Health Information & AI with Dr. Ronda Chakolis-Hassan & Dr. Robin Austin

Introduction and Welcome

Stanton Shanedling: Hello, everybody! Welcome to Health Chatter. Our show today is another show on AI, but we're really going to link it today with health information and how people are going to be getting health information through the AI mechanisms. We've got two great guests who have been on our show before, and we'll get to them in a minute. We've got a great crew, as always: Maddy Levine-Wolf, Erin Collins, Deondra Howard, and Ariana Tordoff, who is new to our group, so welcome, Ariana. Matthew Campbell is our production person; he does our recording and gets the shows out to you, the listening audience. Sheridan Nygard does our transcriptions and marketing, and of course, there's Clarence.

Clarence and I are the dynamic duo that do these shows, and we have a third person, Dr. Barry Baines, who helps us with a medical perspective. He kind of gives us the medical twist, while Clarence and I provide the public health twist based on our backgrounds.

Clarence: And Sam, that makes us the Three Musketeers.

Stanton Shanedling: Three Musketeers, right? Yeah. And we all have beards. The listener, you can't see us, but we all have these matching beards, you know? So, anyway, thank you to everybody. Thank you also to our sponsor, Hueman Partnership. Great community health organization. Check them out. They do really creative things, including one of the things we're going to be talking about today. Check them out at huemanpartnershipalliance.org. Check us out, Health Chatter, at healthchatterpodcast.org.

So, welcome back to our two guests today, Dr. Robin Austin and Dr. Rhonda Chakolis Hassan. Welcome to both of you again. You bring some wonderful talent to our show. Dr. Austin brings more than 20 years of clinical experience spanning pain management, women's health, and integrative health. She has dual training as a chiropractor and nurse, and she has done a lot of research in creative areas that focus on chronic pain and menopause. She is a research mentee with the Menopause Society, as well as a fellow of the American Academy of Nursing and the American Medical Informatics Association. Lots of things, for sure. And she's got a great smile.

Dr. Chakolis Hassan is an alumna of the University of Minnesota College of Pharmacy and has been on our show talking about medication and medication management. She has a diverse career that spans 15 years in pharmacy benefit management and over 5 years as a medication therapy management pharmacist. She's a frontline pharmacist, president of the Minnesota Board of Pharmacy, and a nationally recognized public health leader. Welcome to you, and really thank you, too, because she just got back from Japan. Hopefully… I told her, don't speak Japanese, because our listening audience won't understand it.

Okay, so today we're going to be talking about health information and AI through a really creative program. Clarence, I'll let you start it out, okay?

Introduction to the TRIUMP Program

clarence: Yeah. So, we had this opportunity to work with the University of Minnesota around this program that's called TRIUMP. I think that it's around training nursing students in AI. And so we've been pretty excited about it, because we recognize that AI is here. You know, it's been here a lot longer than many of us have thought about it, and so it was exciting to know that there are also some opportunities for people who are nurses who would want to get involved in AI. That's what we're going to be talking about today. I know that Dr. Rhonda will share some of her experiences as well, but I thought we would let Dr. Robin talk about TRIUMP first, and then we're going to ask you a lot of questions, okay?

Robin R Austin: Yeah, great, thank you. So, welcome, thank you for letting me be here again as a guest. This is super exciting. I'm really excited about this topic as well, super passionate about it.

I'll share a little bit about TRIUMP itself. It is a grant where my colleague, Dr. Priya Rajamani, who is faculty also at the School of Nursing at the University of Minnesota, is one of the PIs. It's a joint grant between the School of Public Health and the School of Nursing, so it's truly an interprofessional grant. We have an opportunity to train individuals on population health, but also informatics, AI, and all the things around data. This grant was coming from the end of COVID dollars, and we realized how unprepared we were to marry public health and health information technology.

Really, data does not flow very easily back and forth from public health to hospital systems or any systems; they don't share data very well. And so this grant is really an opportunity to increase the workforce and truly give the skill sets needed around data and AI. We understand that there's a major gap in our health system, and so that's kind of where the grant stems from. Now we can't talk about data or health information without talking about AI, and we'll get more into that, I'm sure, throughout this podcast, but that's kind of the crux of where the grant dollars came from: workforce development, really wanting to upskill our current workforce, marrying both nursing, healthcare, and public health. That's where it came from.

AI In Everyday Life and Defining the Technology

Stanton Shanedling: So, you know, it's interesting, before the show started here, I asked everybody who's on the show here with us what kind of AI they use. And you'd be surprised, for a lot of people, like, for instance, spellcheck just as an AI-generated thing. ChatGPT, there's a whole bunch of venues. So, why don't we talk about the venues? For somebody who isn't really seasoned with using AI, what should we be connected with as far as AI opportunities?

