GEORGE NETSCHER: Hey, Rhonda. Thanks for having me.
RHONDA DEMENO: We're really excited to have you speak to our audience today about this very timely topic. Our discussion is going to focus on how AI can augment a fall management program through interpretation of AI-generated information that is effective in fall management and keeping residents and families more informed about fall events. Artificial intelligence can play a crucial role in preventing residents' falls in senior living facilities by providing continuous monitoring, early detection of falls, and timely interventions. Now let's begin our conversation by discussing fall events in senior living. George, can you elaborate on how often falls happen in senior living?
GEORGE NETSCHER: Yeah, I definitely can. Unfortunately, frequently. So we today, as an organization, are supporting with about 10,000 falls a month all over the country. I think we're in 36 states today. And what we see is that, on average, an organization will have about one on-the-ground event, not necessarily a fall, but what we call an OTG event. We'll have about one OTG event per resident per month.
The big challenge is that you don't know-- what we have-- frequently what we in the industry refer to as unwitnessed falls, where you come in and you find somebody on the floor and you don't know, did this person fall or did they actually get on the ground to pray and they can't get back up on their own? And they may have cognitive impairments and can't necessarily communicate that to us. And so we have a lot of these events, OTG events that we don't necessarily know whether they're falls or not. And we end up calling them unwitnessed falls.
Some of the kind of really interesting things we've seen from our data is that actually about 40% of the time, that person lowered themself to the ground. So they got on the ground intentionally and they can't necessarily get themselves back up. And we're calling a lot of these falls and we're sending folks to the emergency room that don't need to go and things like that. But statistically, what we'll see is about one OTG event per resident per month.
RHONDA DEMENO: And that's really good information. I did some research myself and noted from the CDC, they're saying that every second of every day, an older adult suffers a fall in the United States. And falls are the leading cause of injury and death in the US for seniors. So talk to me a little bit about how safely you can manage these falls.
GEORGE NETSCHER: The statistics really are sobering. I think what we know really as the society-- I was going to say as an industry, but this is just a global challenge-- is we don't have enough people, right? And so as there are an aging population, more and more folks, falls are just going to be more and more of a challenge. And we can't just keep throwing labor at it. We don't have enough young people to care for all the older folks that really do need more support.
I saw a statistic, might have been in a similar place where I think it was, in 2015, there were seven potential caregivers for every person over 80. And by 2030, there expected to be just four. So as that proportion shifts more and more, we have to find better ways. If we don't have enough people, how do we provide better ways to care for the folks that need care? And then when we think about tools, it's like-- of the tools that can make folks more productive and really take meaningful work off staff, AI is the tool.
I know there's a lot of hype around it and all of that. This is a really powerful enabling technology that can really-- it is going to be the tool that helps us bridge that gap. So I'm super excited about where AI is going and can hopefully help ground that conversation and what is AI really capable of and kind of cut through a little bit of the noise. But in terms of our program specifically, what we do is we put cameras in the private room for folks that choose to have it. It's really designed for folks with Alzheimer's and dementia.
I started the company for my mom who lives in a care community and has Alzheimer's. And a lot of the folks in our company are personally impacted in the same way. And we really have kind of this core vision that for folks like my mom, they lose the ability to advocate for themselves over time. And so for us, it's really about how do we give a voice to folks who can't necessarily tell us what they need. When you think about falls in particular, we really have kind of, like, two things that are meaningful there.
And so we'll put a camera in the private room for folks that choose to have it. Folks are opting in for the program. We take great pride in our median opt-in. It's about 90%. Folks get the choice. And the fact that so many folks opt in for it is really a great kind of source of pride for us that we're really doing something meaningful for folks. And then from that camera we use artificial intelligence to determine if somebody is on the ground, if we have an OTG event. And we don't keep anything else. So we detect somebody's on the ground, we alert the community, and we make just that video available.
So we can really have that kind of right privacy, safety trade off. Again, coming back to my mom, it's like when would people want us to be able to see? They wouldn't want us to be able to see if they're changing in their room or things like that. But they would want to be able to tell us if they hit their head or not or if they just sat on the floor and actually don't need to go out to the emergency room. So we can determine with truly world leading accuracy. We take huge pride in how just well the system works to detect if someone's on the ground, alert the community, make only that video available.
We can see if they actually hit their head or not or if they just sat on the floor. And then how do we actually fix it? So the second big piece of our program is we have our own clinical team that reviews every video, does a root cause analysis, understands really what is the unmet need here. And so we'll see things like all sorts of really person-centered kind of interventions that we can put in place where the folks in the community really need to be a jack of all trades and support with all sorts of different things.
And we can come in and be a real specialist around, OK, how is this person going to the ground? What does their gait look like? Are they getting dizzy when they're getting up from bed? And identify really what are the specific interventions we can put in place for this individual with the key data point they're being-- it turns out that only about 1 in 50 falls results in a serious injury. So we've got-- that's kind of a big misconception that we think every fall is a really serious event. We actually have, like, 49 out of 50 times on average to understand what an unmet need is and put in interventions that can drive real change.
