Ambience Healthcare

So today we’re chatting about Ambience and how it’s used at your company. Before we jump into that, could you give a brief overview of the company and your role there?

I’m an internal medicine physician. I mainly work as a hospitalist, and I also do a lot of tele-hospitalist work and telemedicine for nursing homes. I use Ambience in all these situations, but mostly use it for telemedicine because that’s where it helps me a lot.

What drove you to look for a product in this space?

Especially for telemedicine, I wanted a way for technology to take dictation for me so I wouldn’t have to type it out myself. I was looking before these products were in the market, but last year I started being comfortable with the options I was seeing. I had used some non-AI-driven scribing solutions such as M*modal and Dragon, and still use them at times.

What requirements did you have for a scribe product?

I wanted a product that would cut out the side conversations I had with a patient, which wasn’t possible with something like M*modal. Having to edit those conversations out would take even more time. I wanted to be able to see the patient, have the conversation, and get it transcribed automatically so I don’t have to go in and edit it afterwards.

Did you look at any competing AI scribes? How did they compare to Ambience?

I looked at DeepScribe, Nabla, and Scribeberry. I think they were all at about the same level as far as the conversational transcript is concerned. I’ve been using Ambience more so I know its faults better. DeepScribe gives you the option to select a patient’s pronouns, so it won’t default to “they/them,” and Ambience doesn’t do that. I use a workaround where at the beginning of a conversation I’ll say “the patient prefers he/him pronouns.” Ambience says they’re going to implement a pronoun selection feature, but they’ve been saying that for six months now.

Ambience does have a lot of advantages, though: the assessment and plan that it outputs is superior to DeepScribe’s. But they’re all advancing a lot right now.

I stuck with Ambience because I saw that when I gave them feedback, they actually implemented it, and that gave me the confidence to go with them.

How was the onboarding and implementation experience?

The team was very responsive and very professional in their implementation, and they’re willing to work with various EMRs.

Could you describe your workflow and how Ambience fits into the process?

On a typical day when I’m doing telemedicine, I’ll log into my computer, open my Chrome browser, and activate the Ambience extension through there. It asks for a patient name and then I can start recording. Then I start the telemedicine session through Zoom and see the patient. When the session is done, I click “finished” and the recording ends, and Ambience asks me for a diagnosis. When I enter that, it generates an assessment and plan based on that information. It will generate an assessment and plan without a diagnosis, but it won’t be as focused. I haven’t integrated it with my EMR, but all I have to do is copy the note into the EMR and do any necessary editing.

For in-person visits, I mainly use it for ER admissions. I review the patient’s chart ahead of time, start the Ambience app on my phone, and sometimes dictate a pre-session note with the patient’s history. Then I go see the patient, and have my phone in my pocket with Ambience recording. When I’m done with the patient, it’s the same process as for telemedicine: I finish recording, enter a diagnosis, and copy the notes into the EMR.

What’s the turnaround time for note generation?

It’s typically instantaneous. If I see a patient in the ER, I’ll type in the diagnosis while walking back to my desk, and by the time I’m at my desk it’s finished. At most it might take 10 seconds.

How much do you need to edit the notes afterwards?

For a single note I might have to change three words on average. It saves a lot of time. Sometimes the AI doesn’t recognize a technical term, or it mixes up the names of medications, and sometimes it will interpret a side conversation with a patient as a relevant conversation, but it’s very easy to catch those.

What do you see as the key strengths for Ambience?

I last used the competing products eight months ago, so I don’t have a good sense for how they compare to each other now. But I felt that Ambience was quicker than the others. I also found that it was very good at listening for the names of places and people.

The downtime of the tool is also very low. My colleagues and I have noticed that. Even when there is downtime, the audio is saved locally, so when Ambience comes back up, the note will get transcribed later. That’s an advantage over competitors like Scribeberry, which are browser-based.

What weaknesses do you notice?

The inability to select pronouns is an issue. I don’t want it to use they/them pronouns when the patient has already indicated their preferred pronouns elsewhere in our system. The other main issue is that it hallucinates at times, like all the AI products. Sometimes it will enter the name of a different drug, or an entirely different word, even though the word was very clear on the recording. It happens maybe once a month.

How well does the mobile app work?

It works pretty well now. Initially it was pretty buggy, but it was in beta at the time, and for the past six months it’s been very stable. It works in the background, so I can pull up Epic Haiku to answer a question from a patient’s record and it will keep recording while I do that.

Do you have plans to integrate Ambience with your EMR?

I’ve worked with so many systems, and I know the IT team will ask a lot of questions before we can integrate, and they’ll charge me or Ambience to do so. And Ambience won’t integrate without a fee too. So for me it’s not worth it: it won’t save me more time and it works fine as-is. In the future they’re coming out with some new features that might make integrating more useful.

How has the support experience been?

It’s gone really well. I think they’re a very well-run company. The team is very supportive and responsive and works with an open mind.

Looking back, do you think you made the right assessment choosing Ambience?

Yes, I did.

What might drive you to explore another solution?

There are a few features which Ambience is working on right now, that if another solution comes out with them, I might consider switching because I want to see how it will work. Something that helps me with coding would be very helpful.

Do you have any advice for other people exploring the market of AI scribes?

I think the best way to make the choice is just to start using a trial version and see how it works for their practice. For me, I work in so many different scenarios that I had to choose something that would fit in all those places. They might have to try a few options before deciding which one suits them the best.