Dr. Jen Chatfield (aka Dr. Jen the vet) joins Dr. Andy Roark to discuss how clinicians are already using AI and machine learning in practice, and how these tools can be leveraged to improve clinical outcomes.
You can also listen to this episode on Apple Podcasts, Google Podcasts, Amazon Music, Soundcloud, YouTube or wherever you get your podcasts!
LINKS
Chats with the Chatfields: https://chatfieldshow.com/
Dr. Andy Roark Exam Room Communication Tool Box Team Training Course: https://drandyroark.com/on-demand-staff-training/
Dr. Andy Roark Charming the Angry Client Team Training Course: https://drandyroark.com/charming-the-angry-client/
Dr. Andy Roark Swag: drandyroark.com/shop
All Links: linktr.ee/DrAndyRoark
ABOUT OUR GUEST
Dr. Jen Chatfield (aka Dr. Jen the vet) is a national thought-leader in infectious disease and conservation medicine. She’s been a practice owner, a relief vet, worked in public health, serves as a member of the National Veterinary Response Team, and is a Medical Reserve Corps member. She serves on the advisory board for DVM360, PetVet Magazine, and VPNextGen and as an Associate Editor of the Journal of Zoo and Wildlife Medicine. Dr. Jen is one of the most popular speakers at the largest veterinary conferences and has been quoted in national magazines including Better Homes & Gardens. She was selected as a Future Leader by the AVMA and has been awarded 2 Gold Stars for contributions to veterinary medicine by the Florida Veterinary Medical Association. She loves French bulldogs, Himalayan cats, the dirtiest of vodka martinis and basking on Caribbean beaches.
EPISODE TRANSCRIPT
Dr. Andy Roark:
Welcome everybody to The Cone of Shame Veterinary Podcast. I’m your host, Dr. Andy Roark. Guys, I am here with Dr. Jen Chatfield. She goes by Dr. Jen the Vet. She is a lecturer, she is a podcaster, she is a YouTuber. She has a show called Chats with the Chatfields. You can find it at chatfieldshow.com. Things like that. She is doing a webinar for VetFolio that I saw and thought was really interesting about leveraging AI to improve clinical outcomes. So I wanted to bring her in and talk about that.
We have an interesting conversation. It kind of goes all over the place. She defines AI very broadly, and so you’ll hear us talk about that. Mostly, we end up getting into talking about things like radiographic interpretation using AI. Things like that, that are right around the corner. Things that she’s already using in practice. I just think that’s really interesting.
So if you’re interested in artificial intelligence, how rubber is going to meet the road, and what we’re going to be seeing in practice as far as getting better patient outcomes, this is a really good episode. But anyway, it’s a fun conversation. I hope you guys will enjoy it. Let’s get into this episode.
Kelsey Beth Carpenter:
(singing) This is your show. We’re glad you’re here. We want to help you in your veterinary career. Welcome to The Cone of Shame with Dr. Andy Roark.
Dr. Andy Roark:
Welcome to the podcast, Dr. Jen Chatfield. How are you?
Dr. Jen Chatfield:
I’m fantastic, Andy. How are you?
Dr. Andy Roark:
I’m great. I’m so glad to have you here. I’ve already been enjoying talking with you, because you do a lot of the weird things that I do. You have a YouTube show called Chats with the Chatfields. I became aware of you not from Chats with the Chatfields, but because I saw a presentation you were doing through VetFolio. You were talking about using artificial intelligence to improve clinical outcomes.
Dr. Jen Chatfield:
Yes.
Dr. Andy Roark:
I looked at it and I looked at the walkthrough, the agenda, the points that you were making. I was really impressed with how practical and pragmatic your approach was when you were talking about something that tends to be a lot of hand waving right now.
A lot of people are really excited about it, but you actually had insight as far as use cases and ways that people could immediately pick this up and actually make their days better. That’s what I wanted to bring you on today and talk about. Can you open up for me? Just at a high level, how do you look at artificial intelligence as a veterinarian? What is your larger picture or philosophy? Is it a tool? Is it a partner? What is it? Where are we going?
Dr. Jen Chatfield:
Well, I hate to disappoint listeners. Because that was a super high-level philosophical question right there.
