By John Jeffay, NoCamels -
Patients will soon be discussing their symptoms with ChatGPT before their consultation with a “real” GP.
They will, for the first time, be able to explain to a computer what’s troubling them, in their own words.
Kahun, a startup based in Tel Aviv, already uses AI to carry out a three-minute question and answer session in writing with patients.
It then presents their doctor with a summary of their condition, as a starting point, and offers possible diagnoses.
But the company is now incorporating ChatGPT, the AI chatbot launched last November, so that patients can communicate using natural language and explain themselves better.
Kahun still relies on its own vast repository of medical knowledge – over 30 million insights from trusted sources – but ChatGPT will now allow patients to describe their symptoms in their own words.
Until now it’s been a structured conversation, with the AI asking a question, the patient responding, and the AI working its way through a series of more detailed questions based on the answers it gets.
Integrating ChatGPT puts the patient in control. They describe their symptoms exactly as they would to a doctor, and ChatGPT responds, just as their doctor would.
None of this replaces the doctor. But it means that by the time the patient is in a face-to-face consultation – either physically or remotely – the GP will have a full summary, detailing their main complaint, relevant history, medication, and a list of symptoms that they’re not experiencing, as well as those they are.
The AI will suggest the most likely diagnoses, and offer a list of possible investigations.
“A text chatbot looks like a conversation,” says CEO Eitan Ron. “But essentially, it’s all structured. So behind the scenes, we come up with questions and the patient can say yes or no, or select from a couple of multiple choice options.
“With ChatGPT we identified an option to open it up to a real natural language experience. So now the patient can tell their story in their own words.
“With the new functionality, we let them use free text, in their own language, describe whatever they want, as a starting point, and we take it from there.”
ChatGPT will typically start by asking the patient a very open question – like “What’s troubling you most today?” As the conversation progresses it will start asking more formulaic or yes/no questions.
Kahun aims to offer doctors and patients the best of both worlds. Its own clinical-reasoning engine is built on a wealth of peer-reviewed medical literature, which means it can alert the doctor to extremely rare conditions as well as more likely diagnoses.
But it may be a little rough around the edges in terms of its bedside manner – more questionnaire than conversation.
ChatGPT is not the place to go for professional medical advice. It’s good, but it’s not that good, and it does have a tendency to hallucinate (ie make stuff up).
It is, however, a great communicator. It understands slang, non-medical terms people may use for body parts, and colloquial expressions.
So Kahun combines the best qualities of both – ChatGPT’s friendly approach and its own tried-and-tested medical prowess.
“LLM-based (large language model) solutions will transform many industries, and we see ChatGPT as a strong addition to our core offering to both patients and providers,” says Ron.
“In integrating ChatGPT, we’re really tapping into the best of both worlds by merging our clinical-reasoning machine with an LLM that can interact with patients and physicians with never-before-seen fluency.”
Kahun is currently piloting the ChatGPT-enhanced version of its AI. But the standard version – its XAI Clinical Assessment Tool – has already been used in 250,000 doctor/patient sessions in the US, primarily in telemedicine. Clients include Helix Virtual Medicine, which has clinics across Florida and California.
Other companies have developed symptom checkers, says Ron, as a more sophisticated alternative to Dr. Google. But they don’t “think” like a doctor.
“The main difference is that Kahun is not about telling you as a patient what you have.
“We make sure we ask you all the right questions, that we cover everything that you would expect from an experienced, trained clinician. Other companies are not doing that.
“Our platform can reason like a clinician and explain itself. It shows the doctor the clinical reasoning and points them to the literature.”
That’s the opposite of so-called “black box AI”, which provides good answers, but can’t or won’t show how it arrived at its conclusion.
Kahun, founded in 2018, takes its named from the Kahun Medical Papyrus, a catalog of symptoms and treatments from ancient Egypt that was written in hieroglyophics 4,000 years ago and deciphered 130 years ago.
