Hello, and welcome to the BioSupply Trends Quarterly (BSTQ) Podcast. My name is Abbie Cornett, and I am the patient advocate and engagement specialist for FFF Enterprises. This podcast brings is brought to you by BioSupply Trends Quarterly magazine as an added extension to give readers an opportunity to hear from healthcare experts on trending topics important to them.

Abbie: Unless you’ve been completely avoiding the news, social media and conversations with your friends and at work, you’ve probably been hearing a lot about artificial intelligence, also known as AI. AI seems to be everywhere right now.

Some people are excited about it, some are skeptical and some are hoping if they ignore it long enough, it’ll go away. I don’t think that’s going to happen. Healthcare is one of the areas where AI has the potential to be incredibly useful, especially for providers who are already being asked to do more with less and less time. Think about what happens during your typical day as a provider. You’re seeing patients, reviewing labs, reading medical records, answering messages, documenting visits and trying to keep up with new research and explain complicated medical information to your patients. And somewhere in there, you’re supposed to even find time for yourself to eat lunch.

AI isn’t going to solve all of that. And it certainly isn’t going to replace the knowledge and experience of a healthcare provider. But it may be able to help make some of those everyday tasks a little bit easier. That’s what I want to talk to you about today. Not a futuristic version of AI, but the practical one. How providers can use AI tools available right now to save time, manage information and communicate better with patients, and make the workday a little easier.

I’m excited to welcome today’s guest, Stephen Walters. Stephen is a contributor to IG Living magazine and has written about practical ways AI can be used to navigate healthcare and make complex information easier to understand. Stephen, thank you so much for joining me today. Before we get into AI and healthcare, can you please tell us a little bit about yourself and what got you interested in AI in the first place? When did you start thinking about how it could actually be useful in healthcare?

Stephen: Yeah, Abbie, thanks a lot for having me. I’m happy to be here. And, just to give the audience a bit of a background, I do have some industry experience. I was 12 years in commercial pharma. The majority of that time was spent with Pfizer. So, I’ve always been interested in the healthcare industry. And I also am a rare disease patient, so I can see things through a patient lens.

So, I think having lived experience, kind of going through it as a patient, but also with the industry experience, I bring a unique perspective on things. And they’re not just moonshot ideas. These are things that can be implemented and adopted for patients that are, you know, kind of navigating their journey through the healthcare system. And I’ve always been interested in technology. And, you know, AI was one of those things that, years ago I was always signing up early for the new releases and beta projects and software tools and early days of ChatGPT. So, I’ve been very in tune with what these tools could do to help me advocate for my health. And, so it’s been a huge unlock for me in my journey, and I often want to share what I can with others just based off what I’ve learned and what I’ve gone through.

Abbie: Thank you, Stephen. You know, you truly bring a unique perspective to this as a patient and working with AI. When we’re talking about AI, what are we actually talking about? We can say providers can use AI. What does that mean? I know it doesn’t mean we’re talking about turning over medical decisions to a computer. What are some of the tools that providers already have access to, and what can they realistically do with those tools?

Stephen: Yeah, and I mean, the space is changing so fast so often. So, you know, what we’ve seen two, three years ago has already progressed so fast to what’s happening in today’s environment. And, of course, I speak on behalf of a patient and the commercial lens rep hat that I’ve had when I’ve spoken to both primary care and specialty care providers and how they’re leveraging some of this technology. And you know, it’s interesting because now we’re starting to find a huge use case with citations and transcriptions and how doctors are capturing the interactions with patients and their appointments. And it seems that, you know, a lot of these tools and a lot of the AI use case is trying to figure out how HCPs or doctors reduce the amount of admin burden that they have, right? So that they can really focus the amount of time doing what they love to do, which is practice medicine.

So, a lot of the tools and the feedback that I’ve heard just based off being a patient myself and being in these waiting rooms and being in the appointments, as well as being a commercial rep, is that they’re really trying to solve the admin burden. And again, it comes from just the conversations that they’re having with patients, charting the tools that can help summarize those charts and speed up the billings, as well as just general appointment setting and triaging patients. So, I’ve seen a number of different things in this space, which is quite interesting, and I think it’s only going to improve as time goes on.

