EP451: Hey, Let’s Not Talk About Artificial Intelligence, With Spencer Dorn, MD, MPH, MHA

Episode 451. " Hey, Let's Not Talk About Artificial Intelligence". Today, I speak with Dr. Spencer Dorn. American Healthcare Entrepreneurs and Executives You Want to Know, Talking. Relentlessly Seeking Value. Before we kick into the show today, I just want to make two points. Here's the first point. Together, we can do it. No one said transforming healthcare and elevating patients over profits would be easy. And it is not. It's really, really hard.

I just want to say thanks for all that you have accomplished, Relentless Health Value, RHV tribe members. These are the things that matter to really our entire country, friends, family, patients, members, and in so many ways is really worth it. Second point I want to make is to thank everybody who has left a tip in our tip jar. Some people have even left recurring donations, which, wow, my faith in humanity is restored when I see my people offering their financial support this way.

I feel this way because, A, I don't usually ask for financial support on the pod, even though it's something that is certainly on my mind a lot. And B, those who offer financial support, at least at this time, don't get anything other than knowing that their help helps this podcast continue. Which again, just warms my heart. The show today is a companion, I'm going to say, to episode 446, also with Dr. Spencer Dorn.

In the first show, we didn't talk about the impact of EHR, electronic health record systems, and in a similar vein today, we're not going to talk about the impact of artificial intelligence. I'm phrasing this in this kind of odd way because that earlier conversation with Dr. Dorn was about Kranzberg's first law of technology. And this one is too, where Melvin Kranzberg says, "Don't ascribe any given technology a label of good, bad or even neutral".

Point being, let's not sit around blaming or crediting a technology for downstream consequences. After all, I mean, if we're thinking about just EHR instances, there's EHR instances where it takes 60 clicks for a doctor to order a patient Tylenol. 60 clicks! Then same EHR system installed in a different hospital, it can take two clicks.

Those excess 58 clicks aren't because of the technology itself, they're because the technology was configured poorly by humans involved in configuring the technology. And if that technology then results in burnout or moral injury or someone insisting on measuring 58 quality measures in the most labor intensive way possible, that's a function of how that tool is used or configured, not anything inherent in that technology itself. So yeah, watch where those fingers are pointing.

And all of this is equally relevant to artificial intelligence. As Dr. Dorn says, there's no intrinsic property of the technology, any technology that determines the outcome. It's how we use it, how we implement it, how we put it into daily practice. That really ultimately is the arbiter of what happens and how it impacts lives. I'd also just add even if the tech itself has some glitches or hallucinates. Someone decided to use it in the current form it's in. So yeah.

So the first takeaway from this short show focused on artificial intelligence is going to be the same really as it was in episode 446 about EHRs. Do not ascribe any given technology a label of, as I said, good, bad, or even neutral. That is, as I just said, Kranzberg's first law of technology, and it applies here too.

Second major takeaway, and again, this is the same as in that earlier show about EHRs, but today we're talking about AI, if you're thinking about the ultimate impact of the people and the processes that have some technology in their midst, technology, again, such as AI, artificial intelligence, the ultimate impact will not be a black and white binary.

We talk about some of these nuanced, not binaries in the 10 minutes that follow, but for more, I'll put some links in the show notes for some newsletters, et cetera, to check out. One last thing before we get into the show today. Speaking of AI, I asked Google about myself, and this is what the Google AI bot replied. "Richter is also co-president of Aventria Health Group, a consultancy, and QC Health, a public benefit corporation". Okay, so far so good.

"She has also been recognized for her work on Relentless Health Value by winning the Edward R. Murrow Award." Hm. Just for the record, I did not win the Edward R. Murrow Award, which is actually a really prestigious broadcast media award. So, yeah. This podcast is in fact, factually sponsored by Aventria Health Group. And with that, here is my conversation with Dr. Spencer Dorn about, but not about, artificial intelligence. Dr. Spencer Dorn is a practicing gastroenterologist.

He also helps lead a large academic practice and works in healthcare IT and clinical informatics. My name is Stacey Richter. This podcast is sponsored by Aventria Health Group. Dr. Spencer Dorn, welcome to Relentless Health Value.

Thanks for having me. I feel like it is a legitimate concern that given Kranzberg's first law of technology, like this could go really well with AI. It could be amazing. We could revolutionize healthcare and all kinds of amazing things could potentially happen.

But if we aren't really thinking through the use case of the EHR, clinical decision support, or patient portals, or virtual care, or things that could be analyzed and we can do better next time and incrementally improve, if we're not thinking about how we're deploying AI, then this could go off the rails.

