Graham Walker, MD

linkedin.comAlways Open. Already Full.Why I'm leaving emergency medicine and worried about its future “I was drunk last night and accidentally drank a bottle of Suave peach conditioner because I thought it was juice” is not a chapter in any medical textbook, but in emergency medicine, I’d better know what to do about that. (Yes, this wa

When I told my ER group of 15 years I was leaving, the most common response wasn't "we'll miss you." It was 𝘤𝘰𝘯𝘨𝘳𝘢𝘵𝘶𝘭𝘢𝘵𝘪𝘰𝘯𝘴. www.linkedin.com/pulse/alway... I haven't been able to stop thinking about that. So I wrote about it.

View full thread

And outside our own ER break rooms? I'm not hearing the alarm bells I'd expect from our emergency leaders, chairs, and researchers. So I guess I'll pull the fire alarm myself.

If you're an ER doc, you're married to one, you run a hospital, or you might someday need an emergency department (so… everyone), this one's for you. Full essay above. It's the most honest thing I've written about a job I loved for 18 years.

Here's what I'm hearing from the inside. Physicians at some of the country's best-known academic EDs won't work shifts there anymore. Colleagues barely three years out of residency are already building exits: admin, informatics, industry, part-time, anything. They haven't stopped caring.

These are people who survived residency, and they're wondering whether the highlight of the next 30 years will be managing dysfunction. 𝗔 𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝘁𝘆 𝗶𝘀 𝗶𝗻 𝘁𝗿𝗼𝘂𝗯𝗹𝗲 𝘄𝗵𝗲𝗻 𝗹𝗲𝗮𝘃𝗶𝗻𝗴 𝗶𝘁 𝘀𝘁𝗮𝗿𝘁𝘀 𝘁𝗼 𝗹𝗼𝗼𝗸 𝗹𝗶𝗸𝗲 𝘀𝘂𝗰𝗰𝗲𝘀𝘀.

This isn't a burn-it-all-down essay. I love this specialty. I love the nurses, techs, pharmacists, and docs who stand up when someone asks "is there a doctor on the plane?" That's exactly why I can't pretend everything is fine.

I agree with RFK Jr. (Partly.) On AI and medicine this week, he said: "You may have a medical record that is 1000 pages long. You have 6 minutes with a doctor today. He is not gonna be able to review it, but the AI can."

View full thread

It's weird to me to point at a machine, built by engineering experts, trained on knowledge from medical experts, and then say we're freed from the tyranny of expertise. "Trust the experts" doesn't disappear. It becomes 𝘵𝘳𝘶𝘴𝘵 𝘵𝘩𝘦 𝘮𝘰𝘥𝘦𝘭 — which is, theoretically, experts... compressed.

AI will make numerous scientific and medical breakthroughs in my lifetime. It may radically democratize access to expertise. It will find patterns that are nonsense, promote noise over signal, and lead us on billion dollar wild good chases. Making sense of all of it? That requires 𝗲𝘅𝗽𝗲𝗿𝘁𝗶𝘀𝗲.

AI didn't discover that antibiotics treat bacterial meningitis. It didn't run the randomized trials, or spend decades figuring out which cancer treatments improve survival, or figure out how the kidney works. Or discover vaccines. Or the Starling curve.

Or that tick-borne diseases often causes fevers with headaches and rashes. Human EXPERTS did. Generations of them. Your AI model can explain all of it remarkably well... because it's trained on EXPERTS. Not just Reddit threads.

1️⃣ Reading does not equal "better informed," and "better informed" isn't the same as deciding better. Medicine isn't a trivia contest. The hard part is deciding what *matters*: which abnormality deserves attention, which diagnosis is unlikely enough to ignore.

Sometimes expertise is knowing what NOT to pursue. 2️⃣ Where exactly does he think the AI learned medicine?

That's a business-model problem, not a knowledge problem that I'd love his attention on, as HHS Secretary.)

But then he makes a leap: AI can give you a second opinion "much better informed than any doctor in the country." And then: "We're never ever again gonna be able to be dominated by public officials who tell us, 'Trust the experts.'" You lost me there, Secretary.

Totally agree. AI can read a 1000 pages faster than me, and can retrieve more information than I'll ever hold in my head. That "feature set" can be incredibly powerful in many circumstances. (He's not wrong about "6 minutes," either.

The thing that irks me is that AI sounds exactly like a medical student: fluent, certain, and not yet wrong enough times to know better. That's fine in a med student, because med students have residents, fellows, and attendings to correct them. AI doesn’t have any of those.

