Researching a difficult diagnosis: #PatientsUseAI, example 1

On Monday I called for us to start using a new hashtag #PatientsUseAI, because we in the patient community keep seeing doctors and regulators and businesses talk about industry using AI, and nobody’s taking into account that we out here are using it too – and we may have different priorities. The graphic at right spotlighted five ways patients are already using AI. This post is the first in a series about them.

---

September 11, 2023 will live forever in my memory as bringing the most crystal-clear contrast I’ve ever seen between the paternal old guard and an arriving new reality: patients with real problems using artificial intelligence (AI) to help improve outcomes. That morning, at the annual meeting of the global standards organization HL7, a famous doctor gave a speech about AI, and near the end he said:

“It is just too early for us to give generative AI to patients
and expect that the results will be good.”

(Generative AI is tools like ChatGPT.)

Yet that same day, Today.com (the website of the Today show) ran this story about a kid named Alex who had problems that 17 doctors couldn’t figure out, saying, “His frustrated mom asked ChatGPT for help.”

She pasted in everything she had from the boy’s many MRIs, and it suggested a diagnosis of tethered cord syndrome. A new doctor confirmed it, the boy had surgery, and he’s doing better. (Go mom!)

So which is it? Is it too early, or is it already happening, with real results?

And here’s the kicker on paternalism vs reality: as one audience member said at that speech, “What is he talking about, give it to patients? We already have it!”

We must all wake up and realize #PatientsUseAI. And difficult diagnosis is one thing we’re using it for.

This new reality isn’t just our opinion.

This case was cited by the extraordinary physician Eric Topol in his TED Talk about AI that came out in December, Can AI catch what doctors miss? He also cited another case of misdiagnosis:

… a patient that was sent to me, who was suffering with, she was told, long COVID. She saw many different physicians, neurologists. After getting nowhere … her sister put all her symptoms into ChatGPT. It found out it actually was not long COVID, she had limbic encephalitis, which is treatable. She was treated, and now she’s doing extremely well.

Too early to “give it” to patients? I think not.

#PatientsUseAI. Don’t be surprised.

These cases shouldn’t be shocking – they’re a continuation of what empowered patients have been doing since the dawn of the Web, and even before: they do everything they can to help a loved one when doctors are out of answers. Remember Lorenzo’s Oil, (Wikipedia), the pre-Web story of the Odorne family, whose son Lorenzo had a devastating diagnosis and was expected to die by age 10. With their tireless work he lived to be 30.

Back then (the 1980s) a key factor was that the Odorne family had riches and connections: father Augusto worked for the World Bank and they had access to great libraries. Today, as we’ve just seen, ordinary families have access to much of the same knowledge through generative AI.

By the way, Topol’s AI TED Talk above is pretty mind-blowing – you should watch it (17 min). In addition to those diagnosis cases, he shows many other ways different kinds of AI are helping doctors, and he says this about diagnosis:

These are not just anecdotes anymore. 70 very difficult cases that are the clinical pathologic conferences at the New England Journal of Medicine were compared to GPT-4, and the chatbot did as well or better than the expert master clinicians in making the diagnosis.

Note again: “as well or better” than “expert master clinicians,” using the gold-standard test cases from the world’s most prestigious journal, the New England Journal of Medicine (NEJM).

And that’s just generic GPT-4, not the forthcoming new AI’s that will be optimized for medicine. We in the patient world are eager to not just see those future products but participate in their design, features, and regulation.

(Speaking of NEJM, its new publication NEJM AI has a column about patient use. How’s that for acceptance by the establishment?)

We’re using it. Our needs are different. We need to be involved in regulating it.

As I said Monday, these are exciting times: an explosion of knowledge and power in medicine, and patients in need are making the most of it. I also said it’s essential that we the users get to participate in making the rules, or we’ll be counting on others to know what we want. And that never works: “When someone else speaks for you, you lose,” I said, quoting a disability rights leader long ago.

“So damn exciting.”

Topol ends the talk by saying, “It is really a remarkable time for the future of health care. It’s so damn exciting.” Never forget that his TED Talk included patients as users. Yes, with good results. Because yes: #PatientsUseAI.

Next in the series: Exploring: “I use it to help me think”