Robin R Austin: Yeah, great question. I think AI is here more than we realize. When we look at selecting Netflix, it knows what shows I like now, and it will recommend specific shows based on what I've already liked. Even my Amazon app, I'm very connected with what I typically buy on Amazon. But I think, like spellcheck? Sometimes thank goodness for it, because I'm not the greatest speller, and so I think that's another way that it's already infiltrated within our daily lives.

But I think of the other bigger things, what we call a large language model, or ChatGPT, Gemini, Copilot, Claude. There's many more that I think are coming out more and more. There are free versions of a lot of these, but you also could pay for some of the versions as well. Even now, if I put in something into what I think I'm putting into Google, it's already using AI.

I think of AI as—I'll just kind of go back and define it—computer systems that basically are trying to do what humans typically have done before, which is finding patterns in data and making recommendations based on predictions. Also, taking text and being able to use it to identify more patterns within text than we've been able to do before in the past. I think it has more robust power with much more data than we've ever been able to use. There's many ways of looking at AI, and it's a huge umbrella. It's kind of picking what you would like within AI, whether it's more of a large language model or looking at algorithm development kind of things, like the Amazon or Netflix kind of idea. Then healthcare is a whole other ball game when we think of how we're applying AI to some of these. So, I'll stop there, because I know Dr. Chakolis Hassan—

Stanton Shanedling: Let me ask you something. It seems like we, the public, have been kind of thrust into this. It's just like, whoa! It's one thing, which we'll get into here, and Rhonda, I'll pull you in here—it's one thing training students who have professions that are going to be using this. It's another thing for just the population. It's just like, whoa! So, okay, Rhonda, chime in on this, because I know you've got great ideas.

Ronda Marie Chakolis-Hassan: I do, and first of all, I also have to be fully transparent. Again, we always know the views and opinions that I'm going to talk about don't represent any of those affiliates. But I have to say, I am a direct recipient of the TRIUMP training, and it's not just for nurses—it's for anybody who is looking to expand their skill set in data analytics. One thing about it is it's really helped me understand the roles of engagement and some of those language terms that we're going to be talking about on the show.

But I want to take you back, and I'm definitely going to date myself in terms of... we weren't necessarily thrown into this. The wave of this was paved by Ask Jeeves, which started about in 1996. Ask Jeeves was very different than some of the other things that were on the computer. Again, that was just right around the time when I was completing my undergrad, and so with Jeeves, you could actually put in a question and say something like, "Where could I find this?" and it would kick it back. I think what happened is we weren't necessarily ready for it, and then we kind of had the dot-com boom, and then things shifted. Well, now we're shifting back to these things that are kind of really present in everyday life.

You think about people who now ask Siri, "Where can I find this? Where can I find directions? Where can I get something to eat?" These things are already there. I think what makes people scared about AI is the possibilities, right? Now we know that it does a great job of recommending a movie or where I can get a coupon for a meal, but some of those other bigger things where we're not sure, and we might not have a lot of good governance around, is like how it can be used in healthcare.

Convenience vs. Caution in Healthcare

Stanton Shanedling: Yeah. So, you know, it's interesting. I used ChatGPT last week. I had to get my COVID booster, and recently I started a new medication. Just quickly, I went on ChatGPT and I said, "Okay, I'm gonna get a COVID booster. Is it okay to take with this medication now that I've started it?" Wow! Within, like, 2 seconds! You know, "Yes, it's safe," and da-da-da-da-da.

So, one of the things that I want to address here is the idea of convenience. In the back of my head, okay, I don't need to bother my physician. I can read up on this on my own. Are we using AI as a convenience, especially as it relates to health information?

Robin R Austin: Yeah, I could jump in. I would say possibly, for sure. I think that I've done it myself just to say, like, "Hey, I have this, this, and this, what do you think?" kind of thing. But I also have caveats with that. I mean, the prompt it could give you back, or what it spits out, the paragraph it provides you, may or may not be accurate.

Everybody's own health is very unique. Genetically, we're all very different, and so making sure, especially if it's something much more in-depth or complex—if you have a complex condition—always make sure to follow it up with a provider. There are blind spots I have as my own. Just always make sure you're working in partnership with your healthcare provider when it comes to that. Like, "Hey, this is what I think I see. Is this accurate or not?" I think it's a good resource or good reference to do that. I worry when people kind of just take that as verbatim truth and then go off and do their own thing. That makes me a little bit worried at this stage.

