Lately, a lot of VC money has been spent on ambient notetaking (1,2).
As one of the hottest AI related investments in medicine, I want to take some time to fully unpack my thoughts on this topic. So first, what is the purpose of notetaking/documentation in medicine?
Almost every doctor has to make a record of what they have done for a patient –e.g., office visits, surgery, MRI reading– and this documentation is critical for several reasons. First, it allows you to communicate with other doctors and get them up to speed on a patients past medical history, labs, ect before the doctor actually sees the patient. Second, it a way to remind yourself of the patients main concerns when they return sometime in the future. Third, it is proof to insurance companies that you saw the patient and justification for how much you should get compensated for the encounter –the more you do for the patient, the higher the compensation.
In short, documentation is very important for doctors. Okay, so the natural next question is, how long does it take? It varies based on the doctor’s efficiency, but previous studies showed that it can take up to a third of the doctor’s work day (3) –an immense proportion of the day. To cut that time down not only saves time for both patients and doctors, but also saves money.
Ambient notetaking is an AI-driven solution to this problem. The idea is that for every patient encounter, the doctor can bring in a smartphone and keep it on the side while the encounter takes place. The app on the phone is connected to a large EMR system like EPIC or Cerner. At the end of the encounter, the app converts the recording into a full history and physical exam write up in the way a doctor would typically write. Then, the doctor reviews and makes edits to the draft. In theory, this sounds like a great solution.
A lot of hospitals are beta testing this product with directors of programs or other important attendings. I asked the director of the rotation I am on to test out one of the leading ambient notetaking apps called Suki. I pretended to be a patient and did a mock encounter with the doctor and we saw the results. The notes were not good. The biggest problem was that the summary would paraphrase something from the encounter that was not precise, and technically incorrect. The wording was also clearly written by AI. Other technical inconveniences were present as well; you have to login to the app, then you have to login to the EMR system within the app. You also still have to go back into a chart on a computer to deal with ordering labs so it isn’t seamlessly all done on your phone.
This is one of those things that sounds so great on paper until you try it out in practice. The hype is unfounded. So many people who are investing in it don’t know what documentation is like in practice. Even if you have the best LLM integrated into the software, the way doctor’s take notes in practice is idiosyncratic and not easily replicated by LLMs at this point. There are also just so many little technical problems related to inefficiency as I mentioned.
That isn’t to say this technology won’t get better in the future once there is more data fed into the AI models creating the notes. But, we are far from it being the great solution that it’s often cut out to be.
There is also another glaring problem that most people don’t consider. Ambient notetaking is only one solution to notetaking, and I’d argue that there is already a better alternative on the market. Dragon’s dictation tools are far superior for notetaking at the moment. I’ve been using this software for the past month and it has made my notetaking x2-x3 more efficient. The idea is that you just speak the note into a mic instead of typing it out. It seems so simple, but the innovation is truly mind-boggling when considering efficiency. I think companies like Microsoft know this though since they acquired Dragon’s dictation software for 19.7 billion in 2022 (4). It is a bit odd that the same Microsoft back company, Nuance Communications, is creating an ambient notetaking app too, but I guess it is just to keep up with what others are doing. I don’t really think they need to.
Now, I just want to finish by linking a video where I talk with friend about ambient notetaking.
Also, if you like this content, please subscribe to the newsletter to get emails whenever an article is published!
References:
Heather Landi, “Lightspeed, CVS back Abridge’s $150M round for generative AI,” Fierce Healthcare, February 23, 2024. Available at: Fierce Healthcare (Fierce Healthcare).
Anna Parker, “Ambient Listening Technology for Faster Documentation,” EMRFinder Blog, January 16, 2023. Available at: EMRFinder Blog (EMRFinder | Find EMR Software | USA).
“The hours 23 physician specialties spend on paperwork, administration,” Becker’s Hospital Review. Available at: Becker’s Hospital Review (Becker’s Hospital Review).
Todd Bishop, “Microsoft’s $19.7B Nuance deal is finally final, nearly a year after it was announced,” GeekWire, March 4, 2022. Available at: GeekWire (GeekWire).
