02/12/2025
đź’ Everyone says AI is unsafe in healthcare.
But what if humans are just as unsafe - and no one wants to admit it?
We’re told:
“Don’t trust AI for medical advice- it gets things wrong. Stick with professionals.”
Fair.
But here’s the uncomfortable truth: humans don’t give perfect care either.
🩺 A Personal Test
I gave ChatGPT a common scenario: acute low back pain.
Bread-and-butter medicine. Clear evidence. Clear guidelines.
I compared its advice to Australia’s Low Back Pain Clinical Care Standard (2022).
📊 The Result?
No hallucinations. Just… human errors.
What it got right:
- Screened for red flags
- Avoided routine imaging
- Encouraged staying active
That’s standard care - nothing alarming.
Where it fell short:
- Recommended paracetamol (no better than placebo)
- NSAIDs (without enough cautions)
- Suggested stretches, heat/ice (weak evidence)
Not dangerous. Just imperfect -like us.
📉 The Harder Truth
A review of 200,000 patients across 7 countries found:
- 1 in 4 got unnecessary imaging
- Up to 60% were prescribed opioids
- Only 1 in 5 got evidence-based advice
We didn’t create guidelines because care was perfect.
We created them because too often - it isn’t.
Clinicians aren’t incompetent.
We’re tired, rushed, influenced by old habits and system pressures.
We already tolerate imperfection from humans.
âť“ So Which Is More Dangerous?
Yes, AI shouldn’t replace clinical judgment.
But the status quo has risks too:
- No care
- Non-evidence-based care
- Unnecessary imaging or opioids
Maybe the question isn’t:
“Is AI safe enough?”
But:
“Can AI help us deliver better care - more consistently?”
Because the gap between ideal and actual care is real - and harmful.
đź’¬ Your Turn
Would you trust AI as one input in clinical decisions?
Are we holding AI to a standard we don’t meet ourselves?
Let’s talk. 👇
—
đź’ˇ Inspired by clinical conversations and research.
(Views are my own, shared for education - not personal medical advice. AI-generated image.)
Further reading:
Low Back Pain Clinical Care Standard (2022).
https://lnkd.in/gCt4mdhN
Safe and Responsible Artificial Intelligence https://lnkd.in/g2VpUrAU