But I think that there are a lot of possibilities. I see it as being able to empower others. Like you said, you put in COVID, "Is this medication potentially okay?" Yes, there you go. I think it's empowering to have information for yourself to say, "Is this accurate? Is it not accurate?" and maybe act on it, or maybe say, "Hey, not at this point," or whatever it might be. But just to have another verification, I think, is always super important with that. I'm more conservative-cautious that way, I guess. Yeah.

Data vs. Empathy and Clinical Suspicion

Stanton Shanedling: Yeah, so, Barry, okay, you and I and Clarence didn't grow up with this, and you certainly didn't grow up with it in your medical profession. So, what do you think?

Barry Baines: Well, I think there are a number of issues here that really come to bear. If I can give a broad piece of it, if my patients found information—not necessarily AI-generated—I always welcomed it because I figured I could always learn something. But it also tells you something more about the patient that you're working with and understanding.

So, my big take on this is that you need to understand that AI uses lots of data, it crunches lots of data, and it uses algorithms to basically come up with probabilities. That's how it responds with what's the most likely thing. One of the things that has been alarming is that more and more people are using AI to not only get information, but also to sort of get that diagnosis and just accept it without appreciating that whether you're working with physicians, nurses, etc., AI doesn't have empathy. Okay, it has data. This is really important, because oftentimes what might move you away from just the number-crunching probability of, "Well, this is 37.2% likely it's this, but only 35.1%," is a real understanding of the patient that you're working with.

Without having that conversation with other healthcare providers together, I think you wind up trying to make it too easy and reductive. "Oh, this must be the answer." Most people don't appreciate, from a mathematical perspective, probabilities and reality. Just because there's a slightly greater probability of this versus that, you might be in an area where Ebola is endemic. That would be ranked very low on the probabilities right now, but if you're a practitioner, you might know that, so your level of suspicion would be higher. So on the strictly medical thing, that's where there are just yellow flags and things that we need to be aware of.

My bigger concern is what I'm understanding in relation to mental health, where most of the AI engines generally tend to either just feed back positively and to be very affirmative per person. I think that there's—you read newspaper stories about how it gets people down the wrong track. From that perspective, that's why with Robin and Rhonda and what we're talking about today, I think it's critically important for all healthcare providers to really have an understanding of what's out there, because their patients may be off on their own, giving it a higher level of certainty that really doesn't exist.

For me, those are some of the high-level concerns that I have. This is out there; we're not going back. So now, how can we integrate this into what we can do? Because the positive part of it is tremendous. We don't have the brains to be able to, at an individual level, put all this data together because it's constantly changing. That's the good thing, and so we get better follow-up and better tracking of data to be able to follow, especially with chronic disease in particular. So, that's sort of my take, and I'd certainly welcome other comments on that.

Addressing Gaps, Interoperability, and Bias in Data

Stanton Shanedling: So let—one thing that's come up, and I've talked to a lot of people, is the idea of trust. Our crew put together some things on how AI can improve healthcare, and there are like four different thematic arenas: preventative care, faster care for emergencies (what do we do, boom, you can get that information), tracking changes in your health, and then also streamlining administrative tasks, which I'll link with a show that we did prior. So, Rhonda and Robin, you have this program, TRIUMP. You're trying to train. What are you training? What themes are you using in your training in order for people to understand all this?

Robin R Austin: Yeah, great. Oh, go ahead, Rhonda.

Ronda Marie Chakolis-Hassan: I wanted to say, I think it was a great comment that Dr. Barry was making, and I wanted to make sure we don't lose that: understanding where is the bias, right? We're talking about the bias in AI, but I think this is also a good opportunity to talk about the bias in traditional models. For example, a lot of us, if we see a blood pressure of 170 over 100, that might be the first piece of data that we hone in on. A lot of times, that's what we do. The other thing is some clinicians overemphasize some of the psychosocial things or say, "Well, because you're from a certain zip code, or because you're from a certain demographic, this is what things are like."

But the other thing is the availability and burnout or information overload. I think when you flip that and kind of look at some of the benefits of AI in terms of being rapidly able to gather information, maybe that helps prioritize things. We also have to remember that we have data gaps, and I know Dr. Austin has talked about that extensively. Certain populations we don't have good data on. African American women, maybe in menopause—we won't have a lot of good data on that, so AI might not be especially useful. And so, I didn't want to lose that thought before you went into your next question.

Stanton Shanedling: Yeah, so, okay, good point, really good points. Alright, so what are you training? How are you training to kind of get our professionals up to speed on all of this?

Robin R Austin: Yeah, great question and great point, and thank you for bringing that back to the bias and data and all of that, because that will lead into kind of what we train on. With a lot of our courses, we go back to the basics, especially for health data per se. A lot of it is what's entered into the electronic health record or clinical trials, but if that is not accurate to begin with, or there's gaps in that data to begin with, that is a huge issue future-forward for trying to build AI models. Because if you're building AI models with data gaps, it's going to only amplify those gaps.