View full thread

And sometimes just stop, because "I don't know" is one of the most important sentences in medicine. (Patients hate that one, too.) But that’s what lets you get another test, ask another doctor, or observe the patient longer.

The best doctors I know act decisively while quietly holding the awareness they could be wrong. That tension isn't a flaw in clinical judgment. It IS clinical judgment. Biology taught us that. AI just hasn't been smacked in the GPU by it yet.

When a med student presents with total confidence, the attending says, "Okay, what else could this be?" Not AI. Sure, you could try the surface-level fix and prompt it to "sound less sure."

The harder problem is getting AI to 𝘣𝘦𝘩𝘢𝘷𝘦 like uncertainty matters. Ask for the missing fact before answering. Name the one thing that would change its mind. Offer competing explanations instead of collapsing to one.

AI has the fluency but none of the battle wounds.

Us humans? We’re leaking uncertainty 𝘤𝘰𝘯𝘴𝘵𝘢𝘯𝘵𝘭𝘺: we pause, slow down, hedge, stop mid-sentence while our brain works something out. AI flattens all of it. A well-supported answer and a wild guess arrive in the same streaming cadence and grammar.

Biology loves to smack you across the face right when you start feeling like a pretty good doctor. So your language changes: 𝘗𝘳𝘰𝘣𝘢𝘣𝘭𝘺. 𝘐 𝘵𝘩𝘪𝘯𝘬. 𝘈𝘭𝘮𝘰𝘴𝘵 𝘯𝘦𝘷𝘦𝘳.

Those phrases sound weak to people outside medicine, but I know them as dead giveaways of how experts talk. The experienced physician isn't just accumulating facts, they're accumulating reasons not to trust themselves too much. (Yes, some attendings never got the memo. Hi, ego.)

No shade to med students — I know that voice of confidence, because it was mine, too. We’ve all been there.

Then residency happens. You see weird presentations of common diseases. You watch real, terrible things happen to normal people — things no textbook warned you were even possible, let alone probable — and whatever innocence you had left goes with them.

Everyone finds AI's confidence annoying. But I think physicians find it dangerously familiar. We were trained to treat that voice as a hazard, we learned why on real patients, and now we're the ones expected to supervise it.

Sometimes I’ll hear a confident medical student presentation, and my initial reaction is “Damn, they’re really cocky.” But then I remind myself… they just haven’t been burned by biology yet.

linkedin.comWhat Do You Mean by Meaning?Reflections on Why I Stopped Practicing Medicine, For Now I’ve done my talk therapy via laptop-propped-up-on-shoebox ever since COVID. It wasn’t long ago — February 2025, perhaps? — that my therapist casually asked me a seemingly simple, therapy-ish question: What gives your life meaning? Before goi

𝗥𝗲𝗳𝗹𝗲𝗰𝘁𝗶𝗼𝗻𝘀 𝗼𝗻 𝗪𝗵𝘆 𝗜 𝗦𝘁𝗼𝗽𝗽𝗲𝗱 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗶𝗻𝗴 𝗠𝗲𝗱𝗶𝗰𝗶𝗻𝗲, 𝗙𝗼𝗿 𝗡𝗼𝘄 This all started with a seemingly simple, therapy-ish question my therapist asked me in early 2025: What gives your life meaning? www.linkedin.com/pulse/what-...

every.toAfter automation: AI will make uncertainty cheap to explore without making it cheap to resolveGraham Walker, Emergency physician — AI can surface every rare possibility a doctor once considered and dismissed. The hard part for patients won't be finding information—it'll be knowing when low risk is low enough.

23 Sept

🤔 𝗔𝗹 𝘄𝗶𝗹𝗹 𝗺𝗮𝗸𝗲 𝘂𝗻𝗰𝗲𝗿𝘁𝗮𝗶𝗻𝘁𝘆 𝗰𝗵𝗲𝗮𝗽 𝘁𝗼 𝗲𝘅𝗽𝗹𝗼𝗿𝗲 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗺𝗮𝗸𝗶𝗻𝗴 𝗶𝘁 𝗰𝗵𝗲𝗮𝗽 𝘁𝗼 𝗿𝗲𝘀𝗼𝗹𝘃𝗲. This was a super fun opportunity to think more about the future of healthcare, AI, and automation with Every Inc. Thank you Dan Shipper for the opportunity! My After Automation thesis statement: every.to/thesis-stat...

