Value in Medical Dictation – From Startups to Incumbents

When you picture a doctor at work, what comes to mind?? Perhaps it’s the image of a primary care doctor that's doing a wellness check.? Perhaps it’s an orthopedic surgeon that's fixing a broken bone.? Or perhaps it’s the busy emergency room doctor that’s stitching a cut.? In any of these examples, you’re probably imagining some interaction with a human.? Chances are, your image of a doctor isn’t someone staring at a computer screen for most of the day in managing their inbox and writing notes.? The reality is that for every minute doctors are spends with a patient, they’re spending 2 minutes reviewing electronic health records, writing notes, and doing desk work.? The average clinical note in the US is 4-times longer than those in other countries.? This type of excessive administrative work is one of the many reasons that doctors are getting burned out in record numbers.

So why does the US have this problem?? It comes down to incentives and billing.? The medical record is not only that – a record of your medical issues – but it’s also a legal document and a billing document all in one.?? Doctors must document their medical reasoning thoroughly in the event of a complaint or a lawsuit.? Likewise, how much doctors and/or hospitals get paid is based on the sickness of the patient, their medical conditions (i.e. comorbidities), and the complexity of the medical decision making.?

Take the hospital for instance.? They get paid a fixed amount based on the reason for the hospitalization.? Let’s say a patient gets admitted for a pneumonia.? A simple pneumonia may be capped at a payment of $30,000.? However, a more complex or specific the pneumonia, the more the hospital may get paid.? An E. coli pneumonia, might pay $40,000 instead of $30,000.? Also, if patients have comorbidities like heart failure, now the payment might be $55,000.? If the patient has heart failure and has very low oxygen from their pneumonia, then the payment might be $75,000.? Therefore, the goal of the hospital is to document all the comorbidities and complexities of the case so they can get paid more.?

Furthermore, the quality ratings of the hospital – such as how often patients die in the hospital, develop a new infection while they’re hospitalized, or are readmitted to the hospital within 30 days – are all risk adjusted, which is tied to documentation.? Medicare understands that sicker patients have more complications.? The risk of dying from pneumonia is much lower for a 30-year-old with no chronic medical conditions than an 87-year-old who needs life support and has long-standing heart failure, kidney disease, and emphysema.? To avoid financial penalties and to show that you have a great hospital, it’s absolutely critical to get the documentation and coding right.?

With all these requirements in mind, it’s easy to see why medical records are bloated in the US and why it takes so much time for a doctor to document their work.? As mentioned previously, physicians are getting burned out by this administrative bloat.? So much so that there’s a physician shortage and people leaving medicine in mass numbers.? With a dwindling supply of doctors, it then becomes critically important for technology to make the physician’s day more efficient.? Simply put, investing into technology like digital scribes isn’t just a “nice to have”.? It’ll be as important as it was for businesses to move from paper to computers and email in the 1990s to early 2000s.

So, what are the options?? One of the solutions is to provide dictation services or a digital scribe.? Currently, the most common practice is for people to sit down at a computer and dictate into a phone or a microphone.? People still need to include commands like “new line” or “period”. ?Additionally, it’s a dictation after the visit.? That means the patient and doctor already talked about what brings them into the office, what symptoms they have, any tests that need to be done, and what treatment to offer.? It’s better than typing, but it’s still very inefficient.? The doctor needs to have the conversation twice – once with the patient and again with the dictation software or service.

The newer technology is to ambiently listen to the conversation between the patient and doctor and to create a note immediately based on that conversation.? The doctor then just needs to review the note and verify it.? Think of it like the auto-summary feature on Zoom or Microsoft Teams meeting.? It cuts out the need for someone to take minutes in painstaking detail.? Moreover, these tools are getting good enough to cut out non-medical banter, like the patient’s recent ski trip or what their son is studying in college.? Additionally, there are some added features that are in development, such as providing summaries in lay-person’s language to foreign language interpretation.? Imagine having a tool that will create custom notes for both the doctor and the patient.? Or create notes in English for your EHR system even if you work at a Spanish-speaking clinic and all the visits are done in Spanish. ?These sound like killer features for sure and ones that will massively improve productivity.? There are some limitations, such as if there’s a lot of background noise and crosstalk.? Also, the features like a truly reliable foreign language service may be 1-2 years away, but these issues will improve over time.

So, what the business landscape?? There’s one massive incumbent, a few promising startups, and a bunch of newcomers in this space.? Let’s start at the top of the pyramid.? The 800-poung gorilla in the room is 纽昂司 , which was bought out by Microsoft in 2022 and has a major stake in OpenAI . So, as you can imagine, there’s a significant amount of resources and AI fine-tuning that are being put into Nuance.? On top of that, Nuance has large contracts with the main medical record systems such as Epic and Cerner which includes ambient dictation.? Nuance is not only the gorilla technologically, it’s the gorilla for revenue.? The medical transcription market has a value of approximately $1.5 billion with a CAGR of about 16%. Nuance’s annual revenue as a company has been about $1.4 billion.? Now, I should note that total includes other services Nuance offers such as a customer engagement business.? However, its main business is healthcare.? By the numbers, Nuance is commanding a strong majority of the market and is used by 77% of hospitals in the US.