In our training, we really try to get students to go back to what is quality data. We need to have quality data to be able to understand and fulfill a lot of what these AI models and tools are doing. The other piece of it is we need to also emphasize the interoperability component, meaning that my data at Park Nicollet or HealthPartners may not talk to data within Fairview or within Mayo Clinic. That further limits our ability to look at data across health systems.

Stanton Shanedling: I'm a pharmacist.

Robin R Austin: Exactly, or even public health versus inpatient (which is in a hospital setting) versus pharmacy data, versus... and then mental health data, as what Dr. Baines was saying, is a whole other component as well because of privacy, quality data issues, all of those components.

So, what we try to teach in our programs is really the foundational pieces of what is good quality data, why do we need quality data, and what are the systems we can put in place to improve the data that we have. Maybe to improve the quality of the data for documentation purposes, which now they're hopefully having some AI tools that could help with some of that potentially—meaning ambient documentation, changing some workflows, and trying to relieve a lot of the documentation burden that healthcare clinicians have to do.

It's also kind of a catch-22 in my mind. Sometimes you have to start with the basics of having good quality, sound data in order to build some of these models for prediction, understanding healthcare patterns better, those kinds of things. But the other piece of it is our electronic health records aren't necessarily the best kind of data to get healthcare data from to train models, in my opinion, because a lot of what we put is in a clinical note, which is text or sentences, for example. Getting data out of text data is not the easiest way to do that either. That is another laborious task, adding on top of trying to build AI models at scale. A lot of our good, rich, robust information might be hidden in a paragraph, not in structured fields I could pull from to run a report, for example, or to run other data.

There are multiple things when you think about it. It's not just building a tool and putting it within an EHR, and therefore, now we've got a great model. It has to start with good data. We're trying to teach students to really go to the foundations, go to the basics of having an understanding of foundational data and interoperability pieces in order to build a lot of these more sophisticated tools later on in your careers, in your programs, and within your health systems that you work for. And being able to identify if something is not correct, having the ability to figure out why it's not correct, having the voice to say something about why it's not correct, and then being able to really identify kind of those skill sets.

There was a movement not too long ago around algorithm vigilance, meaning to understand the algorithms and to be vigilant about the data that's being put into it, how it's being trained, and to be able to identify if there's bias or gaps in it. So we're trying to teach students kind of all of these stop points along the way within our healthcare digital systems really, now, and the ecosystem of what they all are becoming. So that... I'll just stop there, because that's a lot of information I just kind of put there.

Critical Thinking and the Human Element

Stanton Shanedling: Oh, alright, so just to kind of surmise a little bit, and then Rhonda, I'll let you... and then Clarence, I know you've got a couple questions, too, here. It appears as though—correct me if I'm wrong—it appears as though part of the training is knowing how to ask the right questions? Critical thinking? Okay, because let's face it, we as humans are developing this AI. So, is that part of the training? Am I clear on that?

Robin R Austin: Yeah, absolutely, and I think it's the critical thinking component—being able to use your clinical skills to critically think about whatever electronic tool may be put in front of you. But I would say one of the biggest skill sets, not only is critical thinking, but it's the ability to discern whether something is accurate or inaccurate, and how we need to change it to get it more accurate. I think discernment will be a huge skill going forward, especially when we're thinking of living in these AI ecosystems and data-driven worlds. Critical thinking and being able to discern if something is accurate or not accurate. If it's not accurate—thinking of the nurse at the bedside, if she's given an error warning—is that really accurate or not? Based on what I see with my patient in front of me, if I'm getting this error warning that's saying something else, that critical thinking and discernment is going to be absolutely critical, I think.

Stanton Shanedling: Rhonda, what do you think on all this training? It seems like it's a huge undertaking, but a necessary one for students going forward and eventual professionals.

Ronda Marie Chakolis-Hassan: I would say it's a necessary training for almost anybody to really understand the true rules of engagement, right? A lot of times we hear the word algorithm. Algorithm seems like a big word, and we're like, "Well, what exactly is that?" Well, to me, when I describe an algorithm to people in the community, I think of it as something so simple as my morning routine, or the recipe that I follow to get to this wonderful cake that I like to eat.

The training really goes with understanding—yes, we know we want input, we have input of data, right? And we have output, we know what we want. I think most of my training, I would say, was really back to Dr. Austin's point: it's on the processes. What makes us get to that output? So if I don't put yeast in a bread that I wanted to rise, I can't be surprised if the bread I get is flat. It's really breaking it down for people like that, just kind of understanding.