21 Sept

Could not be more proud to both be interviewed by @acepnation.bsky.social with my co-founder Dr. Joe Habboushe about @mdcalc.com, innovation, and helping physicians take better care of patients. See you at ACEP Scientific Assembly in October! www.acep.org/sa 🎙️ Podcast episode in the comments ⤵️

blog.mdcalc.comThe Future of Clinical Decision Support Is SilenceWe’ve been ignoring the lesson for twenty years. Now the stakes are higher.The Future of Clinical Decision Support is a new series from MDCalc exploring how AI, evidence, and physician judgment are reshaping medicine. These essays are intended to spark discussion about where technology is taking clinical decision support – and where physicians feel it should go. Blaring “Mary Had a Little Lamb” every 4–19 minutes from 2am to 4am in a fluorescent-lit ER has only two outcomes: insanity, or alert f

I have a deranged hatred for 🎵Mary Had a Little Lamb🎵. The ChatGPT image generated for my article is just as deranged... and terribly accurate. blog.mdcalc.com/post/the-fu...

And now we’re building AI that can generate a warning, recommendation, caveat, differential, or “helpful” suggestion about basically... everything. Ruh-roh, Scooby.

I wrote about why I think the future of clinical decision support is actually 𝘀𝗶𝗹𝗲𝗻𝗰𝗲: AI actually doing far MORE in the background, while having the judgment (and restraint) to interrupt us far LESS.

I have a deranged hatred for 🎵Mary Had a Little Lamb🎵. The ChatGPT image generated for my article is just as deranged... and terribly accurate. blog.mdcalc.com/post/the-fu...

Somehow this trauma has become my philosophy of clinical AI.

For 20+ years, clinical decision support has taught us a pretty important lesson: 𝗶𝗳 𝘆𝗼𝘂 𝗶𝗻𝘁𝗲𝗿𝗿𝘂𝗽𝘁 𝗰𝗹𝗶𝗻𝗶𝗰𝗶𝗮𝗻𝘀 𝗼𝗳𝘁𝗲𝗻 𝗲𝗻𝗼𝘂𝗴𝗵 𝘄𝗶𝘁𝗵 𝘁𝗵𝗶𝗻𝗴𝘀 𝘁𝗵𝗮𝘁 𝗱𝗼𝗻’𝘁 𝗺𝗮𝘁𝘁𝗲𝗿, 𝘁𝗵𝗲𝘆 𝘄𝗶𝗹𝗹 𝗹𝗲𝗮𝗿𝗻 𝘁𝗼 𝘀𝘁𝗼𝗽 𝗹𝗶𝘀𝘁𝗲𝗻𝗶𝗻𝗴.

I have a deranged hatred for 🎵Mary Had a Little Lamb🎵. The ChatGPT image generated for my article is just as deranged... and terribly accurate. blog.mdcalc.com/post/the-fu...

For years, the jingle was the song our ER bed alarms played whenever a high risk patient tried to climb out of bed. At 3am, hearing it for the 700th time that night, you start to wonder how many marbles you're going to have left when the sun comes up.

linkedin.comBack to Where I Started... 21 Years LaterSoooo, ha, um, I have a secret I’ve been keeping from you: I’m joining MDCalc full-time as its Head of AI. I wish I could say that I rage-quit medicine or that I’ve got some incredibly juicy story to tell, but alas.

🤬😡🤬 I wish I could say I rage-quit medicine, but alas. There was no final terrible shift. No stethoscope tantrum. Nobody pushed me out. www.linkedin.com/pulse/back-...

I think medicine is at one of those rare moments when a lot of what comes next genuinely hasn’t been decided yet. AI could make medicine dramatically better to practice. It could also become the most powerful technology we’ve ever invented for asking fewer people to do even more work.

I want to help shape which future we get. I'll do a series of articles over the next few months about what's changed, before ending with what I think medicine needs to become. But, first up: What I'm gonna do with MDCalc! Click to read the article below. www.linkedin.com/pulse/back-...

🤬😡🤬 I wish I could say I rage-quit medicine, but alas. There was no final terrible shift. No stethoscope tantrum. Nobody pushed me out. www.linkedin.com/pulse/back-...

Instead, after 18 years in emergency medicine, I’m stepping away from clinical practice for now, leaving the incredible Permanente Medicine — and joining @mdcalc full-time, the lil' ol' website I started as a medical student 21 years ago.