Where healthcare AI ambient scribes are headed

2024 has become the year of artificial intelligence ambient scribes. AI scribe companies have made headlines, entered countless healthcare organizations, and raised hundreds of millions of dollars. Simply put, Scribe is today's leading generative medical AI use case.

Still, this nascent product class is rapidly evolving. Let me explain how it leaks out.

How AI scribe eases the burden of document creation

Clinicians like me spend an average of 2.3 hours documenting every 8 hours of scheduled clinical time. This work takes up our attention, saps our energy, and contributes to burnout.

AI scribes can help automate this process. First, automatic speech recognition is used to transcribe the clinician-patient conversation. These transcripts are then passed through a large-scale language model to generate structured clinical notes for clinicians to review, edit, and finalize.

Many anecdotes and several peer-reviewed studies suggest that AI scribes reduce documentation time and reduce burnout.

Is AI Scribe a commodity or a differentiated product?

With over 50 products now available (see diagram), AI scribes appear to be becoming a commodity. Still, the products differ in more ways than just price.

Accuracy: Many AI scribes are essentially wrappers for products like ChatGPT, so they are relatively inaccurate and susceptible to fabrication. Conversely, an AI scribe trained on medical conversations will produce more accurate notes. Some products allow clinicians to cross-reference the text of notes with the transcribed conversation, alleviating concerns about hallucinations.

Specialty and setting: Most companies design products for primary care clinics, but two-thirds of clinicians work elsewhere. Some companies tailor their scribes to other specialties (such as orthopedics and cardiology) and settings (such as emergency medicine and inpatient care).

Personalization: Many AI scribes produce standardized notes that read as if they were written by a computer. Alternatively, some allow clinicians to specify the length (summary vs. detailed) and format (e.g., paragraphs vs. bullet points) of notes to suit their preferred style.

Ease of use: Clinicians using standalone AI scribes must go to the vendor's app or website to review, edit, and copy the note text and paste it into the EHR. Users using integrated products can record conversations using the EHR's native mobile app and review and edit AI-generated text directly in the EHR.

Notes are a wedge into the clinician's workflow

As Nabla CEO Alex LeBrun explains, “Clinical documentation is not an end in itself; it is a lingua franca that bridges various downstream processes.” Scribe adds a variety of features beyond note-taking, including:

Clinical Summary: A little-known secret is that specialist and inpatient physicians often write most of their notes before seeing a patient. AI scribes are poised to help automate this “pre-charting” activity by automatically summarizing relevant details of a patient’s medical history.

Orders and Referrals: A human scribe (usually a medical trainee) who sits in the exam room with the patient and physician not only writes notes but also places orders (e.g., tests) for the clinician to review and finalize. , medication, referral). Many AI scribes will soon do the same.

Coding: AI scribes will soon recommend billing and diagnosis codes, making sure notes contain the details needed to justify a claim. Ambience reported that its coding tools increase revenue by approximately $5 per visit. This could be enough to give you a positive return on investment.

Patient summaries: AI scribes can generate post-visit summaries that include patient instructions in English and other languages.

Discrete data capture: Clinicians and nurses continuously enter data into discrete EHR fields within flow sheets, problem lists, clinical registries, etc. AI scribes can automate this arduous task and transform unstructured conversational data into structured text.

Clinical decision support: Reducing the burden of “administrative” tasks is a noble cause, but supporting better clinical decision making is of far greater value. Abridge took the first small step by linking her notes to UpToDate's evidence-based recommendations. It is easy to expect that the scribe will prompt the clinician to fill gaps in care or recommend a differential diagnosis.

Importantly, the title “scribe” is often outdated. Some companies are rebranding their products as ambient and conversational “intelligence,” “assistants,” and “co-pilots.” Importantly, as we'll see, moving beyond note-taking requires deeper EHR integration that only some AI scribes can achieve.

Small Practice AI Scribe Market Remains Fragmented

In small practices, price-sensitive clinicians purchase and use the software. As a result, AI vendors typically adopt a product-led growth strategy, essentially offering clinicians a lightweight version of their product for free and hoping that they will later convert into long-term paying customers. Masu.

But integrating the entire long tail of small practice EHRs may not be worth the effort. As a result, multiple AI scribes with relatively limited capabilities will compete to service small operations. These scribes may differentiate themselves by tailoring them to specific practice types (such as physical therapy or behavioral health). Some vendors may give away these products for free as part of a larger bundle.

Large-scale medical system AI scribe market will consolidate

An AI scribe land grab is underway as countless U.S. health systems test AI scribes directly. These organizations prioritize solutions that integrate with corporate EHRs, satisfy various stakeholders, and meet high compliance and security requirements.

Relatively few AI scribes can clear this high hurdle. “As ambient documentation becomes more at stake, users become more demanding and expect solutions to meet their personal preferences and needs,” said Punit Soni, CEO of Suki. Some players will be eliminated, consolidation will be facilitated among others, and a small number of players will remain.”

Will EHR vendors decide who wins AI Scribe?

Basic EHR integration (usually via the FHIR API) allows clinicians to complete note creation within the EHR. However, other activities such as pending orders and flowsheet updates require deeper EHR integration.

Therefore, EHR vendors may shape the outcome by determining which vendors gain access to privileged integration points. Epic (DAX Copilot and Abridge), Meditech (Suki, Augmedix, DAX Copilot), and athenahealth (Suki) already have partnerships with various AI scribes. Some are even co-developing new features.

Brendan Keeler (aka Health API Guy) believes these partnerships give selected AI vendors a head start. But “information blocking rules (and in Epic's case, rules that incite antitrust violations) mean that functionality created for one AI scribe must then be made available to others.” do.”

On the contrary, Canvas CTO Andrew Hines believes that these partnerships have long-term benefits, stating that “information blocking rules ensure equal access to information (i.e. data reading), but No guarantees are made for features, workflow modification capabilities, or engineering expertise.”

In any case, while many AI scribes aim to become broader intelligence platforms, EHRs remain the platform and distribution channel where these products should reside.

AI scribes are quickly becoming a standard feature in clinical settings. However, as Elion's Patrick Wingo explained, “It remains to be seen whether new vendors entering the market will be able to provide enough product features, ease of use, or cost savings to make significant progress as standalone products. ” Major EHR vendors such as Epic and Oracle are building a variety of AI capabilities.

Penguin AI CEO Fawad Butt added, “The scribe business is highly vulnerable to competition from all sides, including hyperscalers (such as AWS) and EHR vendors.”

So AI scribes, or what will eventually be called AI scribes, are here to stay. Over time, they fade into the background and, in some cases, become part of the EHR. Without a doubt, the market will look much different than it does today.

Acknowledgments: I would like to thank the following people for discussing AI scribes with me and helping me write this article: Ruben Amarasingham, Fawad Butt, Adam Carew, Andrew Hines, Brendan Keeler, Alex LeBrun, Yair Saperstein, Punit Soni, Patrick Wingo.