Dr. Andy Roark:
That’s kind of how I go.
Dr. Jen Chatfield:
I was very deep and intense, and I’m not necessarily that person, to be honest with you. If there is any tool, piece of equipment, piece of information or philosophy that can help me raise the standard of care that I’m able to provide … I’m for that. I guess I approach it like that. “How can I be a better vet?”
Dr. Andy Roark:
No. I like it. Well, that’s it. I kind of did you wrong when I was like, “I’m interested in pragmatism. Let’s talk philosophy.” You were like, “I didn’t come here for that, Andy.”
Dr. Jen Chatfield:
I’m totally happy to wax philosophical as often as the next person, but I really appreciate data. And if I’m going to change something I’m doing, which is one reason that I feel so lucky that I accidentally became a veterinarian … Because if I’m going to change something I’m doing, then I want to either know that I’m changing it because I want to or because I have data that supports that change.
For me, that is where a little bit of my fascination personally with artificial intelligence, with the robots and the machines comes from is … I don’t know. Is it better? Is it worse? Is it the same? And then, from that point, if we answer that and we say, “It’s better.” Well, then there’s more questions. Well, what does better mean? How do you define better? For me, overall, that’s clinical outcome. Hands down.
Dr. Andy Roark:
Got you.
Dr. Jen Chatfield:
Clinical outcome. What we learned is the only reason to do a diagnostic test is if the outcome is going to change.
Dr. Andy Roark:
If you’re going to use the outcome. Right.
Dr. Jen Chatfield:
I should phrase that better.
Dr. Andy Roark:
No. No. I see what you’re saying. Exactly.
Dr. Jen Chatfield:
Not as a medical lookie-loo.
Dr. Andy Roark:
It’s always that question of, “What are we going to do with this information?”
Dr. Jen Chatfield:
Yes.
Dr. Andy Roark:
If the answer is, “Nothing,” then do we really need to run that test? Yes. I’m 100% onboard with that. Got you.
Dr. Jen Chatfield:
For me, that’s the same thing with, “Is whatever I’m doing, whether it be related to AI or not, going to change what I’m doing with this patient?”
Dr. Andy Roark:
Well, start to lay it down for me. What does that actually look like in practice? Again, there’s a lot of people who get really excited about AI. I’m not running into a whole lot of people who are like, “This is what I do in practice.” Talk to me about that.
Dr. Jen Chatfield:
There’s a lot of, like you say, hand waving. People losing their ever-loving minds talking about AI, “Are the machines going to take over?” And this and that. While I do not currently live in fear of the machine I’m talking into right now, I do live in fear of taking the human component out of any job that’s focused on care, because you just can’t convince me they can do it better.
Dr. Andy Roark:
Right.
Dr. Jen Chatfield:
I’m fascinated by that interface and how we use it. So in regular practice, in a very sneaky way, it’s already there. I go into an exam room, a dog is sick, vomiting and diarrhea. Dr. Jen the Vet loves to see vomiting and diarrhea. I say, “You know what, little fluffy white dog with strawberry jam stool? I’m concerned that we have …” It’s not called HGE anymore. What do they call it? Like ADS or AHD? It’s got a new thing.
Dr. Andy Roark:
I still call it HGE. I’m old school like that.
Dr. Jen Chatfield:
Thank you. Appreciate you. Appreciate you. So I say, “I want to roll that out. Let’s do some blood work.” So I get some blood work. I’m going to run a CBC and I’m going to run a chemistry. I put my CBC in the machine. I don’t do a manual CBC. I don’t do a manual differential. Do you?
Dr. Andy Roark:
No. I mean … No. No.
Dr. Jen Chatfield:
Hey. Neither does our doctor. We put it in the machine and the machine kicks that out. That’s AI. That is AI, because it’s scanning, looking for cells approximately this size, whatever other characteristics, and then it’s calling that a neutrophil and kicking out a differential. Or it’s calling that a red blood cell and kicking out PCV plus or minus reticulocytes.
That’s AI. You’re already using it, friends. You’re already using AI. And then, obviously, we have a chemistry going on. Another way I think it sneaks in is when we run urinalysis. All of these tests that we use benchtop in clinic screening tests that we use, that all rely on what I perceive as color. Urine dipstick, I’m looking at you. All these color changes.