And so we take huge pride in the impacts that result from that. We will consistently show on the order of 5x reduction in ER usage, both by not sending folks who don't need to go because they actually just sat on the floor and by really understanding, what is this unmet need? How can we kind of address the specific challenges for this individual?
RHONDA DEMENO: You touched on some really informative, great information talking points. Our first episode 1 and 2, we talked about fall management can't be a one-size-fits-all solution for every resident. And I think what you're saying is you really do-- based on the information that you get, you're able to customize the care for the residents based on the root cause analysis and different information that your care team is collecting based on the type of fall that the resident had. Is that correct?
GEORGE NETSCHER: That's exactly right.
RHONDA DEMENO: How does AI fit into a senior living solution? There's a lot of AI emerging today for many different-- not just for fall management, but for infection control and other areas. How do you see it fitting in for senior living? You brought on some really good points about the reduction in workforce, but is there any other benefits that you're seeing?
GEORGE NETSCHER: Yeah, so I think AI has come in kind of a couple of waves. And to give a little bit of my background on this, so this company actually started as my PhD research in the computer science department at UC Berkeley, where I was working in top ranked AI research lab, specifically, with this kind of core. Before the company started, it was just, how do we apply AI to support people with Alzheimer's disease? And there's got to be some overlap. And so spent a lot of time thinking about AI and working on problems in AI. And it's gone through a couple of different waves.
So if you think about what AI is really good at, it's basically recognizing patterns and data, right? And that's why people say things like data is the new oil, because the more and more data you have, the better and better you can detect specific patterns. So if you're thinking about a lot of the hype 10 years ago was about self-driving cars, as an example. And if you think about what kind of enabled some of that stuff is that the more and more images we have of people driving on the roads and things like that, the better and better we can detect a person versus another vehicle on the road versus, say, a stroller, things like that.
And so that's where a lot of the excitement was, really about 10 years ago now, was around our ability to have kind of human level understanding of the world around us. So if I'm looking at a video feed, I can tell if somebody is on the ground with really high accuracy if I have a lot of data, as an example. More recently with things like ChatGPT, there's been a lot of excitement around kind of flipping the picture and what we call generative AI where now we have-- if you've used ChatGPT, you get an output that looks a lot like a person talking back to you.
And so we're getting really good at the computer not only interpreting an image as an example or a piece of text, but actually generating an image or a piece of text. You might have seen generated artwork and things like that. But it starts to create some kind of really interesting opportunities for can I have-- be much more efficient with my sales and marketing spend, as an example, where I can generate marketing copy? And there are companies that are-- I know Jasper is a company like this that they're a whole business and they're growing really quickly.
Outside of senior living, they sell to, like, the Fortune 500 where they will create marketing content that's in the voice of your specific company. So now I can have one marketing person who can just do much more because they're able to generate outputs for them that they can edit instead of needing to create everything from scratch. So I think there's a lot of opportunities there. And I think we're kind of just scratching the surface and you'll see more and more of it in years to come.
RHONDA DEMENO: Yeah, so you've touched on some broad applicability beyond falls just now. Once you can detect something in the room, where does it go from there?
GEORGE NETSCHER: Yeah, great question. So basically, there's a camera in the room. We determine if somebody is on the ground. That determination is made locally. So we're basically running our AI in the building itself. And the AI is determining that there's an on-the-ground event. And then that triggers a workflow where the community gets notified that, hey, there's potentially somebody on the ground here. The video is made immediately available at the same time. And that data is uploaded to the cloud.
So only when an on-the-ground event is detected is the data uploaded and it's made available in a web portal where folks can come and access it. Not to get too much into the nitty gritty, but we take a lot of pride in our privacy and security protections and things like that. So all fully HIPAA compliant. Everything is encrypted when it's both in transition and at rest. Ultimately, the data belongs to the community. So we are what's called the custodians of the data. But for folks that use our service, they ultimately own it.
If they choose to leave us and not use the service anymore, then that data would be transferred to them and it's fully theirs. Our retention rates are kind of through the roof, which we take a lot of pride in. I can speak to a little as helpful. And so we really don't lose customers. But if folks did leave, then folks do fully own their own data, which we think is an important part of all of this. We've kind of been a rifle shot into the industry around a very specific need around fall management, but there's so much more we can do.
And that's really the way this AI stuff works, is once you start collecting data, you can use that data to do some really interesting things. So we now have the largest ever video data set that's been collected in this space and we can train models that recognize not just falls, but almost anything in the room. And the way that AI stuff works is the more and more data you have, the higher and higher accuracy you can get but also the kind of harder and harder problems you can approach.
And so one of the things that we're super excited about right now is we actually have enough data to tell the difference between a resident and a staff member. So now you can think about, if I'm a fly on the wall in a resident room, what are all the things I would want to know? Well, one of them is our most expensive resource is staffing. And so how do I best utilize my staffing? And so if we can actually measure how much time staff spend in each room in a way that doesn't require them to document anything or wear anything or things like that, we can both earn the right to be in the room by delivering a really important service around fall management.
