Podcast: Assessing the impact of AI in medicine

Ainsley MacLean, MD, and Brian Hoberman, MD, discuss how emerging technology can support quality care

The use of AI in medicine has gone from future fantasy to present reality in just a few short years. From predictive analytics to ambient listening, AI has the potential to revolutionize health care. But what is the best way to integrate this technology into supporting the day-to-day work of physicians while maintaining a human touch?

Potential applications for AI are vast and growing. The technology can be used for patient algorithms, speech-to-text, diagnosis, administrative workflow, and much more. However, expert guidance is needed to help ensure this potential is harnessed in a way that’s responsible and resourceful.

Chris Grant, chief operating officer of The Permanente Federation, Kaiser Permanente, and Permanente Medicine Podcast host, welcomes 2 Permanente physician experts on this topic to dissect the emergence of AI and how Kaiser Permanente can use this technology to augment its approach to high-quality, patient-centered care.

Brian Hoberman, MD, is executive vice president and chief information officer for The Permanente Federation and chief information officer with The Permanente Medical Group. Joining him is Ainsley MacLean, MD, chief medical information officer and chief AI officer for Mid-Atlantic Permanente Medical Group.

Related health care innovation story: Ainsley MacLean, MD: How AI powers health care innovation, enhances care

In this episode, Drs. Hoberman and MacLean share insights on how AI can play a supportive role for inpatient care and enhance medical practices. They also examine how innovation in this space can assist in addressing current concerns around physician and clinical burnout.

This AI wave is putting information to work to decrease burden on physicians.

– Brian Hoberman, MD

“It’s putting information to work to decrease burden [on physicians and other clinicians],” says Dr. Hoberman. “That spreads happiness all over the place, which wouldn’t have been possible without this AI wave.”

Drs. Hoberman and MacLean also address concerns around the increased use of AI in medicine, such as accounting for security vulnerabilities and maintaining a personal touch in patient care. Additionally, they touch on the importance of developing health care policies around integrating AI into the daily work of physicians and care teams.

AI will never replace the expertise and compassion of an individual physician.

– Ainsley MacLean, MD

“There is a demand for this technology, but it has to be done in a way that protects privacy and security,” says Dr. Hoberman, referring to setting standards around the adoption of new technology.

Related story: Brian Hoberman, MD, Appointed to National Permanente Leadership Team as Technology Chief

In this episode

2:22 Get to know Dr. MacLean

4:18 Learning more about Dr. Hoberman

6:33 Key areas where AI is currently used at Kaiser Permanente

9:45 Using technology to address physician and clinician burnout

13:22 Setting standards for patient security and privacy

15:37 Promoting physician education

17:41 Maintaining a personal touch in patient care

19:32 Developing best practices and policies

21:57 Predictions for the future use of AI in medicine

Connect with Chris Grant by following him on X (formerly Twitter) at @cmgrant or LinkedIn.

Follow us: Subscribe to the Permanente Medicine Podcast on your favorite streaming platform.

Podcast transcript

Transcript is autogenerated. Although edited for clarity, it should not be considered an exact replication of the podcast and may also be updated as needed.

Brian Hoberman, MD (00:03):

This isn’t a theoretical wish. This is a reality where many tens of thousands of times a week, doctors are using this technology in their encounters.

Ainsley MacLean, MD (00:14):

That AI assistant will start to sound more and more like their doctor, but it’s always going to be an assistant. It’s absolutely never going to replace the expertise and the compassion of that individual physician.

Chris Grant (00:31):

Welcome to the Permanente Medicine Podcast. I’m your host, Chris Grant, chief operating officer of The Permanente Federation. In this episode, we’re going to explore the hot topic of artificial intelligence and its place in medicine. Joining us are two guests leading a number of key discussions around this issue. On behalf of the Permanente Medical Groups, Dr. Ainsley McLean and Dr. Brian Hoberman. Dr. McLean is chief medical information officer and chief AI officer for the Mid-Atlantic Permanente Medical Group. She’s also their associate medical director of Diagnostics and Interventional Radiology, overseeing more than 800,000 imaging examinations and procedures annually. Also on the call is Dr. Hoberman, who serves as executive vice president and chief information and national information technology leader for The Permanente Federation. He’s also chief Information officer of The Permanente Medical Group, helping to direct the information technology used by more than 10,000 Permanente physicians and 45,000 staff throughout Kaiser Permanente in Northern California. Welcome to the show, Ainsley and Brian, I’m so glad to have you both here to share your expertise on a topic that’s been front and center in health care recently.

Ainsley MacLean, MD (02:00):

Thank so much for having us, Chris.

