No 98: Heidi Hi Patient. How can the AI assisted HCA help you today?
Do you really need to see a doctor?
I'm sure Many of you are already bored of hearing about how AI will change medicine.
However, the machines are coming, whether we like it or not.
Heidi is not registered as a medical device. Therefore, it is not allowed to suggest medical plans or diagnoses. It “merely” transcribes what you say and summarises it.
I have no doubt that it is already capable of some much more and is already learning from us with every use. I suspect they will soon have to undertake the trials required to become a registered medical AI device.
Once that happens, what is to stop it taking over what we do. Bare with me. This is what I can see happening.
Heidi + Google Glass + POCUS + phone + HCA = Cyborg medicine
https://x.com/i/grok/share/Oiezj51qS0pmQMAZazBJmn91GHeidi gets embedded within something, like Google Glass. A healthcare assistant (HCA) could put on the glasses call in the patient and have a guided consultation. The HCA reads off a script on the screen. Asks the questions the AI suggests. Video and audio is recorded. The glasses highlight where to put the POCUS to visualise a heart murmur or an infection in the lungs. The AI combines all the data and spits out a diagnosis and management plan.
This technology already exists.
Primary care triage could be done almost completely without a GP.
After this consult if a second opinion is needed or the patient provides confusing symptoms then they can be booked with a AI assisted GP.
I think that is not very far in the future. For many people, Heidi and Healthcare assistant, will be enough for them to feel reassured.
Might this work even better in secondary care services? Could this mean surgeons no longer seeing patients in clinic and just doing operating?
https://youtube.com/shorts/bIjU20CDmp0?si=oocrp6T5CTN5dGq9As always, let me know what you think. I enjoy reading your comments and emails. Please do like and share.
The modern wisdom podcast recently talked about AI adoption and exponential curves. On the point they were talking about is how humans are really bad at judging exponential curves.
We often judge things by what things look like in the recent past and then predict the Future as a straight line. And that's not how an exponential curve works.
AI is getting better exponentially. This stuff will rapidly change how we work.
As an anesthesiologist (albeit retired) I've been far from being a luddite all my life (built a ruby pulse laser when I was 10, starting coding in 1968 at age 18 using Fortran IV with the Watfor Compiler). Before I became aboard certified anesthesiologist, I practiced as a GP-surgeon. My office was one of the first computerized offices in the state of Nevada — we're talking 1979. I'll give you one example where AI could be fooled. Decades ago, I had 30-year old male complaining of a sore throat. Period. But he looked very slightly off to me. I'd bet AI would not have picked that up because the database of images would not have been sufficiently trained to discern his look. In any event, I got some blood work, which would not have been protocol for a complaint of a sore throat. It ended up he had the 5th case of some bizarre collagen disease, diagnosed at Stanford University where I sent him for consultation. When he finally came back from Stanford, I managed his meds in consultation with the folks there.
I have no doubt that AI will be a useful tool for many aspects of medical practice, even for specialties as anesthesiology. But there will be things AI can't do for a very long time to come. Probably the most important is human-human interaction.