Sue Sheridan is hit with another medical error. This time, ChatGPT prevented the worst.
Patient safety legend Sue Sheridan had never touched AI. But then ...
Sue Sheridan is a living legend in patient safety circles, but she’s not an AI wizard: until this month neither she nor her husband had even touched ChatGPT.
But a few weeks ago they used it in an emergency, with dramatic effect. It ain’t over yet, but Sue is unquestionably in better shape than if she’d been passive. She is very, very glad they wasted no time turning to it when it seemed something was not right.
(Most readers will know Sue’s story of tragedies that drove her into patient safety, but if you don’t, here’s a four minute version and a full 87 minute documentary.)
Days of intense pain
In an online discussion group, I’d asked for #PatientsUseAI stories for this blog. Sue responded:
Thanks for encouraging us to share how we use AI as patients.
I JUST had an experience 3 weeks ago where I was having severe face, teeth, eye and jaw pain for about 2 days. I had been to a dentist a week before for a routine cleaning.
I returned to the dentist and she confirmed that my teeth were fine.
On day 3 the right side of my face was drooping. My husband, a doctor, took me straight to the ED. A brain tumor and a stroke were ruled out. I was told it could be due to stress, they ruled out Bell’s palsy and told me to use benign neglect regarding my symptoms (yes really).
This was obviously not a useful answer (“I continued in excruciating pain,” she writes), and they turned to ChatGPT.
Now, notice something: unlike the tech wizards we’ve described here before, such as Hugo Campos’ story and James Cummings’ story, neither Sue nor her husband had ever touched AI. But they fed the system Sue’s symptoms and circumstances - the same as they’d told the ED, including the suspect dental visit - and learned, in Sue’s words:
The routine cleaning of my teeth had activated the dormant shingles virus, which caused Bell’s Palsy. The virus was attacking my facial nerves.
We went to a second ED armed with knowledge from ChatGPT and the ED was immediately on it and prescribed high dose prednisone and an antiviral. Within five hours the pain was reducing substantially.
This was 48 hours after onset of the pain. She continues:
There is only a 72 hour window to treat the condition or else the facial paralysis can be permanent. I am still waiting to see if my face will return to normal.
As of this writing (three weeks later) her pain is mostly gone, and the drooping is reduced but obviously not gone.
I can’t help imagining what would have happened if they’d gone home and tried the “benign neglect” the first ED prescribed. It could have been a third life-altering medical error in Sue’s life.
Patients have many reasons to use generative AI - and many ways to do it.
In this blog’s first post I laid out five categories of work we’ve identified so far, for which patients are using AI:
Sue’s story illustrates the two use cases at the right: difficult diagnosis and exploring knowledge.
I can’t stress this enough. The whole purpose of participatory medicine - of doctor-patient partnerships - is to help healthcare achieve its potential, to avoid preventable shortfalls. That’s exactly what Sue and her husband achieved when they turned to ChatGPT in this emergency - even though they were complete beginners.
A core question at the root of doctoring is “What do we think we know, and why do we think we know it?” Many factors can affect a diagnosis, and generative AI can play a truly useful role by bringing together obscure information or sometimes even information that’s less obscure but overlooked.
Medicine should welcome patients contributing to their care with AI - and, importantly, we all need to develop methods on how to include it wisely. More on this in future posts.
If you have a story where you’ve used AI in managing your care, drop a note in the comments and we’ll be in touch.
p.s. Sue says she has contacted the management of the first hospital, told them what happened, and asked for a refund of her $350 copay.
Excellent patient education!! learn by example! May I suggest that your next post is a guidance test for newbies in using AI for solving health problems
These stories are so important...Very grateful to you and the #PatientsUseAI community for gathering and sharing.
Can Susan share what prompts she used to reach the information that was most meaningful and actionable? Perhaps that can be a part of future posts to help others get a head start on optimal prompting (what didn't work, what ultimately did)?