May 15, 2023: Robbie Hughes, Founder and CEO at Lumeon joins Bill for the news. How does Lumeon help healthcare providers cope with workforce challenges and the need to do more with less? How can technology help address the shortage of clinicians, and what are some of the potential implications of this technology for the healthcare industry? What are some of the challenges associated with using large language models like ChatGPT in the field? How can subjective narrative in care delivery be turned into a structure that can be used to eliminate interpretation handoffs and create trust and completeness in records? How can technology be used to make care delivery more evidence-based and science-based, while still incorporating human compassion and relationships? Will home-based care become a significant part of healthcare delivery in the next five years? How can healthcare providers balance reimbursement issues with the practical issues of delivering care in the home?
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 This Week Health.
You get people sort of dipping their toe in it, doing one or two patients, a week and you can kind of tie that together with paper and string and handpick the patients and say, they fiddle these criteria fine. But you then try to do that at scale and it's a completely different proposition. (Intro)
Welcome to Newsday A this week Health Newsroom Show. My name is Bill Russell. I'm a former C I O for a 16 hospital system and creator of this week health, A set of channels dedicated to keeping health IT staff current and engaged. For five years we've been making podcasts that amplify great thinking to propel healthcare forward.
Special thanks to our Newsday show partners and we have a lot of 'em this year, which I am really excited about. Cedar Sinai Accelerator. Clear sense crowd strike. Digital scientists, optimum Healthcare it, pure Storage Shore Test, Tao Site, Lumion and VMware. We appreciate them investing in our mission to develop the next generation of health leaders.
Now onto the show.
(Main) 📍 all right. It's News Day. And today we are joined by Robbie Hughes, founder and c e o of Lumeon. Robbie, welcome to the show.
Hey, bill, it's great to be here. Thank you
for having me. Looking forward to the conversation Lumon, you guys are international.
You have London office, you have a Boston office. You have a Denver office. Give us a little idea of what Lumion does.
So we think of ourselves as a care orchestration platform. What that effectively means is that we're bringing automation into clinical care to delivery. Now, this is a really interesting thing, like in Europe, we've been around a long time.
We've done a lot of stuff that's really delivered a huge amount of value in the US bringing automation into this space. It's been kind of interesting, but with the business model and the fever service and the cost plus pieces the desire to really substantially reduce cost hasn't really been the driver with the workforce challenges that we're seeing today and the need to do more with less.
The ability to amplify your care teams and do work that simply couldn't otherwise be done is proving to be extremely exciting for our clients. So it's a great time to be doing what we are today.
Oh, yeah. A fantastic and perfect time to be doing this, I had a conversation yesterday with a cio and we were talking about efficiency in driving automation and those kinds of things.
And one of the things that has become sort of the de facto belief is we will never have enough clinicians again. The, just the demographics just don't work. And so technology has to step in here somehow.
Yeah, that's right. I was in a conference last week and one of the delegates shared that India alone is gonna consume the entire population of nurses for the entire globe over the next decade.
There's not even enough people being trained. In fact, there aren't even enough trainers to train the people who need to be trained to satisfy the demand. It's what we have today is bad. It's going to get worse. It's just not sustainable in the current model.
Yeah. So we have to get creative.
As you look at all the things that are going on technology I put a poll out there on LinkedIn this week. There was five gosh, finding my own stuff on some of these platforms. Now, here it is Monday poll. Which HIMSS announcement will have the biggest impact?
Not the biggest hype, but the biggest impact on healthcare. And there was four announcements. There was Oracle Zoom. So Oracle and Zoom have come together, telehealth, WellStar and clear around patient id, which I thought was interesting. Phillips AWS and generative ai. So one of the few announcements that wasn't open AI and Microsoft and whatever, but Phillips and aws were janitor of ai and then obviously the Azure, Microsoft Azure Open AI available in Epic.
Those were the four choices I put out there cuz with the LinkedIn poll you can only put four choices. There was a lot of announcements, but those are the four I put out there. Which of those, when you hear about 'em, which of those partner announcements do you think is gonna have the biggest impact?
I think they're all gonna have impact on, in their own way, like the Oracle Zoom thing. Whether it's Zoom or Teams or anyone else, there needs to be a decent video conferencing capability in these platforms that needs to exist. I think the Oracle kind of service oriented architecture momentum is interesting, actually.
