HLTH Special: Dr. Stephanie Lahr Spills the Tea on Ambient AI and Computer Vision

[00:00:04] Welcome to High Tea with Grace podcast where we spill the tea on HIT. I'm thrilled to be here with Dr. Stephanie Lahr, president of Artisite at the Health Conference. Stephanie, thanks for joining us today.

[00:00:15] I'm thrilled to be here with Stephanie Lahr, president of Artisite at the Health Conference. Stephanie, great to see you.

[00:00:21] Hi, Grace. It's so great to see you too.

[00:00:23] So tell me, let's like just dive right in here. Stephanie is one of my favorite people in the industry. She is so sharp, former CHCIO, just powerhouse woman.

[00:00:33] And she's here at Artisite. So I want to hear a little bit about the career path that brought you to Artisite, because I think everyone will love it as much as I think it's just so fantastic.

[00:00:42] You mean random?

[00:00:43] Yes! And amazing.

[00:00:45] So I'm an internal medicine physician by background and but really early in my career on a story for another day that we can talk about.

[00:00:53] I ended up in the health IT space as a hospitalist and doing work at an organization and that kind of evolved.

[00:01:00] And then I was the CIO and CMIO at Monument Health and was a wonderful role, dual position where I really got to help spearhead all of the opportunities that we were thinking about from technology and strategy.

[00:01:14] And I've always been one of those people where I know we're getting to places where technology can bring some really cool stuff to us.

[00:01:22] I felt like for many years early in my career as a CMIO, my mantra was like, I can make it suck less, right?

[00:01:30] Not really billboard worthy, but it was better than not at all.

[00:01:33] And in the last several years and in my few years as the CIO, I was starting to see the evolution of technology getting us to a place where, hey, we might make this not suck at all.

[00:01:46] And so I started to be introduced to a number of companies like health.

[00:01:53] It was able to bring together all these startups and young companies with great ideas who really want to help us solve the health care problems that we have.

[00:02:02] And I met the CEO entrepreneur of Artisite and another fellow physician.

[00:02:08] And he was passionate similarly about bringing the joy back to health care and doing that for all of our clinical teams and doing that through automation.

[00:02:18] And so we got involved with him as an organization.

[00:02:23] And I was just so excited about the work he was doing and the opportunity to be a part of something that I do really think can be transformational and bring the joy back to the bedside.

[00:02:34] That about two years ago now, which time flies crazy how that's happened.

[00:02:39] I decided to switch for the third time.

[00:02:42] So from physician to physician leader and CIO in the technology space to now leader at a health IT company.

[00:02:50] Yeah.

[00:02:50] Wow.

[00:02:51] Just so impressive.

[00:02:52] I'm so thrilled you're here.

[00:02:54] I've been hearing little whispers about ambient listening throughout the industry.

[00:02:59] Just lots of innovations happening in the space.

[00:03:01] So I just want you to give me a little bit of understanding that what is ambient listening?

[00:03:05] And can you also talk about beyond reducing administrative burden, what other areas of health care do you see that ambient listening really having an impact?

[00:03:14] The great way to think about ambient listening, especially if you tie it back to even the things we have at home.

[00:03:20] People have Alexa devices and things like that.

[00:03:22] That's a form of ambient listening, right?

[00:03:24] So what does that mean?

[00:03:25] It means you've got a device.

[00:03:26] It's not really listening to everything you're saying.

[00:03:29] It's listening for triggers that it's been trained to listen for, typically with a set of words that would say, oh, I want your help with this.

[00:03:40] I need you to do this.

[00:03:41] And that listening then can be translated into other kinds of outputs.

[00:03:47] So you're right.

[00:03:48] A lot of the buzz in the industry right now has been how do we leverage ambient listening?

[00:03:53] So devices that are in the room and around in our care spaces that will allow a clinician, whether that's a physician or a nurse or a therapist, to have an interaction with a patient, listen to that, and then take it to the next level, which is ambient listening supported documentation.

[00:04:12] Right?

[00:04:13] So a lot of what we've been hearing about is this desire to drive down the documentation burdens for our clinicians because we know that's some of what's driving burnout and those kinds of challenges.

[00:04:26] But I think if we dig into ambient listening, even beyond that, the reality is there's a whole bunch of things that could be articulated in an operating room, in a patient room, in a clinic that could drive process and drive efficiency.

[00:04:44] So maybe that's hearing that a patient needs something and automating a message to a downstream system to make that happen.

[00:04:53] Or maybe that's hearing that something has happened in the room and automating downstream notification.

[00:05:02] One of the things that we've started doing with our technology, which is unfortunate to have to think about needing, but is using ambient listening to help support safety in the workplace.

[00:05:13] Right?