Roles within the Permanente Medical Groups

Chris Grant (02:02):

Alright, well let’s have some fun and dive in. Before we get into the meaty subject of AI, I’d love just to hear from both of you and maybe share a little bit about yourself. Ainsley, could you speak to your role in the Mid-Atlantic Permanente Medical Group and maybe share something about what you enjoy doing outside of medicine?

Ainsley MacLean, MD (02:22):

Of course, Chris, first of all, thank you so much for having me. It’s such a pleasure to be on this podcast. I’m Ainsley McClain, I am a neuroradiologist by training, so I read imaging of the brain and the spine. I like to say that I come to the practice of technology and AI in medicine through both the art and the science realm. I was an undergraduate major in the visual arts, but then through my discipline within radiology and then a career of about 12 years with Kaiser Permanente in the Mid-Atlantic, I’ve increasingly grown more and more fascinated and passionate about the role that technology can really play in implementing large amounts of change at a very broad scale. And over the years I’ve had the pleasure of working with Brian and the CMIOs across all of the Permanente regions to implement a wide range of projects, and looking forward to talking specifically about AI. In terms of what I enjoy doing in my free time, I guess I definitely need an outlet. I love running, I play pickleball, and most of the time just spend driving my kids around to their various activities.

Chris Grant (03:28):

Sounds like a busy life. Just one follow-up question. Tell me about the visual arts undergraduate, what area of visual arts?

Ainsley MacLean, MD (03:37):

I did my senior thesis in photography, and I think that’s sort of how I came to radiology. I’m really fascinated by just how we depict the human body and how medicine really plays a significant role in our understanding of the human body, specifically within radiology. And I think AI has really become a natural extension of that and it’s been so powerful with generative AI, both in the arts but also within medicine and science as well.

Chris Grant (04:05):

I love it. Thank you, Ainsley. And Brian, same question for you. I’ve gotten to know you through our work together at the Federation, but what should our audience know about your professional area of focus and some of your personal interests?

Brian Hoberman, MD (04:17):

Sure. Thanks Chris and thanks for having us for the podcast. I’ve been with the Permanente Medical Group since the early nineties. I started as one of the founding hospitalists when the hospitalist program came online. I worked in a hospital setting with very little technology and was attracted to trying to solve patient safety problems in the hospital and felt that using technology would help us get there. And I was there just as the wave of EMRs was starting to build a little bit and through the luck of being in the right place at the right time, was able to catch that wave and lead the deployment of inpatient HealthConnect, which is Epic, across all of our hospitals in Northern California and collaborate with the other hospital leaders in our other regions as we did that. And I’ve remained with HealthConnect and technology ever since. Right now, we’re catching another wave and it’s this wave of what we can do with artificial intelligence that builds on what we accomplished in the first wave where we have all this massive amount of digital data. I guess I’ve been surfing the opportunities that are available to us in health care as technology got applied.

Chris Grant (05:36):

How about outside of work? What are a couple of your favorite passions?

Brian Hoberman, MD (05:40):

I hate my favorite passion, which is working out because the motto is get comfortable being uncomfortable. And I get very uncomfortable because the workouts are super hard, typically about three times a week, but that’s a major focus of my time, just trying to stay in shape. I have a dog who’s lying on the floor behind me here who’s very demanding and may choose to come on audio at any time. She keeps me pretty busy and other fun at-home stuff. Unlike Ainsley, my kids are grown and out of the house and fully launched and at this point actually living thousands of miles away.

Key areas where AI can support patient care

Chris Grant (06:14):

Well, I’m going to build on Brian’s metaphor here and we’re going to ride this wave forward and shift to the topic of the day: artificial intelligence and medicine. At Kaiser Permanente, there’s been a growing focus on using AI to support patient care and to enhance the capabilities of physicians. Ainsley, I’m going to turn to you and ask that you speak to some of the key areas where this is currently taking place within our health system.

Ainsley MacLean, MD (06:42):

I love the wave analogy and I think to kind of follow along and get back to your question, Chris, is first of all, the current wave in AI and being successful at it requires many years preparation for where we are today. And I think that [at] Kaiser Permanente, we have done a lot over the last several years to get us into the position in which we can really fully leverage generative ai. And I think that involves not only being part of a large national collaborative organization where you have physician leaders who understand the operational problems and what the solutions are, but also a close collaboration with our information technology partners. And then the ability to implement both cloud-based technologies and also AI-driven technologies at a large national scale. But of course, as Brian alluded to, we also have the data, which is absolutely crucial.