I think it's super cool. But that I think that's fine. I don't see that as massively exciting.
It's not ground, it's not earth shattering kind of news. Yeah. But it's necessary.
Yeah, that's right. I think the Phillips stuff I've got a lot of respect for what they're doing in terms of that laser focus around the diagnostics piece and the ability to effectively find a better way to be the term copilots being used a lot.
And I think that's the right term. How do you become a better copilot to your customers? And I think that's smart and I think that's gonna be effective. And again, it's gotta be done on a private cloud. It's gotta work in a way that. Is gonna be trained on sensible, curated data. So I think that's fantastic.
I worry a little bit about the chat, G p T, the large language model kind of piece more generally. I think it's extremely exciting technology. I think it's got a huge amount of potential. I do think that the ability for us to look at what it's doing and interpret it as fact when in fact it could, Unverifiable be something that is, in fact, I think is potentially a massive distraction and a problem.
And so, when you look at putting chat g p t or technology like that into our field, and you say, okay, I'm gonna add the ability for it to summarize my notes, for example. Okay. At face value, that's fantastic. I mean, personally, as an engineer, I strongly dislike the fact that we're taking.
Structured data, turning into subjective narrative data, and then using technology to turn the subjective narrative data into more concise narrative data. I, that boggles my mind. But the problem is we can't decipher the fact from the fiction. And so I think it's a really interesting announcement. I think it's very exciting, but I think it has to be very carefully managed to have the desired impact.
Yeah. And so I was talking to somebody yesterday about how we're going to make these models more accurate for healthcare because clearly we live in a, can't be wrong, deterministic narrow set of data to learn from and whatnot. I mean, you can't learn from the internet.
How to practice medicine. I mean, it has to be very focused on or a lot of weight put on things like periodicals studies, journal of American Medicine. I mean, you wanna really focus it in and say this is the trusted data so that the model can be more accurate. But how do we do this?
I mean, there, I mean you've played with chat G P T I would assume at this point.
Yeah. Yeah. It's, I think
it's really impressive.
Yeah. And it's hugely impressive for where it has taken factual information and then used that to create kind of narrative off the back of it so you know, what's it doing?
And it's basic level. It's looking at a corpus of text, it's then running it through its model. The model roughly being proportionate to the size of the corpus of text. It's absorbing. And then all it's doing is it's generating tokens. It's saying statistically, next to this token, the right next token is this.
It just goes on and on, like that. That's why it generates
words. It just, it's string words together, essentially.
That's all. That's all it's doing. And so if you are training it on stuff that's a hundred percent accurate. Then there's a good chance that well actually, lemme step back. We don't actually know why it works.
We should be clear about that. It works. I know it works because it works. And a lot of the reason it works is because it's been very heavily sort of massaged and curated and refined and there's a lot of work that's gone into that. So then the question is, can we trust the corpus of data that we're looking at today?
And so if you look at it and you made the point, you know it. How do we make it more accurate? I'd say today it's very precise, but we can't validate its accuracy. And if you are training on a bunch of journals and on a bunch of kind of clinical literature, I mean, we know from experience that what happens in clinical trials does not necessarily reflect what happens in clinical practice.
I think as an industry, there's a huge gap between the ability for people to, well, we've made a lot of investments in trying to make good clinical decisions, but actually we haven't really focused on executing them well, and we see that as gaps in in execution, things getting missed.
So if we're training it on a corpus of data or a corpus of text, that reflects what happens in the clinical trial space, is that necessarily reflective of what will happen in operational practice? Don't know. If you train it on what's happened in operational practice, do we have sufficient confidence in an absence of variability in that corpus?
In other words, can we absolutely guarantee that the decisions being made are accurate and that the execution is accurate, such that we would then base a whole bunch of other stuff on it? Again, don't know. I don't think we would. And then how would we curate it? So, It's clearly not a general purpose solution to how do we practice medicine by chatbot, that's not what it is.
Can you narrow it down to very specific applications where you can add within this kind of narrow lane some particular elements of value? I think that's right and that's why I think the Phillips thing is smart, cuz that's effectively what they're doing. They're finding a very specific use case and they're finding something where it can.
Add value quickly in a field where the overall service can be augmented in a, in an effective way.
Yeah. These narrow models we've seen 'em in healthcare, right? So we've seen 'em in radiology, cardiology. We've seen them with, I mean with, quite frankly with a bunch.