[00:05:14] And so for nurses or clinicians that might be in a patient room, if something starts to escalate and they feel unsafe through ambient listening technology, we can hear a keyword or something that identifies that is being triggered.

[00:05:30] And then we can send communication to security or something else to let them know this person feels unsafe.

[00:05:36] Let's get into the room.

[00:05:37] So again, really, I think ambient listening is about documentation and potentially a whole lot more as we think about the kinds of things that are said in a variety of patient care spaces that we could then reduce the friction about how to take those words and make something happen from them.

[00:05:57] When you're deploying this ambient listening technology, what are some of the challenges or limitations that you've encountered just in the past in terms of like just getting it out there and used in that?

[00:06:08] It's about more than ambient listening.

[00:06:11] It's about really a smart hospital platform, which is a variety of IoT sensors that replicate human senses.

[00:06:20] Right?

[00:06:21] So we want to be able to see like a clinician.

[00:06:24] So we leverage cameras with computer vision.

[00:06:26] We want to be able to hear and speak like a clinician, which is that ambient listening side of using speakers and microphones.

[00:06:33] And then we might layer in other forms of technology like ultra wide band and RTLS technologies to understand where things are in space.

[00:06:41] If you start to think about the idea, and I definitely believe that in the next five years, every health system in the country will have a camera and a speaker and a microphone in patient rooms.

[00:06:52] I think that when people first hear that, they're like, oh, wow, really cameras, speakers and microphones in patient rooms.

[00:06:59] Is that invasive?

[00:07:00] Is that jeopardizing privacy?

[00:07:03] How do we want to think about that?

[00:07:05] When it comes to any of these ambient technologies, whether it's ambient listening or ambient computer vision, it's about socializing with people what the technology does do and what it doesn't do.

[00:07:17] We're not recording your voice.

[00:07:19] We're not recording your voice.

[00:07:20] We're not recording your video.

[00:07:21] And we're not listening or watching for every single thing that you're doing.

[00:07:25] We're looking and listening for special cues that tell the system to go and do something.

[00:07:32] And I think that is definitely one of the things that we really have to help walk through with people is I love the idea of the automation and making things easier.

[00:07:41] But I'm a little nervous about this idea about all this observation being in my space and the big brother side of it.

[00:07:48] And then for us as a clinical company started by clinicians for clinicians, one of the other big things that is super important to us and how that we communicate is we only want to use these technologies to make people's lives easier and better and more effective.

[00:08:03] We're not going to listen to something you said and use it against you later.

[00:08:07] That's just not how the technology even works.

[00:08:09] Again, we're listening for key things and then driving fast forward to automation and hopefully efficiency.

[00:08:15] Yeah, it seems like still a lot of education that has to happen in the industry in terms of that understanding of we're not saving your videos for later and selling them or something.

[00:08:24] Absolutely.

[00:08:25] In fact, there's no video to be watched later by anyone to begin with, right?

[00:08:28] It's a window into the care area and it allows us to raise the level of care and observation with a whole new layer of support and technology that we haven't had before.

[00:08:41] Like using computer vision to monitor for falls or using computer vision to monitor for pressure ulcer prevention and skin injury prevention.

[00:08:50] So, yeah, absolutely a lot of education and socialization.

[00:08:55] And I think as people are bringing more of these technologies into their personal life, that's a great easy way for us to be able to have these conversations around.

[00:09:05] If you're in a self-driving car that has any of those functionalities, there's cameras in there.

[00:09:10] They're watching, but they're not watching you.

[00:09:12] They don't care what you're doing.

[00:09:13] They just want to make sure your eyes are on the road.

[00:09:15] So, thinking about those similar kinds of things and creating those crosswalks into regular life applications of some of this technology and helping people see that we already do some of this stuff.

[00:09:27] Really interesting stuff.

[00:09:28] And it's funny because you don't really hear about this a lot.

[00:09:31] I think the industry is very much just starting to really understand it.

[00:09:35] And so, you're really at the forefront of trying to help them understand what's going on.

[00:09:39] So, you mentioned quickly computer vision.

[00:09:42] So, we talked about ambient listening.

[00:09:43] Tell me more about computer vision.

[00:09:46] Areas in the industry really see it having an impact.

[00:09:49] And then also, I want to hear about how it maintains accuracy and reliability when you're integrating it into that high-stake healthcare environment.

[00:09:58] Great questions.

[00:09:59] So, computer vision really is this ability to see like a human, leveraging compute in the background, right?

[00:10:08] So, as a camera is watching and seeing something that they are able, that it's able to understand what it's seeing and then do something with it.

[00:10:17] So, computer vision is in essence our way with technology to try and replicate vision.

[00:10:23] And just ambient listening and natural language processing is our way of trying to use computers to understand what we say.

[00:10:33] Computer vision is much more complicated.

