center for Health Management and Policy. As he notes, there are ongoing debates about how to ensure clinicians are using AI appropriately.

In a recent discussion with Chief Healthcare Executive®, Ullmann explains why he’s excited about AI’s potential to offer better healthcare for patients.

“Accuracy,” Ullmann says. “That's basically a second set of eyes, if you will, in terms of diagnostics. … You have your eyes on things, but you have this extra set of eyes looking at things and identifying things that you may not see, even with the naked eye. So these are things that have become very important in terms of the provision of care.”

“Whether it be in dermatology, from being able to look at skin, at images, basically say this is something that is suspicious in pathology, picking up things in terms of cancerous cells that the AI is able to allow additional information, additional sight on something … this is where I think it's very helpful, and I think that's what I'm most excited about,” he says.

• Read more: Hackensack Meridian CEO says AI could improve health of billions

AI can be a valuable tool in helping clinicians quickly access the latest research. Ullmann notes that keeping up with journals is a monumentally difficult task for doctors and nurses.

“Physicians, nurse practitioners, any individual in the field who is involved in trying to keep up with the healthcare literature, just can't,” he says.

Ullmann cites the challenge of a doctor who has been out of medical school for 15 or 20 years and has a busy practice, and is spending 15 minutes with each patient per day.

“Where do you find the time and keep up with the literature and the advances that are going on? And this is again where artificial intelligence can help, in terms of support for keeping up with the literature, so that when you are dealing with illnesses and an injury and diagnostics, that you have access to newer information that's there to help you make better decisions,” Ullmann says.

Ullmann acknowledges that there are concerns about AI tools that could reflect racial bias in patients from minority groups. Some AI chatbots have provided answers to healthcare questions indicating racial bias, researchers have found. The Coalition to End Racism in Clinical Algorithms has been working with New York health systems to improve racial equity in algorithms.

“AI is only as good as the information going into it and there's a concern amongst minority populations when they utilize healthcare services, including in the hospital setting, that because of some of the history of minority groups … being victimized by some of the healthcare methodologies and technologies from years ago,” Ullmann says.

“There is some weariness of how AI is used with minority populations,” he adds.

• Read more: Health equity ‘is the greatest challenge with AI’

Some Americans have expressed skepticism about the use of AI in healthcare. Six out of ten Americans (60%) said they would be uncomfortable if their doctor used AI in diagnosis of a disease or in developing a treatment plan, according to a survey released in February 2023 by the Pew Research Center. Ullmann says it will take time for some patients to accept the use of AI in their care.

Some AI tools, such as documentation tools that can record doctor-patient conversations, are helping clinicians, Ullmann says. Providers can have more natural conversations with patients, and patients feel like they have the clinicians’ full attention.

“It allows the doctor or the nurse practitioner to speak directly to the patient instead of being at the terminal,” Ullmann says.

As clinicians use AI more often, Ullmann says he has a concern about artificial intelligence being “too good.” AI technologies could detect small abnormalities in tissue that may not normally be identified, and he says that can obviously be a positive.

But he also says that could lead to added tests that may cause additional stress for patients, including biopsies which are not “pain-free.” AI technologies could reduce stress if clinicians can find cancers earlier, but it’s possible that additional testing may ultimately prove unnecessary and uncomfortable for some patients.

“It's a learning curve in terms of all of this,” Ullmann says. “I think it's a learning curve for those who are providers utilizing it, when they are using artificial intelligence to make a decision. Is it giving them the right guidance, or is it leading them down a path that is problematic?”

“The perception is that artificial intelligence will allow for AI to learn from itself and become better over time,” he says.

AI has been touted as being particularly useful in radiology in helping doctors identify diseases. Harvard Medical School researchers, working with researchers at MIT and Stanford, found that some radiologists were helped with AI tools, but the success varied by clinician, according to findings published in March in Nature Medicine.

Ullman says as time passes, patients are likely to have greater comfort and acceptance of AI in healthcare, but he understands why some individuals may have some reservations.

“It’s coming at people so quickly that they don't really know ultimately how it's impacting them,” Ullmann says.

Read more: Microsoft, hospitals hope TRAIN will keep AI in healthcare on track