Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!
Cardiology and preventive medicine fellow Anand Shah discusses his article, “The role generative AI can play in cardiovascular education for patients,” exploring how artificial intelligence can revolutionize patient education and prevention strategies. Anand highlights the challenges patients face in understanding and adhering to cardiovascular disease prevention plans, emphasizing the limitations of brief clinical visits and generic lifestyle recommendations. He explains how generative AI can bridge these gaps by delivering personalized, accessible, and culturally relevant health guidance tailored to individual needs. The conversation also addresses concerns surrounding AI accuracy, bias, and data privacy, while envisioning a future where AI-driven education enhances patient engagement, medication adherence, and long-term heart health. Listeners will gain insights into how AI can empower patients to take control of their cardiovascular well-being.
Our presenting sponsor is Microsoft Dragon Copilot.
Microsoft Dragon Copilot, your AI assistant for clinical workflow, is transforming how clinicians work. Now you can streamline and customize documentation, surface information right at the point of care, and automate tasks with just a click.
Part of Microsoft Cloud for Healthcare, Dragon Copilot offers an extensible AI workspace and a single, integrated platform to help unlock new levels of efficiency. Plus, it’s backed by a proven track record and decades of clinical expertise—and it’s built on a foundation of trust.
It’s time to ease your administrative burdens and stay focused on what matters most with Dragon Copilot, your AI assistant for clinical workflow.
VISIT SPONSOR → https://aka.ms/kevinmd
SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast
RECOMMENDED BY KEVINMD → https://www.kevinmd.com/recommended
GET CME FOR THIS EPISODE → https://www.kevinmd.com/cme
I’m partnering with Learner+ to offer clinicians access to an AI-powered reflective portfolio that rewards CME/CE credits from meaningful reflections. Find out more: https://www.kevinmd.com/learnerplus
Transcript
Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast. Today we welcome Anand Shah. He is a cardiology and preventive medicine fellow. Today’s KevinMD article is “The role generative AI can play in cardiovascular education for patients.” Anand, welcome to the show.
Anand Shah: Thank you. I’m excited to be here, and thanks for having me.
Kevin Pho: All right, so let’s jump straight into that KevinMD article. What led you to write it in the first place?
Anand Shah: Yeah, I would be happy to tell you. I really got into this article as I have started using generative AI for myself, especially with OpenEvidence, and realized that this is a tool that patients could actually benefit from—something very similar to it. From that, it really made me take a step back and realize that when we look at our health care system today, it is pretty clear that we struggle a lot, particularly with patient cardiovascular education. We are really good at managing acute presentations and have the infrastructure set up for that, but when we are talking with patients and managing their day-to-day lifestyle choices and diet, it is clear how much we are lacking. Everyone knows that we need to eat right and exercise more, but to do so is very difficult. In our 15- to 20-minute encounters with our patients, we really do not have the time or training to work in depth with them and help them with their dietary choices.
This leaves us with being able to give them vague advice like avoid fried foods, avoid eating fatty foods, and exercise 150 minutes a week. Unfortunately, that kind of vague advice does not give them the support they need to make those day-to-day changes. I think this is where generative AI can really change the game. I was experimenting with ChatGPT and tried to feed it different prompts and got to the point where it could give me a week-long meal plan schedule to improve cholesterol and hypertension levels. As I prodded further, it turned that into a grocery list and could even estimate the expected price based on my location. I think the potential is incredible. If we look and see how we can guarantee a model that uses validated information that can be provided to patients, it could be game-changing and have a large impact on patient outcomes.
Kevin Pho: So talk a little bit about that knowledge gap. You went from the vague advice typically given on patient handouts to the shopping list that generative AI can create. Generally speaking, what is that gap, and what kind of information, in an ideal world, would you like patients to have—especially given that you are a cardiology and preventive medicine fellow?
Anand Shah: Yeah, I think it has to do with a lot of our day-to-day or week-to-week dietary choices. Unfortunately, as clinicians, we cannot realistically hold patients’ hands or have them ask, “Oh, what should I eat this coming week?” While we can plug them in with nutrition or even give general advice during visits—like limit carbs to this amount, limit fatty foods, or increase fruits and vegetables—it is tough to translate that into a tangible list of meals and ingredients. I think generative AI could potentially fill this gap by using guidelines and what we know in terms of balanced meals. It can screen recipes and say, “Hey, these are the exact meals that fit these recommendations, and here is what you should buy to make them.”
Unfortunately, as clinicians, it is simply not part of our training to dive into such details. We are appropriately focused on medications, pathophysiology, and how to prevent another cardiovascular event, but we are not trained in providing detailed meal plans that directly address those concerns.
Kevin Pho: Yeah, you are absolutely right. I have talked to many physicians about their training in prevention and nutrition, and it is very limited. Obviously, you specialize in this. Talk about that lack of education among most primary care physicians when it comes to traditional medical training in prevention.
Anand Shah: Yeah, and I think it is just a function of the system. It is not necessarily wrong that we have so much to train on—there is a wide swath of medical information tied to medications and randomized trials. For nutrition, there are not many trials that prove definitive benefit, largely because there is no financial incentive to run them. Yes, we know the Mediterranean diet or the DASH diet can reduce hypertension, but even in those trials, there were disparities in how strictly each patient adhered to those diets. So there is not a strong foundation of randomized trials and knowledge linking certain things, but we do know some nutritional components of a varied diet can make a difference. It is about translating what we do know and giving it to patients so they can act on it.