(07:37):

So, I think we’re very well positioned. To that end, we’ve been very strategic within Permanente [Medical Groups] about the implementation of AI. And early on, about a year ago when ChatGPT first came out, I think the first thing we all did was get together and say, where do we need this? How can this help us? Everyone recognizes the tremendous opportunity, but then what do you do with it and what do you do with it reasonably? One is really in the way that we can support our physicians in their day-to-day practice. And I think for Brian and myself, that’s a guiding beacon for us. How do we make our doctors’ lives better? We know that with the EMR there was so much potential unleashed, and yet with it came a lot of excessive documentation. And so, I think we all see AI as a way to continue to document with very high quality, but in a way that doesn’t burn out physicians.

(08:30):

And so, we’re looking in the realm of ambient listening technology, clinical decision support for radiologists as assistants in the interpretation of breast screening studies, for our pathologists as assistants in the diagnosis of breast cancer, and many examples beyond that. An additional bucket of care is how do we support our physicians and our staff in day-to-day work? We’re looking on the backend use of AI and streamlining workflows, whether it’s generating a PowerPoint, or you’re writing an email to a colleague, or you need to summarize email. It’s a whole spectrum, but it’s important that you’re thoughtful and strategic about the way in which you implement it, especially because we know that there are dollars that are going to be allocated and invested and we really have to make sure we’re doing that in a way where we’re getting the biggest bang for our buck.

Innovating to reduce physician burnout

Chris Grant (09:21):

Makes a ton of sense, Ainsley, and I think some folks’ fear is that there’s another technology that gets put in between the patient and the doctor, but actually AI, as you describe it, offers the promise of enabling the physician to spend more time to be more directly engaged with the patient without necessarily needing to be turning to a computer screen or to type in documentation. Brian, I know in your national leadership role there is a focus on how high level of AI integration is used to address concerns around physician and clinician burnout. And Ainsley just touched upon that. Could you speak to the importance of these innovations in relation to reducing burnout and the assistive role that they can play?

Brian Hoberman, MD (10:06):

Sure. This is actually my favorite thing about the AI opportunity at the moment because there are tasks that are not high value-added, but they take a lot of time. One of them is writing a note. And while there are certain parts of the note that will remain in the physician’s hands, there’s a lot of recording for the documentation that today just requires a lot of typing. The patient said this, the patient said that, on exam I found this, on exam I found that. Ambient listening technology is one of those technologies that approaches magical in terms of what it’s capable of doing, because if you’ve seen this in action, when the ambient listening tech is applied to a visit between a doctor and a patient, it frees the doctor from having to make notes about what the patient said. The ambient listening device is not only going to capture what the patient said, it’s going to capture all of it, which the doctor might miss, and it’s going to synthesize it into a pretty succinct professional medical note that captures the important points and the important meaning.

(11:13):

And if the doctor then says out loud, here’s what I’m finding when I examine you, or even going further, here’s what I recommend that we do, here’s what I’m thinking is going on, the ambient technology will synthesize that into a medical note that substantially reduces the amount of time the doctor has to take to create a similar note. And in Northern California, The Permanente Medical Group did a pilot with one ambient listing vendor that is active right now. And what we’re seeing is that this isn’t a theoretical wish. This is a reality where many tens of thousands of times a week doctors are using this technology in their encounters. Patients like it because the patients notice that the doctor’s paying more attention to them. I once saw a little girl draw a picture of her visit with the doctor and in the picture the doctor is behind the computer screen. That’s how she perceived the doctor experience.

(12:10):

Now the doctor can get out from behind that computer screen and be face-to-face with the patient and their family and really engage because that burden of drafting that note so I can get out on time or taking notes and then having to come back and spend more time later, that burden is relieved through the magic of this technology. What it does is it achieves the highest standard of what I’ve always attempted to do with technology and health care, which is make us better doctors, empower us. When we developed the EMR, it was through creating information and transmitting information. And now with this new wave, it’s putting information to work to decrease burden and that just spreads happiness all over the place and wouldn’t have been possible without this AI wave.

Protecting patient information

Chris Grant (12:55):

So, one follow-up question. Some of our listeners may be wondering about potential security vulnerabilities that come with relying on AI to perform the type of tasks that you described. What are some of the steps we’re taking to make sure patient information remains safe and secure?

Brian Hoberman, MD (13:13):

Yeah, safety, privacy, equity, these things are just table stakes. You have to have addressed those things. And in the case of quality, we’ve done formal reviews where we compare the synthetic note with the transcript, and we’ve rated it on about eight different dimensions of quality and synthetic notes are doing very well. And in terms of privacy, we apply the same standards of privacy protections to the data as we do to the EMR data itself. It’s as protected and encrypted as it possibly can be. And in terms of equity at this point, we haven’t specifically looked at this question of does [the technology] have any intrinsic biases, because the scale of the number of patients is not there yet, but we have plans to constantly monitor for that. And so far, we’re not seeing any indication of concerns from our clinicians — who of course read the synthetic notes — that any such thing exists, except we actually did come across one that we gave feedback about.