We've seen it in sepsis. There's really narrow models that utilize AI that are very effective. Because they are there's a reinforced learning that's happening. There's human feedback that's happening, and there's this virtuous loop that's created where the model gets smarter, it gets validated by humans.
It gets smarter. It gets smarter. I, can we create that? I mean, so if I made you CEO of OpenAI today and you're looking at this thing and you have this partnership with Microsoft, and Microsoft is in every healthcare organization in the us. I don't know of one that it's not in. And so there's a, there's an opportunity for every CIO to check a box and say, I want Azure Open ai, and I want to start, I wanna start using this.
And the Epic announcement essentially says now Epic's not careless. In fact, they're the farthest thing from careless. They're very measured in how they make progress, and so they're not gonna apply this to coming up with diagnosis and that kind of stuff. That's not what they're gonna do.
I mean, there's not a chance they will apply it to writing letter, just communication, writing letters, just the general purpose stuff. But is there a way, if I made you CEO today, to get to narrow it down for healthcare and get that reinforced, learning the human feedback into the model?
So where I would apply it, and I think, as a company where we're looking at it today is I've got all this data. How can I make it more accessible? How can I have a conversational in or an engaged conversational interface with something that's going to make sense of what I've got? So I think there's a massive value in saying, show me all the patients that have got this or that or the other, and bring that to me in a sensible way.
I think that's great. Again it's. This has been put into some of the announcements. I think the picked up on something that the eClinical work CEO said where he said it does magic. That helps our customers truly benefit from public cloud computing. I don't like the word magic in this.
I think the,
it tend to know what it's doing. It can't be like, yes.
Yeah, exactly. And when you said, letter writing, things like that, or another one is around patient engagement. Can you direct a bot to achieve something? Like if your goal is to get this information from the patient, go off and figure that fit out, that's fantastic.
And I think there's an entire class of systems that. Are gonna be effectively, I think of the way we're thinking about this internally is this is sort of an exponential leap in capability. And so you need to look at everything you're doing and say, okay, where is that exponential disconnect going to affect and where can we accelerate?
And I think patient engagement is one area where there will be just an exponential change in capability and this ability to get value from things. I think analytics is the other one. I hate the fact that we have to produce letters. I think, I think someone's mission somewhere should be to get rid of the letter.
I think, again, it's back to
this it's one of the practical applications and you don't like it. But one of the practical applications of this is we still have a million faxes going all over the place and ChatGPT four could actually look at the facts, read the facts, and turn it into discreet data element and all sorts of stuff.
It could do. And then move it into the to the record so we could potentially eliminate facts with something like this.
I, yeah I agree. I just, the principle of having subjective narrative and then using a thing to create more subjective narrative, that it's somehow better. Like there has to be, I think the leap here or the unlock is, How can you take the subjective narrative and then turn it into structure that you can then do more things with?
And now someone smarter me will probably say, well, actually with these technologies you don't need to do that, cuz you can just have, you can just keep it as free flowing corpus and it doesn't matter. I don't know if that's the case or not, but the idea that so much of what we do in. care delivery Is intermediated by narrative that has to be created and then read and interpreted, opens up so much variation in execution and practice and delivery that the goal here should really be to try to eliminate those interpretation handoffs and try to smooth the practice and the delivery of what has to happen on the ground.
So we know, like you used the, again, bill used the word deterministic earlier. Medicine broadly speaking, should be should be science in its basis and sort of artful in its delivery. You should use the human aspect of care delivery to engage people, get them to do the right things, build up those bonds and those relationships.
But the decisioning, the execution, all of that should be evidence-based and it should be based on science. And so, can we use this technology rather than trying to rather than trying to take so much of this stuff that is kind of artful in nature and applying a layer on it to make it faster.
And what I mean by that is not the human compassion and not all of that. That's fine. We should be doing that. So much of what we put into an encounter, so much of what's documented is effectively free form text that shouldn't really need to be there. I mean, I'd be willing to bet, and I've done a bit of analysis on this in the past, a long time ago in the uk, if you were to look at all the letters that go around from all of the senior physicians in the UK and you would to train a model on that I put money on the fact that the most statistically likely word you might say might be appearing after the word.
