How my son’s sore knee made me think about the role AI can play in veterinary practice.
About six weeks ago my son came home from school with a sore knee after a particularly energetic PE class. Not ‘something’s ruptured’ sore, just enough to warrant a week of rest and NSAIDs. But halfway through rugby training a week later he had to limp off the field, the knee now worse than before. I had a cursory feel of his leg that night, confirming that it was indeed a bit swollen. If he was a dog he’d be an un-castrated kelpie, so if this was anything that would required actual structured exercise restriction, we’d need some evidence to strengthen our case. So off the the doctor he went.
What the doctor thought…
Doc number one agreed on the swelling and said “I think he has patellar bursitis – let’s send him for an ultrasound.” Great plan. But the scan was a week away and we had to decide whether he should skip the big match that weekend, so I needed to understand the implications of our suspected diagnosis.
Now, until recently I would have just Googled patellar bursitis. But I no longer Google – I AI, or GPT – or whatever we’re going to call this new thing we’re doing. I told ChatGPT the history and signalment and asked it to give me a rundown on patellar bursitis.
What ChatGPT thought…
“I don’t think he has patellar bursitis.” I told Ros. She rolled her eyes at me – twice. Once, because she always does when I try to diagnose the kids, the second time because she’s getting tired of hearing “ChatGPT says…”
Ros took him to the scan. “It’s not patellar bursitis”, the ultrasonography said.
Vindicated, that night I took off my ‘concerned parent’ hat and went into diagnostician mode. “Lie down. Relax your leg. Does it hurt here? How about here?” (Let me tell you – this shit is MUCH easier in a patient that can talk!)
I went back to GPT and told it exactly what I’d found, using appropriate medical terms and specific detail. I recapped signalment, how he injured it, an which activities caused him most pain. “12 year old boy, injured during a jump. Pain on walking up stairs.”
“I’m sorry to hear about your son’s knee. Thank you for the detailed explanation. Based on what you told me I’d consider….”
Chat went into a long explanation, listed dd’s etc, but its top diagnosis was something called “Sinding-Larsen-Johansson Syndrome”.
“Random” I thought. “Probably not.”
Back to the doctor we go
So back to doctor to discuss scan results. Fast forward through the rest of the process: rads to rule out a chip #, another appointment. Four weeks in, still a problem: “let’s send him for an MRI.” Get the MRI done. Wait another week for results.
Last night Ros has another appointment to discuss the MRI findings:
Doc: “He has a condition called Sinding-Larsen-Johansson Syndrome. I’ve never heard of it.”
Ros: “No ways. My husband used ChatGPT 4 weeks ago to ask about his knee, and that’s what the answer was.”
Doc: “Well there you have it. Let’s see what it says about treating it…”
An anecdote is the worst kind of evidence, but..
It’s an n of 1, and anecdote is the worst kind of evidence, but f me! So here’s what I’m thinking about AI:
1. It’s not coming for your job
Your average parent (or pet owner in our case) could not have used ChatGPT to come to that diagnosis. My hands have been sniffing out pain and finding subtle swelling in uncooperative and uncommunicative patients for 25 years. I’m very good at that bit, and I don’t think any AI is replacing that part any time soon.
But it definitely ‘knows’ more than most of us, just like it knew more than my kid’s GP.
Research (and my personal experience) has shown that when it comes to medical knowledge, it’s not that great with non-specific histories. Don’t tell it ‘my son’s knee hurts. What’s wrong with it?’ Or – ‘my canine patient has come on with sudden collapse. Give me some dd’s.’ But narrow down its context and it suddenly becomes laser sharp.
2. Our days of bullshitting clients about stuff we’re not that up to date on are over.
Clients are going to check what you are telling them. And if they’re even half-sensible they’ll get great answers. This is confronting. But I think we need to shift from feeling threatened or undermined by that and embrace it. “Show me what you found. Let’s explore that together. In this context, here’s why I don’t think that makes sense. Here’s what I found. Wow, I’ve never even heard of that before – let’s find out more.”
(Oh, the treatment for Sinding-Larsen-Johansson Syndrome? Rest and anti-inflammatories!)
This post first appeared in The Vet Vault 3.2.1 email. You can subscribe to this weekly email HERE