[00:10:36] And honestly, if you look back to the origins of early artificial intelligence, computer vision was something that people were very excited about, even as far back as like the 1960s.

[00:10:49] But we actually didn't have all the technology that was necessary to make it happen.

[00:10:54] A lot of what's needed is compute and then the visualizations.

[00:11:00] Our ability to have the amount of compute available in a cost-effective way has only very recently become available to us.

[00:11:09] And then the other piece is we need video from, or not video, we need images from a real life situation.

[00:11:18] And so, to your point about how we do this in healthcare, the reality is computer vision needs real environments.

[00:11:25] We've trained, again, we'll go back to the analogy of self-driving cars got to the place that they are today by having images analyzed when a human was driving the car for a super long amount of time and getting inputs from the driver and all these different things.

[00:11:42] And then we brought that together and taught the computer vision what it looks like to have a pedestrian in the road, all those kinds of things.

[00:11:50] Nobody took a car into a TV studio and said, let's drive it around in here for several days and we'll teach it how to drive.

[00:11:58] In fact, we didn't even say, let's take a car that was trained how to drive in San Francisco and pop it over to London and let it drive there.

[00:12:06] Like that at a gut level would make people say, oh, I feel like that's not going to go well.

[00:12:11] And it wouldn't because that computer vision is very dependent on the inputs and the elements that are specific to the location and the setting that it's in.

[00:12:23] So, in healthcare, one of the things for us that has been so important is the computer vision algorithms that we work on are all built side by side with our health system partners who are leveraging the cameras and the speakers and those things that are in the room already

[00:12:39] for things to support things like virtual nursing and virtual sitters and patient observation.

[00:12:45] So, the technology is providing value at baseline, but then we can leverage the interaction between the bedside nurse, the virtual nurse, and the patient to help computer vision understand what is happening in that room and what inputs are important.

[00:13:01] And so, our computer vision algorithms, all of our algorithms are retrained in every environment.

[00:13:06] And I think that's a lot of what we're hearing.

[00:13:08] I've been hearing about today, even at health, is what are we all doing to make sure that AI lives up to its promises?

[00:13:16] And the reality is a good portion of that is making sure that the data that is being used to drive those AI algorithms is effective.

[00:13:27] And computer vision is only going to be effective if we're using data from real health systems and we're working with that in each health system to tailor it to them.

[00:13:37] Very exciting to hear you say this because it does seem like it could just prevent so much unneeded workplace violence, falls, things that could be potentially prevented if there was just another set of eyes ears in the room.

[00:13:52] Absolutely. Yeah.

[00:13:53] So, I kind of want to talk a little bit about the ethical considerations.

[00:13:56] Now, with AI, ambient listening, computer vision, what should we be considering?

[00:14:02] What are things that the industry should be keeping in mind as it comes to that ethical considerations?

[00:14:06] And also, maybe even patient privacy.

[00:14:08] Like, what are considerations the industry still needs to figure out?

[00:14:12] When we talk about outside of using large language models and making sure that the data that goes into those models is accurately representative and those kinds of things,

[00:14:21] which there's lots of great discussion happening right now about.

[00:14:24] And then looking for drift and looking for hallucinations.

[00:14:28] All really important in all aspects of any kinds of artificial intelligence that we're bringing forward.

[00:14:35] When we think about computer vision in particular, there's some interesting elements around bias in that the computers themselves, because just even how they work, might create bias.

[00:14:48] So, there's some great work that's out there about some data that demonstrates that some of the early computer vision models and even more recent computer vision models that are supposed to, for example, identify a person.

[00:15:03] There are racial and gender challenges with their ability to identify based on the way that it is looking for features and different things like that.

[00:15:15] And so, it's really important as we look to deploy these kinds of technologies that we, one, understand who the population is that we are providing that will be studied and included.

[00:15:28] And then making sure that it acts the same way throughout the situation.

[00:15:34] So, again, bias, I think, is a piece that we're going to have to work on.

[00:15:37] And we're still trying to fully understand.

[00:15:40] Now, in some instances, if we start with things that the computer vision is going to see that are less personal and more concrete and about coordinating, is there an IV pump in the room or not?

[00:15:53] Is there a patient in the bed or not?

[00:15:56] It doesn't matter what the gender of that patient is.

[00:15:58] It doesn't matter what the age of that patient is.

[00:16:00] It's just, it's a yes or no.

[00:16:01] Was the computer vision able to see a patient in that room?

[00:16:04] We can start there and then increase our sophistication over time in the tools that we're providing so that we make sure that we are thinking about those biases along the way.

[00:16:14] And then another big part of it, I think, is back to what we talked about earlier, which is the privacy side.

[00:16:20] And if we want to have trust that we should be putting these sensors in a room that are, you know, have a more invasive approach, we need to be able to educate about what they do and don't do.

