This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

  Today on Town Hall

(INTRO)   Just imagine how well a human scribe training primary care would actually work in an oncologist office and how that oncology scribe would perform in the emergency room. The clinician job across specialties is so incredibly different, and therefore the co pilot job must inherently be different

  📍 📍 📍

My name is Bill Russell. I'm a former CIO for a 16 hospital system and creator of This Week Health.

Where we are dedicated to transforming healthcare, one connection at a time. Our town hall show is designed to bring insights from practitioners and leaders. on the front lines of healthcare.

Today's episode is sponsored by Meditech and Doctor First .

Alright, let's jump right into today's episode.

 Welcome to the This Week Health Community Town Hall Conversation. I'm Reid Stephan, CIO here at St. Luke's Health System in Boise, Idaho. And I'm joined today by my friend, Mike Ng, who's the co founder and CEO of Ambience Healthcare. Mike, good to see you. Thanks for making the time. Reid, thanks for having me.

It's a pleasure to be here. So take a minute, Mike, and just share a bit about your career background in a particular, a little bit about the Ambience story, how the company started, what your mission is, what you do, just for those listeners who may not be familiar with Ambience. That's perfect.

My name is Mike Ng, co founder and CEO of Ambience Healthcare.

I fractured my back and had a lot of time to really think about what I really want to do with the time that I have, the skills I've been given. And I've had the pleasure of working with my co-founder, Nikhil Paduma, the author of the textbook Fundamentals of Deep Learning. For eight years across two companies, and I'm not sure if you heard Reid, but one of his other options at the time was actually to be an early team member of a small AI research lab.

People now know it as OpenAI, and I'm sure he and his parents have zero regrets about joining.

Yeah. And how long has Ambience been in

existence? Four years, four and a half years at this point. Um, although this has been eight years in the journey of healthcare machine learnings, people say right place, right time.

We believe one place, long time.

Uh, I love that. Okay. There's a catchphrase for this episode. So curious, like I have my thoughts, but what is it like in the last year or two? Ambient scribing has become such a delighter for clinicians. I've never been involved in a technology that's had more of a positive reception and more of an organic interest in, hey, when can we get this?

What's your take on what's fueled that and what's led us to this moment?

That's a really good question. So maybe we start back at our last company and how we came to some of the insights to start Ambient. We ran a care provider entity. And one of the things we quickly learned that only about 20 percent of our clinician's day was spending time directly in patient care.

And actually a lot of the 73 percent was on charting, coding, and other admin. So we naturally tried all of the different Ambient solutions out there and quickly learned a couple of things. First of all, what works at a demo doesn't always work in reality. And that oftentimes clunky technologies worse than no technology.

At the core, clinicians went into this profession to take care of patients. And the challenges with modern day practice is after I spent so much time entering data in the EHR. And there are two things. They're forced to either multitask during a visit or stay up late to finish charting at night. Few people actually have enough funds to afford hiring a human scribe to help them do this work.

This has led mostly this burden of charting and documentation, along with CDI, being some of the leading causes of burnout across some of our already overworked clinicians. In what we are seeing over across the nation is that many have already left the profession. Others are considering at a time when we have a really large natural shortage of clinicians.

So, when you finally get a chance to build a tool that allow clinicians to actually focus their undivided attention to patients, they have more time to listen, more time to think. And more time to give care. And that ultimately leads to happy clinicians, happy patients, a happy health system. And to me, in healthcare, with very few times you get this triple win, and we'll take it whenever we can get it.

And what's interesting is I think a lot of the healthcare leaders who had the vision of bringing in ambient documentation now feel like superheroes.

Yeah, I think I've certainly felt that way as well. And you described that triple win. It really is that improvement of experience for everyone that's involved.

Uh, and that's certainly been sort of the, kind of the rocket fuel on our ship in our journey. I'm curious though, as I talk to fellow CIOs that are, you know, pilot phase, or maybe they're in production with their ambient listening solution, whatever their vendor is, generally it's the same kind of a theme what Unite has talked about.

It's a delighter. It improves the experience. Because of that, do you feel like in terms of the ambient listening, the AI scribing solution, are we reaching a point where it's becoming a commodity? Like it's just An expected capability that we're going to need to provide for our clinicians and our patients.

And if so, what does that mean for maybe the broader capabilities of an ambient listening solution?

That's a

really good question.