When you're not following that recipe or that algorithm, sometimes that's great. We don't want to always be prescriptive, right? Even to this point, we get this wonderful product that we didn't know was possible if we hadn't taken a step out of that process. So, that's how I would really say the program helped me think. It's a challenge to even what we've been deemed as clinical. We have clinicians on here; we've seen ATP4 and 3, and the updated guidelines and all of these different things, but we recognize that they've had to change.

Going back to anything AI, anything that requires a human element—AI can't read body language. AI can't understand if you're hesitant about disclosing information about a situation. There are things in medicine, I tell people, that are always going to require a human element, at least in my lifetime. I'm planning on living to at least 100, so we've got at least 50 more years or so of that. Anything that requires a human element, we're gonna be around. It's really just understanding how we can take that information and really humanize it.

I think about even on the show, as we're talking about data and everything else, I like to give people real data points. In 2009, the top leading cause of death was heart disease and hypertension. In 2025? The top leading cause of death? Heart disease and hypertension, right? So despite all of these innovations and different things that we have, we still have the leading cause of death that hasn't... we haven't moved the needle. So I think that there is that opportunity to kind of look at those things.

Stanton Shanedling: It'll be interesting to find out what the guy upstairs thinks about AI. Alright, Clarence!

Interprofessional Eligibility and Workforce Upskilling

clarence: Okay, so what I wanted to ask was, I know that this is in the School of Nursing, but I heard something that says that other people could utilize it, other people could be trained by this, okay? And so, what other disciplines could utilize this TRIUMP training?

Robin R Austin: Yeah, great question. I think all of them, all health professionals. The certificate that we have is the Population Health Informatics Technology, or PHIT certificate. For all health professionals, it's a post-baccalaureate, so meaning you have to have a four-year degree to do the certificate. But it's open to nurses, pharmacy, medical, dental, social work, PT, OT... I think I'm getting a majority of them, but you get my point, is that everybody in healthcare could use this training.

The thing that keeps me up sometimes at night, or that I really, really worry about—not only because I'm in academia, so we're teaching the next generation of healthcare leaders and providers out there—is the current workforce. Because these things, as we've mentioned, are moving fast and furious and being implemented at the clinical setting super rapidly, maybe with some training, but maybe not as much as we probably should be doing. And so I really worry about the current workforce being able to manage and have the ability to have the autonomy within their clinical practice to make it work for them, versus adding a layer of burden and burnout, which they already have.

That's another concern: how are we going to upskill our current workforce in the healthcare field right now? It's a huge amount of people, and so that requires a lot of critical thought on how that's going to be done in a way that's scalable and also cost-efficient, because it's super expensive to train the current workforce. That is another piece of it.

But I certainly think that the training that we have at the university, even though it's housed in the School of Nursing, it's open to a lot of interprofessionals. In fact, we encourage a lot of interprofessional course dialogue just because if I'm in a class with Rhonda, I know she has the pharmacy perspective that I might not have as a nursing perspective, and vice versa. I think it's a great way for students to learn. It's how we practice out in the field anyway; we don't do things by ourselves. It's a great way to start to understand how to dialogue and really work interprofessionally in the classroom first before you get out into other practice or extend it into your everyday work that you do.

I wanted to make one more point when we were talking about critical thinking. I saw this quote and I found it, so I'm gonna say it around critical thinking and AI. People are very afraid of it losing their jobs, and it says: "AI does not eliminate the need for critical thinking. It amplifies the cost for not doing it." Which I think is a really key component. It's that AI is not going to take your jobs, but you have to have that critical component, or critical thinking. It's not going to take away the ability to do that. In fact, if you're not doing it, that's where the cost really might be.

Stanton Shanedling: You know, Einstein—I'll quote Einstein, I've used it on this podcast before—he would say, "Remember, not everything that counts can be counted, and not everything that is counted counts." So, inherent in that is the idea of trust and also knowledge going forward.

Public Trust, Privacy, and HIPAA

Stanton Shanedling: So, a couple things here. 32%—this is what the research is saying, though—32% of adults are turning to AI for health information and advice. That's today. 32%, alright? The majority, about 77% of the public, say—now this is interesting—they are concerned about the privacy of medical information out there. All right, so can you two respond to that a little bit when we're talking about privacy here?

Robin R Austin: Yeah, sure, I can... Rhonda, do you want to go ahead? Sorry.

Ronda Marie Chakolis-Hassan: Oh, you go first. Go ahead.

Robin R Austin: Sure, yeah, that's always a very interesting component. People are interested in using AI and other tools and digital apps and things like that. We recently did a survey on midlife women using technology, and they're super excited to use it, very digitally engaged, and willing to share their information for research purposes if it was de-identified. But then I asked, "How many read the privacy policy statements?" and probably like 90% said they don't read those privacy policy statements.