This is the word delightful, because if you go and look at a letter or any kind of inter physician communication, you'll find, I saw this delightful lady about this, or I saw this delightful gentleman, and there's so much of that stuff that takes up time. And we should just be able to look at the record, trust the record to be complete and say, okay, based on that, this is deterministically what we need to do, and this is how we're gonna get it done.
But I feel like there could be a lot of uses for these technologies and sort of perpetuating that old way of working, which again, as an engineer, I don't like, I would prefer to say is the record is objectively complete. We trusted and then we act on it. And I, from our company at Lumion, a lot of the things we struggle with are, we go in and we look at a clinical delivery process and we say, okay, well why are you doing this?
Why are you doing this? Why are you doing this? It's a lot of, it's down to trust. They don't trust what's in the record. They don't see what's in the record to be complete, and so they revalidate things and they check things. They end up with all this kind of repetitive steps, and so can we use these things to, to somehow create trust and completeness in these records?
And my fear is that actually it goes the other way. We summarize it, but we don't trust the summary because the summary is based on something that's incomplete. And so we get further away from that trust and that completeness rather than closer to it. So I don't wanna see that. And yeah, I don't, it's tricky.
Yeah, no 📍 it, it is tricky. 📍
Alex's lemonade Stand was started by my daughter Alex, in her front yard. It By the time she was four, she knew there was more that could be done. And she told us she was gonna have a lemonade stand and she wanted to give the money to her doctor so they could help kids like her.
It was cute. Right? She's gonna cure cancer with a lemonade stand like only a four year old would.
But from day one, it just exceeded anything we could have imagined because people responded so generously to her.
We are working to give back and are excited to partner with Alex's Lemonade stand this year. Having a child with cancer is one of the most painful and difficult situations a family can face at Alex's Lemonade Stand Foundation, they understand the personal side of the diagnosis, the resources needed, and the impact that funded research can have for better treatments and more cures.
You can get more information about them at alex's lemonade.org.
We are asking you to join us. You can hit our website. There's a banner at the top and it says, Alex's lemonade stand there. You can click on that. And give money directly to the lemonade stand itself
now, back to the show.
We could stay on this topic for. Another hour if we wanted to. We only have a couple minutes ago. The Kaiser Geisinger thing just happened. We're not gonna talk about that. It's too new.
We need to consume a little bit more information. Before we get there, but there's a interesting story that came out. So Three health systems are joining Amazon's CVS Best Buy and supporting home-based Care expansion. Let me give you a little excerpt from this, A coalition featuring Amazon Best Buy, CVS Health, and three Health Systems.
Is supporting federal legislation aimed at bringing more healthcare into the home? The moving Health Home Group says, at-home care increases access and reduces cost without sacrificing quality and safety. The pandemic has taught us that in polls confirm that care in the home is preferred by many patients with increasing demand for at home options.
The coalition founder talks about the representatives a little bit, and then it says, the co coalition includes Ascension, it includes Hackensack Meridian, and includes Intermountain Health. As well as Best Buy and CVS and others that I mentioned. If you were to project out five years, what are your thoughts?
I mean, obviously we're gonna see more home-based care over the next five years. How much, I mean, do you think we will see a significant move into the home?
I'd like to think so. I think there are some fairly materials. Issues. I mean, putting the reimbursement aside for one minute there are some quite practical challenges in delivering this stuff at scale.
So, on the one hand, putting acute care into the home, fine, do it at home. Then how far is the home from where the staff are? How do we think about the logistics of getting things there? It's not just remote patient monitoring. It's all of these other things around it. And the problem I have with this is
as a capability, and I think there are very few people who can solve this is if you are, for example, Ascension and you solve this at one hospital, the logistics challenges, the staffing challenges, all those things will be solved at that hospital and around that hospital. You then solve it for another hospital and you're starting from scratch.
You do another one. You're starting from scratch. The logistics and the challenges of actually providing this care in the home. Is essentially a ground up problem every time you solve
that. So, so essentially it's not scalable is what you're describing.
It doesn't scale in the way that we think.
Most of these sort of digital first or consumer-centric models, scale, consumer-centric models that are digitally enabled scale extremely nicely and work very well, where you're having to put a logistics backend onto it where you're having to actually put humans into these things, deliver things, make sure they happen.
That's a very hard challenge to scale, and there's a few people who are attacking it.