So I'll break it down to two components. The first is, I think it's a very good thing that ambient scribing solutions are now becoming more widely accepted as the future.

Costs are going down, accessibility going up. However, if you As an industry, across vendors, we're really seeing good traction in primary care. But the data is showing that solutions generally struggle to expand into specialties. Inpatient and ED. And so this is one thing we learned early on in building Ambience, that the job to be done in every specialty and every sub specialty is actually vastly different.

Just imagine how well a human scribe training primary care would actually work in an oncologist office and how that oncology scribe would perform in the emergency room. The clinician job across specialties is so incredibly different, and therefore the co pilot job must inherently be different. So why it took us so long to build the core platform of Ambience is because our philosophy here is that to build.

The co pilots for each specialty from the ground up. It takes exorbitant amount of time to chase perfection by specialty. So on this problem alone, we've spent. Gosh, now our team is probably 20 plus on product integrations, designers, and then also a dedicated ML team and a clinical team that at this point is five full time doctors working with specialized clinical advisory boards by subspecialties in order to tackle this problem.

And that has been a big part of our secret to actually successfully expanding across specialties. And one of the hard earned lessons that we had to take over time is that you're not building one co pilot, you're building more like 50. The others. Obvious challenge here is, AI is evolving so quickly, and it's really hard for health systems to really understand the performance.

So, the insight here is not all AI solutions are created equal. And so, nowadays, health systems are speaking with their colleagues to compare notes. And they're quickly realizing that there is a vast difference in performance across various metrics, from adoption to time saved ROI. We're actually incredibly far away from commoditization.

And selecting the right partner really matters. And more often now, we're seeing groups test multiple different AI scraping solutions. or deepening a diligence by conducting what we call these side by side comparisons. There are many times where we go into a health system and they have multiple vendors sending up their machines and then they have these pre made recording mock visits by different specialties and you have to actually produce the output together.

Then you actually have A panel of clinicians by specially reviewing the notes for quality and accuracy, just to see the difference between what the different vendors are producing out there. I think what folks are starting to realize is we're actually really far away from perfection and actually selecting right behind it matters.

Especially if your roadmap goes beyond AI scribing for primary care.

You said you're really honing in on what fascinates me. So beyond the AI ascribing of primary care, and you talked about just the complexity of different specialties and care service lines, what are some of the capabilities beyond the AI ascribing that a CIO should think about as they approach this from the idea of a platform versus a point solution to solve the in room note taking problem?

So we

took

a step

back, and one of the things we did was make sure that we truly understand the jobs to be done, the problems to be solved, and you, as many people are starting to realize, EHR documentation is used for so many different reasons. There's the charting component, but there's also coding, prior authorization, utilization, management, and quality and reporting.

And an ambient documentation solution that doesn't really fully appreciate all of these use cases actually end up saving the collection of time in the short run, but creating all of these downstream issues. So an earlier version of Ambience, we, we built was not what we call coding aware, which means it didn't understand CDI.

And for example, in a clinician's visit, they may explain colloquially to a patient the word wound, but from the AMP setting, if you're looking at a professional coder, they needed to use the word ulcer. So in this case, from clinician's time. But the HCC code worth thousands of dollars in a capitated reimbursement environment was ultimately rejected by the payer.

So Coding Aware Solution is something that anticipates all of the coding rules for documentation and writes the documentation in a compliant way. At the point of care. And people often mistake AScribing as a summarization problem, but in reality, every single line has to be written for a reason. That's why we have to train our models to not only understand medicine, but also coding, prior authorization, quality measures, UM, and more.

And so this leads to another insight, which we had, which is to get from a scribing solution to a true co pilot, we actually had to do a lot more. And the goal of getting to that from the 27 to 73 percent actually required us to also look at how much time a clinician is spending before they see the patient.

For example, in oncology, you might actually have these 20 to 30 clicks to actually understand past patient history and context, and if a clinician doesn't have time to actually pre chart properly. They actually have to spend more time asking questions in the visit, which obviously is not a great use of time for the clinician, it's a terrible experience for the patient.

And then after each visit, Think about the post visit workflow. Clinicians now have to make several coding decisions to finalize the CPT codes for each interaction. So if we think about if clinicians didn't go to med school to chart, they most certainly didn't go to med school to learn all of the literally tens and thousands of coding and billing rules.

