When it comes to the potential impact of AI on healthcare, I consider myself cautiously optimistic. More optimistic in fact than I feel about other aspects of how AI will affect our lives. Politics, news, social media and information security are just a few that raise major concerns.
However, I see many ways AI can augment and improve healthcare and the patient experience. Hence my optimism. On the flip side, there are things that could go horribly wrong without adequate human oversight. Hence my caution.
I’m not going attempt to lay out all the pros and cons for every AI use case in this blog post. But there’s one that makes perfect sense to me: notetaking in the exam room. I mean who likes taking notes? It’s not what any doctor went to medical school to do. Yet, it’s critical to document patient interactions.
Ironically, I’m writing this blog post just a few days after Open AI’s notetaking tool made the news for some not so great reasons. More on that later. But first, let’s start with what the technology in the exam room is all about.
AI listening: Coming to an exam room near you
This case study from the American Medical Association offers a good overview of what's happening with AI listening tools.
Using the microphone on a secure smartphone, the ambient AI scribe transcribes—but doesn’t record—patient encounters and then uses machine learning and natural-language processing to summarize the conversation’s clinical content and produce a note documenting the visit.
Why doctors like it:
A primary care physician, Dr. Lee noted that she has had decades long relationships with some patients and the ambient AI scribe filtered out all the talk about children and pets and the exchange of Christmas greetings while producing a note that she would have had to type into the record herself.
Adoption:
The Permanente Medical Group rolled out the scribe technology tool, offering it to 10,000 physicians at 21 locations in Northern California. It was adopted by 3,442 physicians who used it in 303,266 patient encounters during the 10-week period before Christmas.
At the start of the study period, the ambient AI scribe was used almost 20,000 times a week and then exceeded 30,000 times in seven of the 10 weeks studied.
This was the quickest spread of technology and quickest adoption of new technology in the medical group ever.
My experience with AI notetaking
A few weeks ago, I went in for my annual physical. Yes, I am of that age! Before we got started, my primary care doctor asked my permission to use his phone to listen in and take notes.
He admitted to being a somewhat slow adopter of the technology, but he had found that it was saving him time and allowing him to focus more deeply on conversations with his patients rather than looking at the computer for the entire visit.
To me, the experience was not invasive or off-putting at all. In fact, I thought we had a more complete conversation than in years past.
Following the appointment, he encouraged me to check the after visit summary to verify the accuracy of the AI-produced notes. I was all over that.
It was amazing to see how well the technology filtered out the unimportant parts of our conversation. Like when he asked me what 80s TV show theme song my phone ring tone used. He guessed it was the A-team. Nope. Magnum PI of course.
Sorry, I had to do that!
Back to the point, how was that summary? Pretty great. I caught one error – it added 10 years to my age! That seems like an especially odd error given that you would think the technology would have allowed for some cross-reference with the data in my chart to get that right. If you’ve consumed healthcare lately, you know how often you provide your birthday as a data verification point. Maybe that ability will be released in a subsequent version upgrade.
Like all applications of generative AI right now, ambient listening does need healthy oversight.
Let’s go back to the AMA’s case study with Permanente Medical Group.
While most notes were accurate, a tiny percentage did contain “hallucinations,” which is the term used when AI provides false information without a sound basis.
For example, one physician mentioned scheduling a patient’s prostate exam, but the AI scribe recorded that the exam had been performed.
Another time, the doctor discussed issues with the patient’s hands, feet and mouth, which the AI-generated summary recalled as “the patient being diagnosed with hand, foot and mouth disease."
OK, that’s a big difference!
What the future holds
With a workforce that’s stressed out and demand for care that far exceeds physician supply, tools like AI ambient listening clearly have an important role in the future of care.
If consistently done right – and that will take more time – it truly seems like it can be a win-win for everyone.
Working at Providence, I’m really excited by how our organization is approaching AI as an opportunity to improve care. Providence is focused on four specific domains:
Assisting and augmenting clinical caregivers
Consumer and patient self-service, engagement, and navigation
Supporting our workforce in administrative areas
Streamlining back office functions
Providence is evaluating how AI can assist, augment or be autonomous in these areas. For example, AI can assist in collecting patient information before appointments or augment clinical decision-making during diagnosis. AI can also work autonomously to complete administrative tasks.
Exciting! You can read more about that here.
One final point: Earning public trust
Trusting AI in the exam room is going to take time. Last year, a Pew survey found that most people are skeptical.
“More broadly, the public is divided on the impact of AI in health and medicine. While 38% say it will lead to better outcomes for patients, 33% say it will lead to worse outcomes and 27% say it won’t make much difference. One widely expressed concern is that AI will make patients’ personal relationships with their providers worse.”
So, returning to where we started: my cautious optimism. If AI tools can empower doctors to be more efficient with their time AND help support an empathetic, personal relationships with patients during office visits it will be a win-win for everyone.
Let’s hope setbacks are quickly overcome, but with safety and accuracy in mind.
Comments