Being able to understand what tool you're using and what they do with the data on the back end is very, very important for your own empowerment and own stewardship of your data. But I would say, too, on normal practice, if you're using a health app, they're not covered by HIPAA. I think that that's an important fact to know, that it's not a covered entity, and to try to maybe not use your full identity information in some of these apps if you can handle it.

Same thing with like ChatGPT: don't put your identifying information in there. These are just kind of normal practices if you're trying to be careful of where and how your data is being used. Within healthcare institutions, it's a little bit different because they are covered by HIPAA, but you still have some say if you want your data to be included in research or not; you might have an opportunity to opt out. Being able to inquire about some of those things as well is important for your own safety around what your data is doing and where it is. I feel like everything is out there already; however, there are still some safeguards that can protect us, particularly around mobile health apps. I think it's really important to understand where that third-party payer is going around your data, and if your data is being sold, that's something you might want to know. But I'll stop there and let Dr. Chakolis jump in.

Stanton Shanedling: Alright, hold on one sec. Clarence! Clarence, I mean, you're always good at grounding us a little bit here from a community perspective or a population perspective. What do you think about all the people that you talk to? Are there any concerns around AI, or are people just saying, "Well, I guess it's out there, so I guess we have to kind of deal with it?" Give me your perspective on it a little bit.

clarence: I think it's a little bit of both, yeah. I think people are just like, "Well, it's just the way life is, you know." Other people are saying they're afraid that it's going to interrupt their lives, and I think that part of what's important is to have conversations like this where we can talk about what's going on. One of the things I was going to say—I was gonna use this at the end, though—one of the reasons why I strongly liked this program is because I watched Dr. Rhonda's journey with getting involved in this program. I knew her before she talked about AI and those kinds of things, but I also watched how, as she went through it, what she was saying about AI. So that was one of those kind of things where, even though it's not in my lane, because I see people that I know and I can trust their judgment, I can be much more open to the conversation.

That is one of the reasons why we're doing this program today, is because it is something that we're going to be talking about. We need to figure out where are the places where people can interface, and that's why I was asking the question about who could be eligible for this kind of program, and why should they be eligible for this program? What are the outcomes of this program? I think Dr. Barry mentioned empathy. All these kinds of things are all there, but they're not going to go away, right? And so we have to figure out a way how to answer this. So that's my long answer to what you're saying, is that people have got to find a way to interface with this technology in a way that's going to be comfortable and safe for them.

Stanton Shanedling: Yeah.

Interoperability Challenges and Governance Models

Ronda Marie Chakolis-Hassan: I have a comment. As you know, I do stuff with the Minnesota Board of Pharmacy. Recently I attended our national conference and had the opportunity to witness this wonderful keynote by Dr. Gupta, who's an expert on CNN, talking about data and visualizations. One of the things that Dr. Austin kind of pointed out is the burden of things. Well, what happens a lot of times in pharmacy is people are like, "Go ask your pharmacist about this, go ask your pharmacist about this. Did you get your vaccine? Did you do this?" There is this burden on the pharmacist, right?

I had an opportunity to ask a question in a room full of hundreds of people, and we had a mic-drop moment when I actually asked this expert, "How do we solve for interoperability—that is, the sharing of information from what happens at the pharmacy to other health entities?" It was radio silence. So imagine being in a room full of experts and people not having the answers. What this tells us is that anybody can kind of enter this field, and it's really about that process for solving. We haven't solved for everything. There's a perception that maybe there is an answer, but it hasn't solved anything.

Kind of circling back to your question about community perception, I think there's a couple of things that happen. One, people have become very complicit and feel like they don't have a lot of autonomy or control of their data. But in a big population, it's only heightened mistrust. I know Clarence and I can attest that if there's a meeting in the community and there's an AI transcript recording trying to take notes, people will immediately get off that meeting because they're afraid of how that information will be utilized. But again, it goes back to thinking: how do I want my information to be utilized? So often, we don't take the time to start thinking about, "Well, do I want my information shared? Do I want it with a research study? Do I want to be notified if it's being utilized? Do I want to be notified if the company is sold?" We need to take a little more control and start thinking that way.

I've also said this before: we need to demand that that important information be at the top instead of being way at the bottom. Yes, you have to get through all of these disclosures, but it's really easy to say, "Your information may be utilized as follows," right at the top. That way, you don't get that fatigue. I probably wouldn't have thought about that before because I was so rooted in that clinical way where you sign the consent form, and the last thing on the consent form is, "Do you want your information to be in a research study?" Being actively involved in training in the TRIUMP program taught me that I can challenge existing norms and really think about a better way to solve things for clinicians and for patients. A lot of times, those last few questions are what gets people hung up, and then that mistrust builds, and then here we are.