(14:25):

It was very interesting that when a male parent brought a child into pediatrics office, there was a period when the system assumed that the parent was a female. And that has since been addressed. So, we’re constantly on the lookout for any of those types of things and will remain so. Bias and quality issues that we’re facing, this is computing we’re talking about. It’s based on software and software constantly evolves, so we never ever stop looking for these things. And a quality program that addresses all of those dimensions is just part of the table stakes of using technology in any setting but includes certainly using AI technology like this in a care setting.

Clinical guidance for evolving technology

Chris Grant (15:09):

Makes a ton of sense. Ainsley, I love your background and you’ve spoken directly to how Permanente physicians and researchers have been long leaders in adopting new technologies and innovations. What do you see as the role of physicians and researchers in the development and use of AI technologies and programs?

Ainsley MacLean, MD (15:30):

Absolutely critical. And I love this table stakes concept. I would add to that physician education in AI. And we’ve really promoted that through a series of internal lectures, specifically speaking with leading companies in AI and also encouraging board certification in AI or other educational courses for our physicians. And [I] just have been so impressed by how Permanente doctors have really leaned into this moment. I love the word assistant and as a radiologist, our radiologists in the Mid-Atlantic region have been using AI as an assistant in our radiology reports for some time now. And I think that when we look at it, and we think of it that way, who doesn’t want an assistant helping them with their work? And so, if we can do that through technology, it’s a win-win for all. I think an important bucket that Brian and I haven’t touched on yet is that of the in-basket.

(16:25):

We know that so much time for our physicians is spent in the office speaking with patients. And that tends to be incredibly rewarding, having that one-on-one time with the patient to guide them through their care. But we know that the care doesn’t continue, right? And in fact, it continues 24/7 through the weekends because our patients have lots of questions for us and really rely on us as doctors. And so, we’re now nationally also looking at a pilot in which we’re assessing generated draft responses to the messages that our patients send us. And so that gets back to what you were sort of alluding to, Chris, which is this concept of the physician as the leader, as the educator. And so, our doctors and our technology teams right now are essentially doing prompt engineering sort of these responses because when it comes out of a box, it may not be the right answer for what a Permanente physician would say to a patient.

(17:19):

And we have so many streamlined processes within our systems that allow us to provide this incredibly complex care virtually or digitally. So, if a patient writes in with a request for something, we can very quickly have their medication mailed to them through a series of smart phrases. And this requires someone with an in-depth knowledge of our care delivery, and it has to be personalized as well. And so, the portion I mentioned with our radiologists, the generated response is actually based on that individual radiologist’s last 10,000 reports, what they said in that particular case. We’re not doing that quite yet in the generated in-basket responses, but my hope will be over time, that’s where we will get, where that AI assistant will start to sound more and more like their doctor, but it’s always going to be an assistant. It’s absolutely never going to replace the expertise and the compassion of that individual physician. It’s really just augmenting, to use the word of the American Medical Association, augmenting that expertise of that doctor.

Creating policies for AI at work

Chris Grant (18:23):

The picture that both of you are painting for me is one of greater engagement for the physician and the patient, and perhaps reducing pretty meaningfully some of the administrative burdens that get in the way of a physician’s workday. With the emergence of AI tools like ChatGPT, [which has] been on the headline news for the last many months, one challenge many organizations face right now is developing policies around how employees use these resources at work. I have friends that lead financial services, consulting firms, telecom, and every one of them is focused right now on how we apply certain policies within human resources or within my organization, and health care is no exception. Brian, in your national technology leadership role, I know you’re taking a close look at this dynamic. Could you speak to how health care systems and Kaiser Permanente in particular is developing policies that address the use of AI and daily work?

Brian Hoberman, MD (19:28):

This ultimately comes back to a privacy question, Chris. The information that is entered into a public tool for using large language models such as OpenAI, it’s like a Google search you do, it actually belongs to Google. Corporate information is private, and then patient care information has even higher standards of privacy. So, the idea of using any kind of a tool like that that doesn’t protect that level of privacy for patient information or corporate information is, well, don’t do that. That’s a really bad idea. And so, what we had was the excitement of, wow, look what these tools can do, against the concern that on the one hand, yeah, you’re decreasing your administrative burden or getting a lot of work done, but at the cost of something that is unacceptable. And so, policy had to stop it and word got out, that’s for sure. And we’re not the only organization that said, please don’t do that. But please don’t do that wasn’t the end of the story.