There are a few Venture Capital -backed startups trying to disrupt this space, such as Abridge , Ambience Healthcare and DeepScribe .? These companies use ambient listening as an app on your smartphone or through a computer’s microphone to create a note in specific formats.? The doctor will review the draft, make any edits, and quickly complete the notes.? Furthermore, these software tools can track the edits so the transcriptions improve over time and be written in the individual style of each doctor.? These companies may either have APIs that interface with the electronic health record (EHR) to import the finished notes, or be integrated directly into the EHR.? That said, this space is starting to get crowded.? With the rise of LLMs (large language models) and NLP (natural language processing), the technology is more widely available.? This makes it easy for even more startups to join the race.? Consequently, the differentiating factor will be in features and workflow.? Unless you have an absolutely killer app that’s intuitive to use, you’ll lose.

Now, I would argue that this technology is nice to have, but it’s still not the holy grail.? With a clear market leader that is investing heavily into AI and many startups creating medical transcription solutions, the actual value of just the medical transcription will slowly be competed away.? The transcription technology will eventually merge around a core set of features.? This will ultimately turn ambient dictation into a marginally priced commodity that all healthcare organizations will have.? This will result in relatively low profit margins for the overwhelming majority of medical transcription companies.? We see this historically with Nuance’s profit margins.

So where is the value?? I would argue there are two enduring pillars of value.? One is the captured data itself.? Remember the note includes the summarized interpretation of the patient’s issues, key diagnostic results, medical decision making and treatment plan.? In the past, all this data was locked in the individual EHRs in the walls of Epic, Cerner, Stanford, MIT, etc.? Instead, the distilled information may be collected and summarized by these dictation tools.? If companies are able to take the transcription and turn it into deidentified meta-data, this will provide a robust dataset for further analytics.? This can include details into patient behaviors to how doctors order tests to the reasoning behind treatment plans.? The insights gathered here will be immensely valuable for startups, diagnostic companies, insurance companies, and pharma to name a few.? Now, if you’re thinking that patient information (including yours) is sacred and can’t possibly be used for business purposes, you’d be dead wrong.? Healthcare organizations and researchers may already be using your deidentified data, whether you realize it or not.

The second enduring pillar is in billing, coding, and insurance authorization.? As we discussed at the top of the article, how hospitals and doctors get paid to how quality metrics are determined is based on documentation and coding.? If instantaneous note production is the 10x solution, then the 100x solution is building features beyond transcription to include coding and billing.? This may include copilot prompts to ensure each note contains the pieces needed to maximize documentation and coding, and pre-loading certain diagnostic codes for review.? Likewise, it’s common for notes to be missing elements for insurance to authorize a diagnostic test or treatment.? In the future, the AI may spot-check your note to ensure it contains the elements necessary to justify the test or medication included in your treatment plan.

Looking further, the holy grail will be the ability to layer AI tools within the EHR in addition to dictation services to populate specific results and find overlooked diagnoses.? Imagine for instance if a patient has kidney disease but isn’t documented in the clinical note.? However, let’s say that adding that diagnosis increases billing collections by 20%.? Having a tool that can find these hidden or missing diagnoses and add them to the doctor's note is extremely valuable. When we consider that the cost to provide healthcare is rising and insurance payments are compressing, these types of solutions to maximize reimbursements may be crucial.? The main limitation to a comprehensive tool that can pull in data, document, and bill/code for the doctor will be regulation.? Right now, finding or clarifying diagnoses is mainly done manually by coding and documentation teams.? There are specific Medicare rules on what these coders can do to prompt doctors to maximizing their billing.? How much AI will be allowed to find missing diagnoses and automatically generate prompts will be a key decision to watch as transcription tools build out more features.

While I'm sure that companies like Nuance or a startup will unlock a positive cash flow business in the transcription service alone, that’s only the tip of the iceberg.? Transcription is a relatively low margin business with a total market cap in the handful of billions.? The real value is in the data pipeline and adding value to all components of the medical note.? The total addressable market (TAM) then turns from several billion to hundreds of billions in the US alone.? The AI scribe companies can then create an enduring moat with real lock-in to medical practices and hospitals.? This will certainly be a rapidly evolving space over the several years, and will undoubtedly fundamentally change the doctors’ workflow.?

For now, the market is still up for grabs. My bet? The company or companies that are most integrated into the EHRs will win. It'll be a race to get build out the features quickly and get as close to the physician workflow as possible. For now, Nuance is in the lead. We'll have to see if they continue to dominate the market or if another startup can upend them.?

I have no affiliation or investments in any of the companies mentioned in this article.

Abridge has been backed by investors including Bessemer Venture Partners , Spark Capital , Wittington Ventures , and others. Ambience Healthcare is backed by Andreessen Horowitz , Kleiner Perkins , AIX Ventures , and others. DeepScribe is backed by Index Ventures , Bee Partners , Stage 2 Capital , and others.

#healthcare #venturecapital #medicine #ai #startup