AI medical scribes: Understanding who's out there
Cofounder & COO
We’ve done our best to represent ourselves and our fellow AI scribes as accurately as possible, with our sources linked. Please reach out if you see something that needs to be updated! We’re happy to make the change.
Real-time clinical note generation
Nabla Copilot, Ambience and Abridge all return clinical notes in real-time at the end of an appointment, with Ambience coming in “within seconds” as they say, Abridge just over a minute, and Nabla Copilot falling around 20 seconds. Neat, huh?
Deepscribe’s note generation begins as soon as the consultation is ended, though their demo specifies that generation time is often several hours. You use the tool with other patients while the note is being generated.
Currently, Nuance DAX’s speech-to-text (Dragon Ambient eXperience) involves sending transcripts to AI-empowered medical scribes, who then return the finished note after varying lengths of time. They do promise real-time note generation in the not-yet-released DAX Express.
Human scribe in the loop
Somewhat of a philosophical difference has developed among intelligent scribe companies. For some, including a human in the loop is an essential layer of quality assurance. For others, the clinician themselves is an irreplaceable part of the process.
For both Nuance DAX and Deepscribe, AI-generated notes are passed through a step of human quality assurance, before being sent back to the clinician. Deepscribe’s model operates on the understanding that AI alone won’t complete tasks to the same standards as humans, thus each of their notes is reviewed by their team of human medical scribes. The whole process takes roughly a few hours.
In Nabla, Abridge and Ambience’s workflow, notes are generated and automatically available for review by the clinician, opting for a model that allows clinicians to add final touches — the insights that come only from having shared a space with the patient — before closing patient files immediately following each visit.
Note templates customized to your specialty & note-taking style
Note template customization is a big one: needs vary drastically between clinicians with their own note-taking styles and specialties that need significantly different medical information to be pulled out from consultations.
Ambience, Deepscribe, Abridge, and Nuance’s AI models use the history of their users’ sessions to understand their note-taking styles and learn to imitate them.
At Nabla, we never keep recordings of our clients’ consultations: we work with customers to adapt our models to their individual needs, taking into account everything from specialty (from Internal Medicine to Lactation Consulting) all the way down to the details (bullet points or text blocks).
API integration into your existing tech stack
Let’s talk tech! Players have various approaches here: ranging from little to no integration capabilities to full integration into the platform you already use.
Nabla Copilot, Ambience, and Abridge support direct API integration into the tech stack you already use, both through existing partnerships with EHRs and telehealth solutions and direct integration with proprietary software. This allows for note-taking to be hosted within the workflow already in place.
Nuance DAX & Deepscribe don’t mention access to APIs on their sites.
Supports asynchronous care (chats, voice notes)
With the rise of digital healthcare, asynchronous chats and voice notes are increasingly used in hybrid and digital care models.
Nabla Copilot is capable of taking these asynchronous interactions and structuring notes from them. The same goes for Ambience.
Abridge is able to use audio files or transcriptions to generate notes through integration with their APIs. We found no mention of compatibility with asynchronous care on the Nuance or Deepscribe sites.
Medical Code Extraction
Though notes alone are certainly helpful for saving documentation time, we don’t think we need to tell you that most of the admin comes afterward.
Using Nabla Copilot, Nuance DAX, Abridge, and Deepscribe, you will receive automatically suggested associated medical codes from consultation notes.
We weren’t able to find any information on automatic code extraction on Ambience’s website.
No Data Stored
It goes without saying that data privacy & security is central to any conversation around AI in healthcare. The stakes are high, so security has to be higher.
Nabla Copilot is the only solution that never stores user data: audio, transcript, and notes from consultations are only briefly stored in the browser of the clinician and patient information never touches Nabla’s servers. In addition to opting out of data retention with our LLM vendor, we have implemented an additional layer of security with a pseudonymization algorithm that systematically removes all portions of the transcript that contain personally identifiable information.
In most intelligent documentation solutions, including Deepscribe, Abridge, Nuance DAX, and Ambience, ML models are trained based off of the storage of user data. Each applies varying privacy & security practices to ensure said data is secure.
They have put in place a range of security and privacy measures, including employing multi-factor authentication, applying HIPAA security standards and committing never to sell data.
We have represented pricing models to the best of our knowledge, based on websites, conversations with various sources, and this article. For 100% accuracy, you’ll have to directly contact each team.
According to our sources:
- Nuance DAX currently falls at about $10k per seat per year
- Ambience ranges from $2800 to $3200 per seat per year
- Abridge comes in at $2500 per seat per year
- Deepscribe’s pricing evades us!
- Nabla Copilot is $119 per seat per month
For those strapped for time (we’re talking to you, clinicians): there’s only one solution you can get started with without going through a demo. You can try Copilot from wherever you are, on any device, for free, by just clicking on this link.
Nuance, Ambience, Deepscribe, and Abridge all make it quite easy to request a demo or sign up for a group session through their websites.
Additional resources for your exploration
Don’t just take our word for it:
- Andrew Lundquist has written an in-depth comparison after several months of testing
- Blake Madden, Paulius Mui, and Ben Lee mapped out the current landscape in AI Has Entered the Chat
- STAT went into the differing pricing models in their article, “Hospitals want to buy doctors’ happiness with Nuance’s AI scribe. What they’re paying varies”
- Bertalan Mesko wrote an article for the National Library of Medicine: “The ChatGPT (Generative Artificial Intelligence) Revolution Has Made Artificial Intelligence Approachable for Medical Professionals”
- Paulius Mui released a mapping of the Clinical Documentation Market
- Akilesh Bapu, Founder & CEO of DeepScribe, gave an interview to Unite.ai
- Nuance announced DAX Express, and Digital Health published a related article