Roundtable Discussion: Crew Perspectives on AI

Stanton Shanedling: You know, we've got some of our research gang with us, and it would be interesting to hear either Ariana or Aaron or Matthew's perception or feelings about AI. Are you comfortable with it? Are you not comfortable with it? Do you use it? What do you think? Matthew, I know you use it in your work.

Health Chatter (Matthew): I do use AI a lot, yeah. How I feel about it is... mixed emotions, if you will, right? Like, how do you trust it? How do you validate it? Is it making my life easier if I'm spending 20 or 30 minutes developing the prompt to get me the actual data to do the work that I wanted to do? Then once I get that work, I'm having to spend more time validating that work. At what point do I just do the work? And so... I wouldn't say I have found that answer yet, but it's technology, right? We're figuring it out as we go, and I'm cautiously optimistic, I suppose, is the best way to describe it. I feel comfortable with it, but also understanding its limitations.

One of the things that I would love to see different about AI is I hate that it can't admit it's wrong. I don't like that it doesn't know when it's wrong. If it can't find anything, it just starts predicting or making stuff up. I think that's a major issue. If you're asking it a question, I want you to tell me you don't know the answer, that you don't have the information for that, because then I can go look somewhere else rather than putting out all of this bad information and bad ideas. That's where—I know it's kind of been mentioned on the show—you have to really piece out the truth there. It makes it a lot more difficult if you're being fed information and the AI thinks that it's correct. So, I don't know. Cautiously optimistic, I guess, is a great way to describe my current feelings.

Stanton Shanedling: Ariana, you have any thoughts on it?

Ariana: Sorry, just needed to unmute. I mean, I guess I have less experience than Matthew with using it, but I think about the impacts economically and environmentally. It's another very high-earning industry, and having lived in San Francisco, I feel like I'm a little bit more sensitive to the gaps between the classes, basically. So, I can't comment much on the actual use of it, but yeah, I'd say I'm less optimistic.

Stanton Shanedling: Yeah. At this point, right? At this point.

Ariana: Good point. Yeah, yeah, I could be proved wrong.

Stanton Shanedling: Yeah, yeah. Aaron, are you there? You want to chime in on it? Are you there? Hello, hello, Aaron!

Erin Collins: I am here.

Stanton Shanedling: What do you think?

Erin Collins: I mean, I was only here for the second half of the conversation, but I think I echo similar concerns to Matt and Ariana. I think the idea is really fantastic. When I think about AI and healthcare, I think about—and biasly—people with type 1 diabetes who are accidentally killed by an insulin overdose that could be completely prevented by something like AI use. I love the idea of that, but I also think about data centers, the use of water, and global warming, and how we could possibly have the best of both worlds without harming the planet in a horrible, horrible way that it looks like it's going towards. So I'm just so curious to see how things progress and how things are regulated and controlled as well, but the idea of having it in healthcare is also extremely interesting and hopeful to me.

Future Potential and Closing Thoughts

Stanton Shanedling: You know, I think cautiously optimistic is a good way of putting it. I'm not negative about it. It's kind of cool, you know? But on the other hand, there are some questions that come up. All right, Barry! Circle us around. Alright, no, you're on mute.

Barry Baines: There we go. My last words on this are that I think everybody pretty much agrees that the potential for AI within healthcare is huge, and it's great. Then the flip side of that is that the road there is littered with potholes and big stones to trip over. Some of these other issues that came up are things that not only do we expect individuals to have to deal with—which can be very overwhelming—but society at large is wrestling with some of these bigger ethical issues, privacy issues, and data issues as well. So, the potential is large, but I think it's gonna be a rough journey until we get to the promised land for what AI can offer.

We certainly see reports of how it's been helpful in some of the research now. Even from our last week's program on longevity, looking at some of the genomic things that AI can get into to really be helpful in that area and in a lot of different areas. There is the Mayo study where AI has been able to find earlier heart failure issues. Reflecting back on what Rhonda said at the beginning about cardiovascular issues and hypertension—again, the potential is there, but we haven't figured out the way to get there yet because there are these pitfalls along the way.

Part of this that we all agree with is we need more conversations like this to bring these issues out into the public to get more of us just talking about it so that we can identify those kinds of things. I like the idea with the disclaimers, having it at the top of those policies that we read, because I'm in the 90% also where my eyes start to go...

Stanton Shanedling: Right, and the font should be bigger than, like, .4. That's right.