As a friend who’s a hairstylist told me, “Everyone sees color differently.” We’re looking at this dipstick with a bajillion colors, and I don’t know about you, but I’m like, “I’m holding it right here.” I’m turning the canister as the seconds tick by, and I’m like, “Nope. Not that color. Maybe this color?” Hey, guess what? That is open for error with interpretation of the color changes. What do we do now?
Dr. Andy Roark:
Our dipstick reader.
Dr. Jen Chatfield:
We have a system. And that removes that subjective interpretation error. That’s AI.
Dr. Andy Roark:
Okay.
Dr. Jen Chatfield:
Another piece. This is my favorite piece, because I am the fluids doctor. I am not the blood work doctor. I am not the x-ray doctor. No one has ever accused me of being a surgeon. I love that we now have sophisticated AI available on the market that will support interpretation of diagnostic images. X-rays. Love it. Love it.
That’s probably the scariest one for people, because we’re accustomed to blood work. Just trust in that machine. But it’s reading a picture. “It’s reading a picture, Dr. Jen. What the heck?”
Dr. Andy Roark:
I think that stuff is super exciting. I think that stuff is really cool. I think it’s going to open up doors for us to do ultrasounds and things like that in the clinic. There will be a referral part of it, where you say, “Well, let’s get a second pair of eyes on this.”
It’s not something I would race to right away. I think there’s going to be a long learning curve, but I fully believe that we’re going to be feeding our radiographs and then our ultrasounds and things like that into AI and just getting our results right back.
Dr. Jen Chatfield:
Dr. Andy, I’m already feeding my radiographs to AI. Every radiograph.
Dr. Andy Roark:
Interesting.
Dr. Jen Chatfield:
Every radiograph. Now, the better question is ask me, “Why? How come?”
Dr. Andy Roark:
Okay. Tell me.
Dr. Jen Chatfield:
Because the first question I had naturally a skeptic. I’m Texan. I’m naturally a skeptic. The first thing, I was like, “I wonder how often it’s wrong?”
Dr. Andy Roark:
Of course.
Dr. Jen Chatfield:
I wonder how often I’m wrong? And so, I wanted data. Because if I’m going to start adding this component in, what does the data say? Oh my gosh. The data is amazing. It’s incredible. First of all, you have to accept that … Again, I’m not banging on radiologists at all. I would love to have a radiologist in my pocket. I would love it.
Again, I was not the x-ray doctor. I don’t like it. I like x-rays for counting puppies and kitties. So it turns out that is partly because I’m human. The diagnostic error rate for radiographic interpretation ranges anywhere from 30% to 65% or 70%.
Dr. Andy Roark:
Okay.
Dr. Jen Chatfield:
Let me give you a specific data. So it’s very different in human land, because radiologists read only one body cavity. That’s all they do. Or one body part. In vet med, we’ve got to read it all. We take catograms for heaven’s sakes. Or at least, I do. You can put your nose up in the air if you don’t, but I do.
So if I’m looking at this x-ray … There was a study that was published fairly recently. They had four veterinarians. They had two that were reported radiologists and they had two that were surgeons. I think that’s fair. Surgeons … They look at tons of x-rays.
Dr. Andy Roark:
Sure.
Dr. Jen Chatfield:
They analyzed how often they were accurate when they were looking at images to determine intervertebral disc disease. Whether it was extrusion, protrusion, viscosity, whatever. IVDD. That’s what they were looking at. They were accurate 30% of the time. Now, when they told them clinical context, then they bumped up their accuracy to … I think it was almost 60%. That’s fair.
Dr. Andy Roark:
I think it’s fair.
Dr. Jen Chatfield:
Okay. 30% with four people who were boarded specialists who spend a lot of time in the dark room.
Dr. Andy Roark:
That’s brutal.
Dr. Jen Chatfield:
Or used to be in the dark room, I guess. I’m dating myself by saying that, but I’m like, “Well, what chance does Dr. Jen the Vet have?” It’s tricky. And then, if you look at the perceptual bias and the cognitive bias that’s available for our human minds. They looked at human radiologists. Mammograms is where this was all pioneered, because it’s one of the most difficult tissues to image well and to a diagnostic standard.