(20:27):

That was just that moment in time because what leaders needed to do was find a way to do it safely, because there is a demand for it and it does do useful things, but it has to do it in a way that protects privacy and security. At this very moment in time when we’re recording this, that issue is not fully resolved, but people are actively looking at methodologies that would address the privacy and security concerns and at the same time allow folks to use these things to do their jobs. So, what you’ll hear from folks like Ainsley and myself, is this is coming online or that’s coming online when they’re mature enough for us to be able to say, this is ready for use, this is safe for use. But until we’ve achieved that base table stakes of protecting privacy security, ensuring the lack of bias, and ensuring quality, these tools are going to have to be evaluated and then released in a way that corresponds with our own internal standards of what’s acceptable. It’s not to say we don’t want to adopt the technology; we do, we just want to do it in a way that is safe.

Predicting new frontiers for AI in medicine

Chris Grant (21:47):

I love it. It’s taking a deliberate and measured approach to the implementation and adoption of emerging technology. I’d like to wrap with a question for both of you about the future of artificial intelligence. You each have a direct line of sight into the discussions currently taking place in areas of focus for AI. So, here’s where we take out our crystal ball, and if you had just one prediction about the future and the new frontier for the use of AI in medicine, what would it be? Ainsley, why don’t we start with you?

Ainsley MacLean, MD (22:24):

I’ll just pick one use case. A woman goes in to have a screening mammogram. She has an imaging done on a 3D [mammography] machine interpreted by artificial intelligence and a world-class Permanente radiologist. And the AI basically is an extra reader to that radiologist. It then goes immediately to her kp.org app where she gets the result and either can go on and enjoy her day or may have an additional imaging that’s required. If she requires a biopsy, a pathologist will interpret that study the same day [or] next day. Using AI as an extra reader to increase the sensitivity of that study, a diagnosis of breast cancer or no breast cancer can be made. And I can tell you, Chris, that 90% of that is happening today in the Mid-Atlantic, and AI will just be that extra icing on the cake to provide incredible high quality and service to patients, not only in the realm of breast cancer but in every disease pathology. And I’m really, really looking forward to that.

Chris Grant (23:26):

What a fantastic example and something that I think everybody can relate to. Brian, I’m interested in your thoughts, in your crystal ball. Tell us what the future of AI looks like.

Brian Hoberman, MD (23:39):

I don’t have a crystal ball, but I did live through the age of cellphones coming online and then smartphones coming online. And I’ll be honest with you guys, I didn’t see either one of them coming. I was one of those Blackberry people who said, why would I want to give up my Blackberry? So, it’s really hard to predict other than to say that a decade from now we’re going to find it indispensable and it’s just going to become part of how we operate in countless respects and it’s going to create conveniences like our smartphones have created for us that we will ultimately take for granted. I think that what’s fun is going to be helping to shepherd and curate this era in. But what I find really fun is watching how many super smart, super creative people are recognizing what this wave of technology potentially can offer.

(24:35):

And so, it’s a little bit maybe like what’s happened with the app stores, where this thing came out with no apps and now there’s too many to possibly be able to understand or imagine. I think we’re going to see countless applications of AI. And what that means is that some of them aren’t going to be meeting our standards. And so, we’re going to need to help influence creative people to do wonderful things, and we’re going to need to be super vigilant about protecting our organization, our patients, and the industry from the potential misuse of AI. And that means we’re going to be paying a lot of attention over the next many years, but I think ultimately it will turn out to be incredibly useful for patients, providers, and in countless applications in the rest of the economy as well.

Chris Grant (25:22):

Well, you both have gotten me very excited about the future and I was already quite optimistic and excited, and I agree. I think it’s going to allow for greater accuracy, faster responses, and perhaps even more importantly, greater human interaction and the ability to engage on a personal level with more time, increased compassion, and meeting patient’s needs. We could truly devote a full season to this topic, but I’m positive, absolutely positive we’re going to have many more conversations together in the future, and I look forward to that. Thank you so much for joining, Ainsley and Brian. It’s been an absolute pleasure to spend some time with you.

Ainsley MacLean, MD:

Thanks for having us, Chris.

Brian Hoberman, MD (26:04):

Thanks very much, Chris. It’s always fun to talk to Ainsley.

Chris Grant (26:08):

It is. It’s fun to talk with both of you. And thank you once again to our listeners for tuning in. If you’re a new listener and enjoyed this episode, please take a moment to subscribe or share a review. We’ll see you next time.