[00:16:34] And, again, I think it's reminding people it's not about recording.

[00:16:37] There's nothing left over later.

[00:16:40] It's about educating and notifying people.

[00:16:42] Like, yes, we are using cameras and speakers in the room.

[00:16:45] And here's the value that they provide to you, that they're providing to your family, that they're providing to the health system team that's supporting you.

[00:16:54] And that is a piece that then we can relatively quickly get past.

[00:16:58] But it is a piece that needs education.

[00:17:01] And then I think the last piece is the tools themselves.

[00:17:04] We need ways to be able to say, we're not going to listen and watch right now, right?

[00:17:09] And let people make some of those decisions so that if there are situations that we still think we need to establish more comfort in, that we can do that and give the humans the control over when that technology is used and when it's not.

[00:17:22] Very interesting thoughts.

[00:17:23] So thank you so much for sharing your insights with us.

[00:17:25] Now, on Hi-T with Grace, we like to really get to know our women leaders.

[00:17:29] And so I want to hear from you a little bit about as you've gained all this wonderful knowledge, you've built this amazing career,

[00:17:36] you're really pushing forward this technology that is forward thinking.

[00:17:41] What are things that you do to just keep grounded and stay yourself, your own human woman self?

[00:17:48] I have children.

[00:17:50] Yeah.

[00:17:51] They'll ground you real quick.

[00:17:52] Nobody will remind you that you know nothing faster and more frequently than your kids who are like,

[00:17:58] Mom, I don't care what to do for your day job.

[00:18:00] You forgot to pay for my school pictures.

[00:18:02] And you're like, oh, yes, of course that.

[00:18:05] But in all honesty, I think family is such a big part for me.

[00:18:10] I'm involved in lots of my kids' activities.

[00:18:13] My husband, to his dismay, I'm now the chair of the board for my kids' ski team.

[00:18:20] And I'm out there with the ski team all the time and going to meets and things like that.

[00:18:25] But it's maintaining that connectedness to people.

[00:18:29] And I think, honestly, it's important in the world that we live in where there's so much technology around us and that our utilization and dependence on technology and artificial intelligence and thinking about things like bots and all those kinds of things coming into our lives.

[00:18:46] At the end of the day, who are we without human connection?

[00:18:50] We are humans on this earth to be humans with each other, to support each other, to learn from each other, to grow with each other.

[00:18:59] And for me, a lot of what I think what grounds me is personal human interaction, getting together with friends, getting together with family, traveling around the world and seeing how people do things in other places and learning from that.

[00:19:14] Because it's very easy to get into your own space and your own little world and then depend on a bunch of technology and kind of depersonalize yourself from the world.

[00:19:27] And I just, to me, again, bringing the joy back to healthcare, that is why that statement is so important to me.

[00:19:35] That's about human connectedness.

[00:19:37] That is about my ability as a physician to sit in a room with a patient and their family and potentially have a life-altering conversation and have it mean something and have an impact.

[00:19:50] I've got patient families from more than a decade ago that I don't have any clinical contact with anymore that I still maintain human connectedness with.

[00:19:59] I might do it through technology, but I maintain that connection.

[00:20:03] And so, I don't know, I think that if I had to distill it into one thing, it's human connectedness and my ability to appreciate the differences and wanting to be out there and meet and learn from all kinds of people and just enjoy this precious time we have on earth.

[00:20:20] That is very inspiring.

[00:20:22] And I think all of us needed to hear that today.

[00:20:24] So thank you for that.

[00:20:26] Now, before I forget, where can our listeners that have been able to hear so many great insights from you today find you online?

[00:20:31] I'm all over LinkedIn.

[00:20:33] So just look, Stephanie Lahr at LinkedIn.

[00:20:37] Less active on Twitter, but I'm there.

[00:20:39] You can find me.

[00:20:40] And certainly happy to have folks reach out to Artisite as well.

[00:20:44] And we're at artisite.com and have an ability to contact us there as well.

[00:20:48] Terrific.

[00:20:48] Thanks for joining us today.

[00:20:49] Yeah.

[00:20:50] Thanks so much, Grace.

[00:20:50] It was wonderful to see you.

[00:20:51] And thanks to you folks for joining us too.

[00:20:53] Check out the High Tea with Grace podcast for more interviews with great guests like Stephanie today.

[00:20:58] Cheers.

[00:20:59] Like a Girl Media is more than a media network.

[00:21:01] It's a community.

[00:21:03] We want to meet you and amplify your voice and the voices of outstanding women innovating in healthcare.

[00:21:09] Interested in starting your own podcast or hosting an event near you?

[00:21:13] Connect with us online or in person.

[00:21:15] We're here to support and empower you.

[00:21:18] Thank you.

[00:21:19] Thank you.