They looked at this. They discovered that a conversation happening across the room while a human radiologist was looking at images, they’re not in the conversation. Consciously, they’re not listening. It can impact their diagnostic interpretation.
Dr. Andy Roark:
I believe that. That totally makes sense. Our minds are constantly picking up the things around us. Even if we don’t want to be influenced by them, we are. I could 100% see someone having a conversation about a different case and influencing what you see as you look at this radiograph.
Dr. Jen Chatfield:
Yep. I could nerd out with more. There’s more data, but it just scares you. It just tells you this is a crapshoot. So if you go to cut section imaging, this is one of my favorite things. I have a friend that’s a human surgeon. She’s like, “I need an MRI for a surgical plan. I need an MRI.”
I’m like, “Sure. No problem.” That’s basically a crapshoot with interpretation. When you look at the error rates on interpretation of cut section imaging like CT and MRI … On the human side, for radiologists, if it’s 3% to 5% on plain films, it goes up to 30% or 40% on cut section imaging. Again, when we get down to the practicality of it, you could flip a coin and you’re 50/50.
Anyway, I’m very excited for AI. Now, I don’t always agree with it. Because again, it’s the same thing as … And this is why you can’t replace the veterinarian friends. How often do you look at a patient and you get the blood work back and you say, “Yeah. No.”
Dr. Andy Roark:
Well, it’s another maxim from vet school was, “First, look at the patient.” It was the whole hammer on you that the most valuable diagnostic tool you have is your physical examination. Putting your hands on the pet is the most valuable thing that you can have. I think that there’s a lot of wisdom there.
Dr. Jen Chatfield:
Frequently, the technicians that work with me, they know I say this. I’ll be like, “Nope. Nope. I don’t want this blood work. Give me a different one.” Even though it’s for that same patient. I’m like, “No, no, no. This is not it.”
Because the white count will be 8,000 or 9,000 and I’m looking at a dog that, clearly, I was expecting septic. It was going to be either through the roof or it was going to be in the basement, but it wasn’t going to be right in the middle. And so, I proceed with what I think.
It’s the same thing when I get this newfangled thing, this radiographic interpretation from the robots, from the machine. I look at it and say, “I don’t think that’s true. No. No. I see this here.” That’s okay. And I tell clients that, because that’s the hardest thing too. How do you talk about that with the client? I’m disagreeing with artificial intelligence? Yep. Yep. I am.
Dr. Andy Roark:
Hey, guys. I just want to jump in here real quick with a nice little bit of CE love for you. That’s right. On September the 5th at 4:00 PM Eastern time, 1:00 PM Pacific, I’ve got a webinar that I am putting on. I am hosting it for my good friend, Dr. Andrea Eriksson.
She is a board certified veterinary cardiologist, and she’s doing a webinar called CARDALIS: When to Utilize RAAS Suppression and Why. Guys, this is all about the new drug CARDALIS, which is out. If it’s not in your practice yet, it will be. This is talking about what CARDALIS does, which patients should be considered for it, and why suppression of the RAAS system should be utilized in what cases.
Anyway, that’s it. It does have an hour of RACE CE attached to it. It is going to be free for you guys. I’ll put a link in the show notes, but come get some cardiology CE with me and my friend, Dr. Andrea Eriksson. Guys, let’s get back into this episode.
Do you think that the rate of improvement is going to be pretty high? Do you think that you’re going to continue to see … The number of times when we don’t match up. Is that going to go down and down? Or is this going to be a constant thing, do you think?
Dr. Jen Chatfield:
I don’t know. I don’t know. I only know my own experience thus far, which is, it’s not that often that I will disagree with the artificial intelligence interpretation of the image. What it does for me, it’s more sophisticated than that. It’s not a right or wrong. It’s the fact that the artificial intelligence looks at the entire image.
For me, when I have clinical context … Because it doesn’t care. It doesn’t care that you’re looking at a Frenchie. It doesn’t care. And that heart looks giant and you’re like, “No. I dismiss it.” One of the most common perceptual errors that we make in radiographic interpretation is wrong decision. We search and we find it … Or I’m sorry.