Abbie: You know, you’ve mentioned some of the areas where physicians and providers can use AI. And you know, there’s no doubt about it. The providers already have more on their plates than they can frequently get to in a day. And one of the frequent complaints that I’ve heard from providers is they don’t have time to be with their patients anymore. Where do you think AI can make the biggest difference in their everyday workload? And, can you go over some of the things that you’ve touched on that it can help with like organizing notes, summarizing information, helping provide educational materials to patients? And, you know, just some of the repetitive tasks that physicians and providers have to do every day that eat up so much of their time.

Stephen: Yeah, I think again, like you mentioned it too, right? Like these tools are never going to replace the physician or the expert. They’re never going to diagnose. But I think they can take a lot of the admin and the nuance on the paperwork side of things. At least, again, that’s the feedback I’m getting. I’m not a doctor, so I don’t have first-hand usage of the tools in that context. But the feedback has always been, you know, how can we try to reduce the admin burden, the charting? And, again, it comes back to the ability to dictate into the computer for the notes and whatnot, you know.

So, I do think it’s interesting to see what the uptake is from the different providers. Like we also know that they might be using LLMs to just get tidbits of information on certain drug doses or contraindications or what’s on label and what’s not. You also have platforms like OpenEvidence and UpToDate that are more of a reference point for clinical studies and trials and trying to get information there. But I think, you know, what they’re looking for really is just how can you synthesize a large amount of information really quickly. And that’s what some of those tools are doing. So, like I said, it’s going be interesting to see what continues to evolve in that space.

There are so many different players and so many different use cases, but a lot seem to be, at least in the physician’s office, it seems to be around AI. And we can obviously get to the questions later around where we think the future is, and I mean that’s a huge unlock with drug discovery and whatnot.

Abbie: That’s the unknown. You know, one of the places that I potentially see AI could be beneficial are for patients who come to providers with a complicated medical history. A lot of the patients I work with have chronic or rare diseases and have multiple years of records from multiple specialists. They have medication changes, labs, imaging and treatments, and that’s a lot of information for a provider to absorb in a relatively short appointment. How do you see that AI can help providers make sense of all that information and identify what they really need to pay attention to?

Stephen: Yeah, it’s a great point and a great question because that is a huge gap and a huge challenge. And I’ve seen that, as a rare disease patient, your information is so siloed and fragmented among different providers, right? And so, we’re actively trying to explore and build in that space where we use our own tools or, at least, I am right now. How to capture those different meeting notes or consult notes, and what the labs are showing and imaging and really trying to piece it all together so you have really valuable information to bring to your appointment that is collected and curated. And, again, that comes back to advocating for your health. So, you know, if you were to ask me, “Steve, how would you build that five years ago?” I would have no clue. I would need to hire an actual software engineer to even try to remotely piece that together. Now, with AI, and the vibe coding market, I can at least put together a wireframe or an MVP that allows me to show a proof of concept. So, I think that’s where some of these AI tools. It’s not just ChatGPT and typing something into a chatbot. I mean, there’s a lot more use case and capabilities than just doing that.

Abbie: We’re way past that.

Stephen: Right. And so, a lot of people don’t know that. And I think that’s where some of the work that we’ve done and what I’ve spoken to you on in the past is AI literacy, right? Helping understand how to use these tools.

And even from a physician’s perspective, you know, looking for really niche clinical studies or potentially looking at certain populations within those studies — how can they capture that or gather that information? So, the tools are out there to do that now. But you’re right. The ability to piece all of that fragmented information is a huge challenge.

Abbie: And of course, understanding the information is just one part of it for a provider. Then there is explaining it to the patient. Talking to a patient in plain English. Medicine has its own language and providers use it every day, and patients don’t. What are some things providers can do with AI to help patients understand their complicated diagnosis, treatment plan or test results, and help explain it in a way that makes sense to someone who doesn’t have a medical degree.

Stephen: Yeah, and I think this is a big challenge, right? Because, you know, oftentimes you could leave an appointment and you don’t get a lot of time with the doctor. And so, I think that’s why you want to make sure you come in prepared and you know you’re ready to ask the right questions or you’re ready to get the answers that you want.

But, yeah, it’s really challenging sometimes to hear, you know, the medical jargon or the terminology and understand what it means as a next step. So, I do find that’s where I will heavily rely on AI and tools to try to break that down and explain it to me in an easy-to-understand fashion, right? So, you know, most patient records are available digitally. You just need to ask for them and have access to them. And so that’s an easy way just to port those into AI. And again, obviously there’s security and privacy, and we can talk on that too, but that would be an easy way to walk out of that appointment and use some of these tools to better understand what the terminology is and what it means for you as a patient.