Right. So Amazon's going after this. They've proven the ability to take a digital workflow and scale it with a logistics backend. I mean, will it take somebody with that kind of experience and background to scale this across many major cities?
I think it's gonna take someone with a huge amount of understanding of the digital problem, kind of what I would call the orchestration problem of all the digital and human assets and logistics and everything together, and then has got the firepower to actually muster the suppliers to do that.
And I'll give you an example of this. So let's say for example, in one region, and let's agree for a second that. You've found a geographic region in which you can deliver this within some sort of, I dunno, 50 mile radius and let's agreed for a second that maybe the patients who are in the 51st mile radius can or can't get that service.
You've got some way of making that a rational and sensible kind of boundary that you can cater to. So it's not gonna be one company that does this, it's gonna be sort of a coalition, great. What we're talking about here. But it's also gonna be after coalition of downstream providers who are gonna have to sign up to SLAs and service levels and bas, and all of these protections where they have to accept a standard operating model and a standard set of requirements that they may not like, for example.
And what you can get in one region may not be what you can get in another. And remember today, When these hospitals are opening up virtual wards or or hospital home programs, they're not big, like 50 patient, 50 beds, for example. Many smaller than that. So you've gotta find vendors and suppliers who are prepared to do this at pretty small scale on pretty tight, clinically oriented delivery time scales, meeting service performance agreements that.
Maybe they don't wanna do because the economics aren't there to do it. So I think there's just a whole bunch of downstream logistical issues that if you're focusing it on one region, fine. If you're trying to do it nationally, much, much harder problem. So I'm excited to see how this plays out.
Yeah, I, clearly you have to determine there's levels of acuity.
What levels of acuity can you actually deliver? And if it's a higher level of acuity, it has to be closer. To the to the primary hospitals and that kind of stuff. I think and the policies, the procedures around it, going into people's homes. The technology and how you set that technology up to work.
Is it gonna have to be cellular versus cellular? We don't use cell towers anymore, but essentially cellular versus wifi based, because we, I mean, there's a whole host of things. I think what we're gonna see happen though, is there's gonna be a demand for this. And companies are gonna come alongside and create that those sets of workflows integrations to the EHR and processes and procedures.
And they're going to approach the local healthcare systems. And they're gonna say, Hey, look, we have this offering. we know there's a demand. We know that you don't have the wherewithal to stand it up in every one of these hospitals, but you could partner with us. And we can knit that together.
I think it is a different capability. I think we think, oh, the health systems do this on the campus, therefore they will do it in the home. And I think they can deliver, they can be a partner in delivering the care in the home and almost must be, but they don't have to stand it up.
Yeah. It's not enough.
And. What we've seen of these is you get people sort of dipping their toe in it, doing one or two patients, a week or whatever else and you can kind of tie that together with paper and string and handpick the patients and say, they fiddle these criteria fine. But you then try to do that at scale and it's a completely different proposition.
And anyone who's been involved in any business that has any kind of supply chain dependencies will understand that's a very different problem set. And I think there's just a lot that we have to learn as an industry and how to do this well because again I love the guys at Best Buy. I think they're extremely well positioned to do this.
when that deal first happened, people might have been scratching their head saying, well, why does this make sense? It makes sense because of all these other bits that you need to do. And this, the industry's gonna learn all this stuff very quickly over the next couple of years. Yep.
Robbie, I wanna thank you for your time. This was a phenomenal discussion and I look forward to having a few more in the future.
Thank you, bill. Great to be here. Thanks for your time.
📍 And that is the news. If I were a CIO today, I think what I would do is I'd have every team member listening to a show just like this one, and trying to have conversations with them after the show about what they've learned.
and what we can apply to our health system. If you wanna support this week Health, one of the ways you can do that is you can recommend our channels to a peer or to one of your staff members. We have two channels this week, health Newsroom, and this week Health Conference. You can check them out anywhere you listen to podcasts, which is a lot of places apple, Google, , overcast, Spotify, you name it, you could find it there. You could also find us on. And of course you could go to our website this week, health.com, and we want to thank our new state partners again, a lot of 'em, and we appreciate their participation in this show.
Cedar Sinai Accelerator Clear Sense CrowdStrike, digital Scientists, optimum Pure Storage. Sure. Test Tao, site Lumion and VMware who have 📍 invested in our mission to develop the next generation of health leaders. Thanks for listening. That's all for now.