We search, we don’t find it. We search and we see it, but we don’t recognize it. And then, the final thing is we search, we recognize it, we actively decide it’s not abnormal. Or we make the wrong decision. 60% to 80% are perceptual errors like that. So if I’m looking at this and I look at a frenchie, I’m like, “God. That heart looks giant. It looks super globoid.” Alan Spear, if you’re listening, globoid is a real word. It looks super globoid, but it’s a Frenchie, so I’m sure it’s fine.
That’s wrong. My AI isn’t doing that. It doesn’t know it’s a Frenchie. It doesn’t care. It’s calculating vertebral heart score and going to tell me what it is. It doesn’t care. So it finds things like that, and it actually helps me. At least in my experience, it’ll help me with a more sophisticated interpretation in order to either more proactively encourage a visit today to the neurologist … Even though your dog is walking around.
Dr. Andy Roark:
We’re all so strongly affected by our own personal experiences. You have those cases and you say, “Oh my God. I did this.” Or, “One time, I missed this.” And then, you’re looking for it. If you’ve been a vet for any time, you have your own biases.
You’ve had medications that a pet had a side effect to. Now, you’re just like, “I hate that medicine.” The truth is, the data on that is great. The safety numbers are wonderful, but you had the one in a million side effect, and now you never want to do it again.
Dr. Jen Chatfield:
Please. One of my favorite pathogens is leptospirosis. I go all around the world talking about leptospirosis. The metadata says that leptospirosis as a component of your vaccination is not … It doesn’t say, “It could be.” It is not a risk factor for adverse reaction.
Millions and millions of doses and dogs. And I still run into lots of veterinarians who are like, “I’ve seen it once.” It’s hard to get over. I don’t blame them. It’s super scary.
Dr. Andy Roark:
It is. It’s the way that we’re made. When something scary happens to you, you hold onto it. It’s wired into your being that this is a thing that you’re going to remember. It’s really hard. But that is the benefit of having AI or outside data that comes in, when you can say, “I have my biases. There are not biases here.” Or they’re not the same biases as the one I have.
Dr. Jen Chatfield:
Right. Exactly. They’re different. They’re not better or worse. They’re different.
Dr. Andy Roark:
Jen, when you look into your crystal ball five years from now, how does it look different from today? Are we interacting with AI in a different way? Do you think it’s going to be about the same? What does practice look like? Five years is not a lot of time. This is right around the corner.
Dr. Jen Chatfield:
It’s not. It’s really not. My hope is that we are, because I feel like it’s rapidly becoming such a powerful tool. But the difference will be whether or not … When you grow up as a veterinarian, you grew up in vet school. Whatever you learn in vet school is the way we do it.
This is the way. That’s your Mandalorian experience, “This is the way.” I feel like if vet schools don’t begin to embrace some of these tools, then it’ll be a harder struggle for the industry. But I think that they’re so cost-effective. I do think they enhance our ability to provide care at a level that now our clients are expecting. I think if we’re going to see more and more AI, I don’t think in a rapid acceleration.
I don’t even think it’s going to even enter the realm of replacing a veterinarian. I can order a hamburger from a kiosk in a screen … My expectation of that hamburger is a little different, but I cannot hold my pet up to a camera and truly get that diagnosis in a sophisticated way. And I don’t think we’re going to be able to do that.
Dr. Andy Roark:
I love it. All right. Dr. Jen Chatfield. Dr. Jen the Vet, where can people find you online? Where can they learn more about you?
Dr. Jen Chatfield:
Well, thanks so much for asking, and again, thank you for having me on your show. It’s my first time on your show. It’s been wonderful. If you can’t find me on the internet, you’re just not trying. You can find me at chatfieldshow.com. You can find me at drjenthevet.com. You can find me on the YouTube. Almost anywhere. And if you want to find me IRL, I’m usually speaking at whatever conference you’re going to.
Dr. Andy Roark:
Awesome. Thanks for being here. Guys, thanks for tuning in today. Take care everybody. And that’s it. That’s what I’ve got for you guys. Thanks so much for being here. Thanks to Jen for coming in.
Guys, if you enjoyed the episode, let me know. Leave me an honest review wherever you get podcasts. That’s the nicest thing you can do for me. It’s how people find the show. That’s it. To anyone, just be well. Take care of yourself. I’ll talk to you later on. Bye.