Abbie: We’ll touch a little bit on the security coming up here in a minute. This is one of the things I find from providers that are interesting. I’ve worked with providers in the past, and we’ve discussed this before that, you know, there was, what was it, Google Docs. And before that were the pharmaceutical commercials. And providers would be very impatient when a person would come in and had used the Internet to self-diagnose. Now patients are doing that with AI. How does the physician need to address that from a patient’s perspective? Because they are coming in with information about their symptoms.

Stephen: Yeah, and I think, you know, we all kind rely or at least have used Dr. Google to kind of jump to conclusions. And now, you can add in like Dr. LLM, right? The large language models and chat GPTs, you know, for lack of a better term. So yeah, you’re right. There are studies that have shown that patients are using these tools to better understand and educate. And I think the biggest thing here is, you know, these aren’t regulated tools, right? So they’re really going to grab information and synthesize that into a window where it’s going to try to answer your question.

So, you know, what you prompt or what you’re putting in there and giving the information is going to be highly, highly important, because it’s kind of garbage in, garbage out. But to answer your question about coming in there and self-diagnosing yourself based off what a chatbot’s telling you is obviously very risky because you need to understand where the sources and citations for that answer are coming from. Is it coming from a reputable source? Is it coming from an earned media site? Understanding where the answers are getting pulled from would really educate the conversation. So, if you went in there and said, “I was doing some research, I have these symptoms” (never really diagnosing yourself until you let the expert and the doctor do that), but just saying this is what I came up with, this is my research, and here are the sources that came behind it. Something coming from a Cleveland or Mayo Clinic is a lot different than you know a Reddit thread. So just coming in and having that kind of context for the physician as opposed to just saying, “Nope, this is what I have, and I’m convinced of it because I asked Dr. LLM” type thing.

Abbie: You bring up an interesting point when you talk about the amount of information and the quality of the prompt. You know, but it isn’t patients who are the only ones dealing with this overwhelming amount of information. Providers themselves are having the same problem when it comes to research in that there’s so much of it that it’s hard to keep up. There’s absolutely no realistic way for any single provider to read every new study that comes out in their field. How can providers use AI to find relevant research and summarize that research and find treatment guidelines or emerging therapies?

Stephen: Yeah, and it’s a good question. I do have real experience doing this. So, you know I’ve worked with some specialists who are very interested in certain research areas or they’re very niche in a specific population or therapeutic area. And so, you can build agents, right? So, think of an agent as something that’s kind of doing the work in the background, where you’re kind of scanning a directory like a PubMed or clinical trials.gov. Or, you’re looking for certain publications or studies. There’s the ability to try to grab all that information, break it down and give it an output based on something that’s digestible for the doctor to read. So, you know, what we’ve set up in the past is something that’s very, very niche in their interested therapeutic or disease area. And then, on a weekly basis, on Friday mornings, they’re going to get like a one-page report of all the stuff that potentially would be of interest — whether it’s pipeline products or new clinical studies in that field. So, there are lots of ways to be able to not spend the time doing the manual research yourself. That’s where AI would be very helpful — trying to grab that, curate it. But again, it all comes down to building that properly, setting up the infrastructure through proper resources and databases that are trusted.

Abbie: You know, AI, and we hear about it all the time in the news, has been known to confidently give you information, but it’s completely wrong. I believe it’s called hallucination. As a provider, how do you make sure that AI is giving you reliable sources and that the research is valid?

Stephen: It is definitely a concern. As the models improve, they’re going to reduce the hallucination. But where you run into trouble is when you have vague prompts or vague search queries and you’re having it just go into the open web search to try to gather and curate and generate an answer.

LLMs never are going to disappoint. They never want to disappoint you. So, that’s where the hallucination issues come in. It doesn’t have the proper context; it doesn’t have the proper rules or skills built into the user to actually say, no, this is just not something that I’m confident in my answer. So, I think, you know, just understanding the biggest ways to tap into reputable sources, databases. You know, Claude, for example, has something called MCPs where you can go right into PubMed and use search within that.

And, you know, Claude Science has a dedicated arm just for doing those types of search queries. So, I think it really just comes down to where is the source coming from. And your best bet is to always, you know, tap through a direct database that is trusted and reputable and vetted.

Abbie: Now, when you talk about trusted databases, obviously one of the things providers cannot overlook is patient privacy when they’re using AI. What are some of the safeguards, and what should never, and this is both for patients and providers, what should never go into an AI platform?