Barry Baines: That was my circle around. I don't know if that captures stuff together, but that's one retired physician's perspective.

Ronda Marie Chakolis-Hassan: I like to do this.

Stanton Shanedling: Robin, so... lasting things that the public should know at this point in time, because I really hope that you and Rhonda can both be back on the show as we progress in this AI venture. But go ahead, your perspective, last thoughts.

Ronda Marie Chakolis-Hassan: You can go ahead on that.

Robin R Austin: No!

Ronda Marie Chakolis-Hassan: That's worse! Oh, sure. Yep, please do.

Robin R Austin: Oh, okay. Sorry. Yeah, so last thoughts. I think this has been a great conversation. I think more education for the public in and of itself is needed, but I'm super optimistic. Cautiously optimistic, I would agree with as well. Being able to train our upcoming next-generation workforce in the proper way to really be able to handle this is key, but also thinking of our current workforce—how are we going to upskill a lot of our current health professionals as well is another big component.

I'm optimistic that we're going to hopefully figure this out, knowing that there are some caveats along the way. But it's coming, it's here. I think we just have to kind of jump in and go with it, but within the confines of what you're comfortable with, and also being mindful of what you're sharing out there. We try to think we're de-identified, but really, the risk of re-identification is always there. Just kind of lead with that, I think, when you are putting any information into any sort of digital tool. But yeah, thanks again. Great.

Ronda Marie Chakolis-Hassan: I'm a student of history by nature, so I think sometimes we always feel like we have to reinvent the wheel. When I think about ethical care in healthcare, we know that there are four principles. One, autonomy: patients should have the right to make decisions, right? Then we have beneficence: providers must act in the best interest of the patient, which sometimes even means some patients don't want to use AI in a room, and you need to respect that. The other thing is non-maleficence: again, not doing harm. And then, of course, justice: that these resources are going to be spread equally.

Here's what happens, and I feel great in terms of my lifetime of what I've been able to witness in terms of healthcare. In healthcare, just slightly before I was born, came Medicare, but right after that, we saw things like Medicare Part D, right? Before then, we never saw prescription drugs even being covered by Medicare. People can't even remember that time, but my grandparents used to have to pay full price for their drugs. But we also have this legislation called HIPAA, and so many of us forget what life was like before HIPAA. I think we have a model there that we can take in terms of helping with governance, because it's really hard to govern something once it's already been out there. I think if we go ahead and kind of use what we've done with HIPAA—that law, that legislation in terms of insurance and patient information, and even being notified when your information is being shared or breached—those are models that we can use. That is why I'm very hopeful in terms of AI. I think that doesn't just apply to AI being utilized in healthcare; I think it applies to any sector that we need to focus more on governance.

Stanton Shanedling: Clarence!

clarence: So, you know, the great thing about Health Chatter is that we want people to enter the conversation. And I think that's one of the reasons why we're so energetic in some of our questions and our thoughts and things like that. But I know that there's going to be some people who are going to want to know more about AI and about TRIUMP, and so could you just quickly give a quick way in which they might be able to find out more information?

Robin R Austin: Absolutely. You can go to the University of Minnesota School of Nursing website, and underneath our certificates, that's where you'll find more information about the PHIT certificate, which is Population Health Informatics and Technology. We also have one called Leadership in Health Information Technology, or LHIT, if you're interested in that one as well. So there's one with more of a population public health focus, and one more of a leadership focus. School of Nursing website, University of Minnesota. Feel free to reach out to me personally if you'd like to do that as well; that'd be a great place to find information for the programming.

clarence: Thank you.

Stanton Shanedling: So... you know, I can't help but think that AI is a tool. It's a tool in our toolbox. It's a kind of an all-encompassing tool, but a tool nonetheless. There's an article in today's New York Times about writing—just writing. And you know, okay, tell an AI tool to write a good chapter on blah blah blah. Well, wait a minute, what about you writing something? So there's a balance.

The other thing that's kind of struck me in all of this is, keep in mind it's called artificial intelligence. It's artificial, okay. Will there be a day when we call it real intelligence? In other words, will there be a day when we don't think of it as artificial or fake or whatever, where we really depend on it? At that point, I have a feeling that it will be called some different form of intelligence.

So, thank you, everybody. Boy, wow, a great crew here to provide some useful insights into artificial intelligence. Like I said, Robin and Rhonda, we reserve the right to get you back on this show, or anytime that you'd like to be on the show, you just give us a call. So, really, thank you for your insights about all the different things that you're doing in this arena.

To our listening audience, we've got an interesting show coming up on disabilities research and policy, and there will be a link, I'm sure, to AI in that arena as well. So, everybody out there in listening land, keep Health Chatting away.