I think we have a lot to learn from our experience with digitizing healthcare, that will serve us well as we move to adopt artificial intelligence tools. And going back to Kranzberg's law again, we should not expect AI to fix all our problems. But we should be optimistic that AI can help. At the same time, just like EHRs, there will be unintended consequences.

So, in my opinion, discussions around AI right now are really charged and polarized with people on one side saying like Marc Andreessen, this will fix everything, and people on the other side saying we're headed for doomsday. We need to step back and say, look, EHRs were kind of like this. They've been a mixed bag. There's been some good. There's been some bad. How do we apply these lessons in ways that we could maximize the good and minimize the bad?

I think there's a lot that we can and should learn from the past 15 years of digitizing healthcare that will hopefully help us do much better when it comes to AI adoption. Again, it's kind of all about how it's deployed, how it's used, what does good look like, making sure that we are fixing things fast when things do go horribly wrong. And speaking of just because you're a GI doc, did you know that, rocks are important for digestive health yeah, you should eat at least one small rock a day.

That was, someone asked Google the first day. I didn't see that. I know Google is, Google got some bad press for recommending glue on pizza. as a topping, but I hadn't seen the eating rocks recommendation. No patients have come to you? Not yet. Apparently Google was training its algorithm on The Onion. The Onion is a parody website. But anyway, at some juncture, they wrote an article about eating rocks and Google took that as sound medical advice. So, you know, just there's pitfalls, right?

And it did get yanked relatively quickly. But again, I think maybe just back to what you said before, it's not going to go flawlessly. It's just not. And so there has to be kind of a recognition and eyes on what is going to go wrong because something's going to go wrong. And then probably what good looks like is to fix it really fast. Yeah, I think there are a few categories you could look at how AI

could make things better and also at the same time make things worse. For example, physicians, healthcare workers, nurses especially have too much to do, right? AI can potentially take some of that work away. Some of the drudgery, some of the monotonous, boring work that we're bogged down with so we can spend more time speaking to patients. So that's an example of major potential benefit and there are various companies and health systems and physician practices working on this.

Yet at the same time, AI could create more work, right? Sometimes using AI to do a job is harder than just doing it yourself. And there's a study recently that came from UC San Diego, where they were looking at using ChatGPT to answer some patient messages, some of those patient emails we talked about earlier, and paradoxically, the doctors who use GPT to respond to patients spent 20 percent more time on messages than those who didn't.

And the reason for that, I've experienced this personally, is that when you use GPT to respond to a patient message, you become an editor rather than a writer. And I don't know about you, but for me, I can write faster than I can edit. So it makes the work harder. So that's just one example of how it can make some work easier, but it also can make some work harder.

And, you know, there are, several other examples of different domains of clinical medicine where we'll see that same benefit on some ways, drawback on others. Yeah. Again, back to Kranzberg, AI, similar to every other technology that we have talked about and hope to deploy, it's not good, it's not bad, it's not neutral. It's how it's used, how it's deployed, how it's operationalized.

And then I think most importantly, as we step into new terrain here, what the plan is to incrementally improve when we find an issue. And I think, you know, the EHR case study is a really, really good one that if we think to ourselves, okay, we're going to deploy this. The end.

Like no further department effort, budget, we're not throwing our backs into necessarily the improvement process as much as we've thrown our backs into the, let's just get this stood up and maintained area, then we could wind up finding ourselves at odds with where the net starts to veer on the negative side relative to what is possible.

Yeah. And to me, that's why we need to look beyond the headlines. And beyond this cloud of hype and also the excessive pessimism, and we just need to move towards the middle ground and carefully consider what are we gaining, what are we losing? To me, that is what it ultimately comes down to at technology is a clear eyed assessment of how is this going to make things better? How is this potentially going to make things worse?

So that we can take agency, make good decisions, build systems around our technology to mitigate the possible downsides and maximize the potential benefits and kind of proceed that way. We have, like, like we discussed, almost one to two decades experience doing this with EHRs. Let's make sure we apply some of our learnings to this next wave of technology. And that is probably a really good place to wrap up this conversation because that was so well said.

Dr. Spencer Dorn, if someone is interested in learning more about your work besides LinkedIn, which I would highly recommend that anyone listening should follow Dr. Spencer Dorn on LinkedIn, but is there anywhere else that you would refer people? We're very proud of the work we're doing in the UNC Department of Medicine. You could visit our website. It's a phenomenal group of people doing phenomenal work.

Anyone who's interested in our work, just check out the UNC Department of Medicine website, learn more about us, and you can find me on there if you have specific questions you want to connect with me about. Dr. Spencer Dorn, thank you so much for being on Relentless Health Value today. Thank you, Stacey. Hey guys, it's Marty Makary. I want to let you know that I love Relentless Health Value. I follow it and get the newsletter and it's great stuff.