Stephen: Well, I think you want to be cautious of just dumping personal or sensitive information in. You’re never going to put in data bursts. I just I would be very cautious and…

Abbie: redact any personal information.

Stephen: Yeah, I typically would lean that way. And, I don’t think any healthcare provider is ever going to do that anyway. A lot of the tools that they have access to within their charts and their EMRs, those are all HIPAA-compliant. They have SOC 2 Everything there is probably vetted as a technology or as a software to safely read the information from a privacy level and have those safeguards put in place.

But yeah, I mean for patients, just to sit there and dump something into an LLM or a model with not really thinking through if that would be something I’d be cautious about — like a lot of financial details, you know. Not just personal health information, but also on the financial or banking side, I’d just be cautious of that. That’s where maybe local models — where it doesn’t leave, doesn’t go into a cloud, it’s going to stay on your own computer —that would be a little bit more appropriate. It’s not to say that you can’t put in, you know, labs or imaging and things like that. I still leverage that to try to get some understanding of the data myself. But it’s a risk reward, and you have to be cautious. And you have to understand what you’re signing up for.

Abbie: You know, that brings me to my next question. We touched on it a little bit before, and there’s no way to predict what the future will bring, particularly with technology that is growing at such an exponential rate. And if we were having this conversation three years from now, not so much of where do you think the technology will be, but how do you hope providers will be able to use the technology, or how will the platforms have been grown since this time?

Stephen: Yeah, it’s a great question. And I think, you know, just to kind of shift the gears on my answer from earlier about how clinicians or how the medical environment is using it. I was very focused on the admin side of things where I see the future. And I I mean it’s not to say this isn’t happening now, but it’s exciting to see where drug discovery could go. So how are we, you know, going to market? How can we get, you know, faster access to breakthrough molecules through AI, right? Like that that’s quite exciting.

Abbie: I was going to say those are some of the potential applications with CRISPR.

Stephen: Yeah. I’m not familiar with CRISPR. What do they do?

Abbie: It’s basically rewriting the DNA. So, I forget what CRISPR actually detects. I’d have to look it up again. But it’s cas9, and you actually go in and are editing the genes.

Stephen: OK. So that is the kind of stuff that would be really interesting to follow.

Abbie: I’ll send you some of the articles that I have. It’s really, really cutting-edge technology.

Stephen: Right. And I think that’s only going to get better, right? And so, you know, two, three years from now, you know, hopefully there are even more improvements on the diagnostics side of things, especially on cancer screenings and pathology.

You know, I still see that maybe there are ways to improve the current AI scribes that are happening in the office. And I assume that the more data that they get and collect, those can only get better and better. You already have technology kind of going into the EMRs and trying to proactively look for trends on potential early diagnosis of rare disease and all sorts of things there. So again, that goes back to like a screening play. But yeah, I think that there are a number of different buckets here that are interesting to kind of keep an eye on. But like drug discovery, just based off my past and my history and knowing how long it takes for a drug to go to market, hopefully they can fast track that based off the technology that’s coming.

Abbie: Stephen, I want to thank you so much for joining me today. I really appreciate you helping us get past some of the hype surrounding AI and look at how providers practically can use it. I think really that the point of this conversation is that AI doesn’t have to change the way providers practice medicine to be useful. It can help summarize an article — help you turn a complicated explanation into something patients can understand more easily.  And it takes one of those annoying repetitive tasks off your plate like charting. That’s a win.

But this is healthcare, and that means we also have to be very careful. AI can be wrong, and it can make things up. We talked about hallucinations. And just because it gives you an answer in two or three seconds doesn’t mean that answer is correct. You still have to do your due diligence. And the most important thing to remember is the provider is still the provider. AI is a tool that you can use to enhance the patient experience and take some of your workload off.

Stephen, thank you again for joining us and thank you to everyone listening to the BSDQ podcast. If you’ve enjoyed today’s conversation, please subscribe with a colleague. Until next time, thank you for listening.

Abbie: Thank you again for joining us today. Additional information regarding this podcast can be found on our website at www.BSTQuarterly.com. If you have a question that was not answered, please contact me at [email protected]. Look for the next BioSupply Trends Quarterly Podcast on our website for the opportunity to submit your questions.

The BioSupply Trends Quarterly Podcast is a copywritten production of BioSupply Trends Quarterly magazine, a publication of FFF Enterprises Inc. The views and opinions expressed by the speaker are their own and do not necessarily reflect the views of BSTQ magazine, FFF Enterprises or its affiliates.