It also makes sense that in our traditional patient encounters, we do not have the time or flexibility to dive into these day-to-day issues. It comes down to how we balance and use the time we have with patients to make the most impact.
Kevin Pho: Tell us about some of the prompts you use on ChatGPT to create a grocery list, recipes, and specific dietary changes for your patients. How did you come up with these prompts, and what were some of the exact outputs from ChatGPT?
Anand Shah: I would be happy to share. I have not done much in terms of actually recommending patients dive into this, mostly because I know ChatGPT and similar models are not extremely validated in this area. I do not want to risk giving patients wrong information or something potentially harmful. But from my own exploration, I have started prompts with instructions like, “Use American Heart Association or dietary guidelines for reference,” hoping it draws on established literature. If I do not preface it that way—like telling it to follow guidelines—I can get very different responses. In a general approach, it might find popular but unhealthy recipes, whereas focusing it with guidelines yields a healthier Mediterranean-type diet.
What impressed me most, and what I mentioned in my article, is how personalized it can be to each individual. It can start with a weekly meal plan and schedule, and then if I say, “Give it a Southeast Asian or Indian spin,” it creates a meal plan accordingly. That is truly impressive because generative AI can be customized and personalized to the individual, adjusting for language, education level, or cuisine type, all without having to wait a few weeks to meet with a nutritionist for a new meal plan.
Kevin Pho: From your use of ChatGPT and your perspective as a cardiology and preventive medicine fellow, is the information you get generally pretty good? Is it something that can be relied upon? I do not expect it to be one hundred percent medically validated—doing a Google search is not, either—but as a physician, how do you view ChatGPT’s information?
Anand Shah: Overall, especially if I prompt it with, “Use these guidelines,” the meals it suggests tend to include lean proteins, plenty of vegetables, and fewer fried or fatty foods. Surprisingly, it captures that really well. What I find great is that it reduces the burden; when I am trying to figure out my own meal plan for the week, it takes time and effort. With people’s busy lives, the easy thing is to grab fast food or something familiar. If we can integrate a way to reduce that barrier, giving people an easy, healthy meal plan without much planning, they might be more likely to follow it.
Kevin Pho: Give us some scenarios where you would incorporate generative AI into a typical patient visit. How would you introduce it, and how would you expect patients to use ChatGPT?
Anand Shah: That is the million-dollar question. For me, if a patient expresses interest—like, “I’m trying to eat well but can’t figure out how”—I might say, “Why don’t we pull up ChatGPT together and see what comes up?” I’m in an academic center and a trainee, so I have a bit more time with patients, but that is not the case in most clinical settings. Time is limited and valuable. It really comes back to working one-on-one with the patient and walking them through it. We also rely a lot on nutritionists. It could be, “Let’s see what this suggests, and when you meet with the nutritionist, you can go over it and make sure it aligns with their expertise.”
Kevin Pho: From what you are seeing, what is the general AI literacy of patients? Are you finding that patients already use ChatGPT or other large language models to answer questions about their conditions, or are they still primarily in the “Google and read about diseases” stage? What do you see in your clinic?
Anand Shah: It is a wide spectrum and depends on demographics and the clinic setting. Some patients still have flip phones and struggle with video visits, while others come in with the latest clinical trials in hand. For the most part, there is an element of privilege—more well-off patients have the resources and awareness to take advantage of these tools, while less privileged patients do not. I hope generative AI, and AI in general, can help bridge that gap rather than widen it, offering an easily accessible resource that benefits everyone without worsening disparities.
Kevin Pho: How about among the medical community, such as the attendings you work with, other fellows, and residents? Are they embracing AI to the extent you are? Do they see it as a positive tool going forward for patients?
Anand Shah: I think so. Among my co-fellows and attendings, there has been a lot of talk about OpenEvidence, and many are using it. It is a valuable resource, though not perfect. I have noticed flaws and inaccuracies in its answers, but it condenses information nicely. It can pull literature reviews or data on specific questions and make it easier to engage with that information. People are excited about it. The question is how to integrate it into our workflows so it truly helps us care for patients, rather than becoming cumbersome. There is a lot of good emphasis on it, and everyone is curious about how it will evolve to really enhance patient care.
Kevin Pho: You mentioned OpenEvidence. For those unfamiliar with it, it is a large language model similar to ChatGPT but based on established medical literature. Is that correct?
Anand Shah: Yes, it is. That is what makes it great for clinicians—we know the model draws on peer-reviewed journals rather than random internet sources. Still, it is important to understand medical fundamentals so we do not simply rely on whatever output the model provides as if it were the absolute truth. We need to parse and question it, recognizing that part of what it says might be correct while another part may not be. Having that conversation and pushing the model is key.
Kevin Pho: We are talking to Anand Shah, a cardiology and preventive medicine fellow. Today’s KevinMD article is “The role generative AI can play in cardiovascular education for patients.” Anand, let us end with some take-home messages you want to leave with the KevinMD audience.
Anand Shah: Absolutely. My take-home message, though a bit cliché, is to follow your interests. My interest in preventive medicine and cardiology led me to explore AI, and I see its potential to help us take better care of patients and help patients manage their day-to-day lifestyle and dietary choices. Each of us has unique gifts, backgrounds, and interests, and any one of those can drive positive change in medicine. There is a lot of low-hanging fruit, and if we all dive into our interests and keep exploring, we can improve outcomes for patients.
Kevin Pho: Anand, thank you so much for sharing your perspective and insight, and thanks again for coming on the show.
Anand Shah: I am happy to be here, and thank you again for having me.