But if the clinician never wrote down the pieces of information down, Either because they didn't have time or they didn't know the rules, then the code or review would never have known it actually happened. And the reality is the minimum viable product, as we see it, is much larger than AI scribing alone.

And to really achieve that vision of 73%, this requires a much more holistic copilot from ambient documentation to ICD 10 CPT support to after visit summaries to referral letters and soon to be read pre charting as well. And all of this is done by one platform. One system, one set of integrations, not a collection of point solutions.

Yeah, I love that. And you've used the word, described it as the co pilot. The augmentation beyond just the AI scribing and resonates, it makes complete sense to me. I think that really is the exciting thread of this capability. Like the immediacy of what the joy that we hear from clinicians is really around the AI scribing.

But then like I step back and appreciate that, oh, but wait, there's so much more. So you talked about pre charting, maybe as we wrap up this conversation, give us a shadowing of, of what you're thinking about in that space or maybe doing today, what the future might hold with that kind of capability and in particular, again, how that's going to drive experience and outcomes in a positive way.

That's a really good question.

I think, first of all, the, there's a couple of trends that's really working in our favor. And the first is, EHR integration is getting faster and faster. So the standardization of APIs mean that, say, for example, at St. Luke's, we went from kickoff to everything ready to start in five weeks.

That's including ambience embedded directly into Epic as an activity. So it's not a separate app. And also one click launch from Haiku. And this level of integration not only is fantastic for the core ambient visit for summarization of certain parts of the note during the visit, but also helps with downstream actions, such as order entries, for example.

And it's actually really important to have those set of integrations before you can approach problems such as pre charting, given that you actually need connectivity to be able to pull out certain pieces of information, have the clinical intelligence to understand what to summarize for that clinician, for that visit type, in that specialty.

And so, what we're super excited. about is this idea of having one co pilot across the different use cases before the visit, during the visit, after the visit. And then also having the ability to expand across different users within the health system. So we think about is how do we support the workflow for outpatient, inpatient, and every patient visit.

And we see that there's incredible amount of synergies. So you think about Nursing workflows, there's so many pieces of information that the patient explaining to the, to the nurse, that's actually helpful for fracture care, critical care, which is important for the entire ED workflow. And then as we get, as the patient is moving from the ED workflow.

and intake into inpatient. There's so many pieces of information. If you didn't get right at the beginning, you're actually losing the context as you flow through the system. And that ultimately impacts in a very different way. So that, for example, you actually have a lower CMI score because you're actually not writing down all the care that's actually provided for that patient.

Lastly, the part we're really excited is once we start banking early wins with a health system, we are now Started to be seen as a care transformation partner. For overall strategic initiatives. So not a tech technology solution, but how do we help use technology to drive operational change to achieve an overall objective?

One example, when we first started with a group in psychiatry, the greatest limitation factor there was the intro evals, where because it's an hour long visit, documentation takes about 50 minutes afterwards. The Scarce resources of psychiatrists can only see about four patients a day for intake. We were able to compress enough documentation time without concurrent documentation to get them to be able to increase patient access.

And it turns out that this group was trying to expand into adolescent care. And the challenge here is now you have a version of the note for the school, for the parent, for the child, and another one for the billers. So think about the documentation paying for that. And so partnered up to co generate all of the documentation with the right Tone, context, and compliance for each of their stakeholders.

And instead of that group expecting for that care transformation to take three quarters, we finished in about four and a half weeks.

Wow. It's so exciting. Mike, every time I visit with you, I learn something and I leave inspired. I really feel like we've entered and are at the start of this kind of golden age of being a healthcare CIO.

And I think companies like Ambience and the capabilities you're helping bring forward are a key part of ushering that in. So just thank you for your time. Thank you for your mission and what you're doing, and I wish you the best of luck going forward.

That's incredible. Thank

you so much,

Reid. (MAIN)  ​  📍

  📍 📍 📍 Thanks for listening to this week's Town Hall. A big thanks to our hosts and content creators. We really couldn't do it without them. We hope that you're going to share this podcast with a peer or a friend. It's a great chance to discuss and even establish a mentoring relationship along the way.

One way you can support the show is to subscribe and leave us a rating. That would be really appreciated.

Special Thanks to today's episode sponsors Meditech and Doctor First

Check them out at thisweekhealth. com slash partners. Thanks for listening. That's all for now..