'Nurses won't document:' Here's how Mayo Clinic plans to use AI to relieve burden

Most health systems are at least thinking about their artificial intelligence strategy — but still treading cautiously. After all, this is a risk averse industry that tends to move slowly. 

The Mayo Clinic is an outlier. And it’s been sharing its strategy more broadly.

This week, for instance, it offered The New York Times a look into how it's using AI in its radiology department. The takeaway? Rather than replacing physicians, it’s seeing improvements in speed and accuracy, and the number of radiologists has actually increased 55%.

“At Mayo, we have long recognized AI as a tool to enhance human expertise,” Gianrico Farrugia, M.D. , Mayo’s president and CEO, wrote in a post on LinkedIn.

Only 40% of health systems have implemented generative AI, according to a McKinsey & Company survey conducted in December. That’s compared with 48% of health insurers and 47% of the industry overall.

They were also far more likely to say that the return on investment has been unclear (31% vs. 21% of insurers). And payers were twice as likely to say they’ve seen substantial ROI, with 8% of that group responding that they’ve seen 4 or more times their investment.

Health system executives have told me in interviews that they’re looking to their clinician workforce to help them figure out what to prioritize, but that also requires healthcare professionals to understand the potential. Getting clinicians up to speed on AI will be a monumental undertaking for health systems and professional associations, many of which are actively trying to figure out how to upskill the workforce at scale.

Executives are also confronting a good deal of skepticism about the technology. Even though clinicians have been warming to AI, as many as 37% of nurses, for instance, said they have concerns about how it will impact patient care, their workload and even their job opportunities, according to a separate survey last year from McKinsey and the American Nurses Foundation . 

Nevertheless, health systems like the Mayo Clinic are forging ahead. The Rochester, Minn.-based network has announced AI partnerships at the past two JPMorgan healthcare conferences — one with startup Cerebras Systems in 2024 and the other with AI juggernaut NVIDIA this year.

Nurses, of course, are the backbone of a health system, and a key part of any AI strategy.  

“The priority for me is looking at how we are creating nurse-informed solutions that make it easier for them to do their work,” said Ryannon Frederick , Mayo’s chief nursing officer, in a recent interview. “Specifically, we are committed to looking at solutions that remove some of the burden from nurses.” 

In our conversation below, Frederick and I talk about the tools that Mayo is building for nurses, its efforts to upskill its workforce on the technology and the skills nurses will need to take a leading role in its adoption.

And tell me: Where do you think AI will be most useful for clinicians? Where are you finding opportunities to expand your skillset?

The transcript below has been edited for length and clarity.

LinkedIn News: Tell me about how nurses are using AI at Mayo.

Frederick: Our approach is for nurses by nurses. About three or four years ago, we sat down with a group of nurses and asked them, ‘Where are the biggest pain points in your day? Where do you need the most help?’ That was the start of our road map and our journey around AI. One of the things that they shared with us is documentation — without a doubt, it is the biggest burden in their day. It's a necessary part of the role, but it's time that they're taking away from their patients in order to complete that record. 

We started on a journey where we wanted to radically change the way documentation occurs. We came out with a bold statement saying that at Mayo Clinic nurses won't document. We wanted to start with challenging the status quo and challenging our boundaries of what is realistic and what isn't realistic.

LinkedIn News: How are you doing that? 

Frederick: We entered into a co-development deal with Abridge and Epic to look at an ambient documentation solution developed for nurses by nurses. When nurses do their assessment with patients, [they] oftentimes will share what they're seeing. We call it ‘narration of nursing out loud.’ Imagine a world where you have AI running in the background that is taking this narrative conversation between a patient and a nurse, and automatically putting that into the documentation for you. It doesn't require them to change their workflow. It can be tailored to the way that they do their patient care and their assessment, and it captures just that conversation between the nurse and the patient in the background.

We always get consent from the patient because that's one thing that's really important. Patients opt in for this. It's not a mandate. Nurses opt in for this. It's not a mandate for them either. And then after the nurse finishes that assessment and that narration of care, they review all of the proposed documentation from the AI for accuracy before it flows finally into the chart. And they also can see the transcript and see where the AI pulled everything from so they can validate accuracy if they have a question.

LinkedIn News: How are you quantifying the results?

Frederick: We have seven units that are using the very first prototype of the tool.

We had a little over 400 nurses that were trained to use it and we've made it 100% voluntary. We want them to use it because they love it, not because they're forced to, and we're seeing about 74% of people using it because they want to. I don't know that I've ever seen that high of an adoption from a technology that was voluntary for nurses.

LinkedIn News: Where else are you using AI?

Frederick: Another example we implemented last year was augmented response technology. Oftentimes nurses will get long messages from patients and they'll spend a long time going back through the medical record and finding the relevant information. What the AI technology does is it helps do that deep dive into the chart. It pulls up the relevant information and proposes a response to the patient. The nurse will review all of that information. They can go back and check the sources to validate accuracy. They tweak and change what they want to and then you can go ahead and send it to the patient. 

We're responding to our patients faster. That's our outcome at this point in time. And nurses are telling us that this technology has been really helpful, especially towards the end of the day when their energy's lower.

LinkedIn News: How are you upskilling your nurses on AI?

Frederick: We did an event in our Department of Nursing called Nursing Nexus, where we brought over 100 people together to explore and imagine new ideas that could make patient care and their role better. We had industry there side by side with nurses hearing what works, what doesn't work, and problem solving together. That's one example of us not just upskilling but making sure nurses are in the center of telling us where this technology could help them.

We also have a robust nursing education team at Mayo Clinic and we have a significant amount of specialty education that we give. Anytime we work on rolling out a new innovation, we're also providing education specifically to help ease fear, to help people know how to utilize the tool, and be very clear about their responsibility with the tool. We want to be very, very clear about that: It does not replace you, it supplements you. You're still responsible for the clinical information education that you're providing to patients. It's really important that we develop that foundation.

LinkedIn News: How do AI responsibilities fit into the leadership team?

Frederick: I created a structure to propel this work forward. We have a CNIO who also works on our clinical system. She plays a dual role for Mayo Clinic, but also functions as my partner. I also designated a nursing leader who reports directly to me, who is responsible for nursing transformation, and that is where, not all, but much of our AI work sits.

And then under that, there’s a vice chair of nursing transformation and a couple of nurse administrators too, who are really leading these projects. We've supplemented their teams not just with nurses but with people who have project management expertise because those skill sets supplement the expertise of nurses in the creation, development and deployment. We also have a nursing optimization team and they are the ones that are trying to work on AI automation technology.

LinkedIn News: How can nurses get more involved in an AI leadership role?

Frederick: A lot of people want to be a part of our transformation team and what people miss about that is this is exciting work, without a doubt, but it's so much more than just knowledge of AI and technology. It's deeply embedded in communication, change management and project leadership. I always encourage nurses to start with getting some leadership experience because you cannot be successful in some of these massive projects if you don't have the leadership acumen as a foundation to build upon that.

As a nurse leader, all of my learnings on AI are self-taught. I don't consider myself to be an informaticist or an expert, but I believe we're leading the way at Mayo Clinic. And the reason I think we're leading the way is that we have the soft skills in the development, the creation, the execution and the implementation, and those things are really what makes the difference.