But also once we're in the room, we can now collect information that this industry has never had access to before. So we can understand things like levels of care and staff utilization and how to make the best use of our staff and things like that so that we can support them in the best way. In other industries where staffing is the number one cost, which is often the case, if you think about Amazon as an example, you know that they are measuring exactly how their staff are used and how best to utilize them.
And optimizing over time and understanding when they try a new solution, does it actually make an impact for their staff and things like that? And bring this very data driven approach to staff management where we see an opportunity to just really drive the industry forward at a time when we know margin compression and all of that is happening. As inflation has made such an impact and interest rates have gone up so much, there's such an opportunity for us to really drive both top line and bottom line performance for folks.
RHONDA DEMENO: So from all this data, I mean, it's amazing that you can determine the difference between a staff member or a resident. With all the information that you've gathered, are there some common fall prevention interventions you recommend or is it really unique for each individual?
GEORGE NETSCHER: Yeah, great question, bringing it back to falls. It really is unique for each individual. And I think this kind of talks to how medicine works in general. In medicine, we're great at hitting kind of the center of the bell curve. So what folks have had historically is, like, we consistent lighting is helpful, things like that. And so folks will try things that are-- we're going to put a nightlight in place because we know consistent lighting can be helpful. But then when we actually look at specific use cases, it's really individual.
We just had one as an example where the resident was trying to get out of bed independently and move himself into a wheelchair. The wheelchair is kind of facing the bed and he's got his arms on the arms of the wheelchair and unfortunately ends up slipping out from under him and he has a fall. Really unfortunate, but what we can see from that is, oh, this man does have the ability to transfer himself but this wheelchair is sliding out from under him.
And the intervention that the clinical team identified was that there's actually a device you can put underneath a wheelchair, which I didn't know this thing existed. It's not an expensive device. But what it does is it locks the wheels when there's no weight placed on the seat of the wheelchair. And then when somebody actually sits down, it releases the wheels. So now this man is able to actually have the wheelchair be locked and be kind of weight bearing so he can transfer himself over.
And then once he sits down, it releases and he can move the wheelchair. I had no idea something like this existed. But it kind of shows you, one, falls are really multifactorial. So there's both also the ability to come in and support with gait training and core strength training and things like that. But, two, we can really understand these specific unmet needs for folks where you can't expect the clinical team within a building to know and stay up to date with all sorts of different fall prevention best practices.
But we can have a team that's-- that's their whole job. They live and breathe fall management. And we can identify exactly what's happening here, identify the root cause and get the specific right intervention for this person at this point in time, which will change over time as folks move through their journey and being able to be that really person-centered care is our whole mission.
RHONDA DEMENO: Let's pivot a bit to the cost of falls. WTW completes an annual claims benchmark study and we have some data, but can you share information that you have on the cost of falls?
GEORGE NETSCHER: Yeah, we actually did a big-- it focused on a Google. We did a big survey on this last year and released what we call the state of falls. And what we did was we went out and interviewed folks throughout senior housing, what they estimate the cost of a fall is. And this is specifically a cost of a fall with injury. And what we found was that the cost of a fall with injury is about $5,000 for a senior housing operator. That doesn't include spend that other people are paying. So if you send somebody to the emergency room, then it's going to cost the health insurance money, it's going to cost the family money.
What we're interested in here was, like, what specifically is the cost for a senior housing operator? And it came down to about $5,000 that comes from a couple of different buckets. One is certainly the cost of a potential litigation that certain number of falls will result in claim. We know that falls are the number one cause of claim and all of that, particularly when we repeat falls and particularly when somebody with Alzheimer's and dementia and we don't know what's happening and all of that. But there's also a bunch of other cost centers.
So the amount of staff time, the loss of revenue, both-- if somebody spent some time in the hospital, often folks will still charge rent but they won't be charging care services during that time there is a likelihood that person is going to move out for whatever reason. And so there's a whole bunch of buckets that can be hard for folks to track and trace. But it ends up being about $5,000 for each fall.
RHONDA DEMENO: And that's probably not even taken into account employees' injuries because we know that oftentimes, when a resident falls, employees may get injured from improper moving and positioning. So there's just many different ways that we could look at that data for sure.
GEORGE NETSCHER: Yeah.
RHONDA DEMENO: So does SafelyYou have any data showing a reduction in the cost of claims?
GEORGE NETSCHER: Yeah, we do. And we don't have a ton because, obviously, only a certain number of falls will result in claim. And so we're still a pretty small sample size here. So what we have shown is we will consistently reduce the number of falls and the number of emergency room visits that come from falls. And we're very confident in that. We show it every time. And then what you'll see is, over a five-year period, a certain number of claims coming out. And we can start to compare the number of claims that were there with SafelyYou versus the number of claims that were happening before SafelyYou.
And we hear a lot of anecdotal stories of this of, hey, we didn't have to settle something we would have settled beforehand because we could show that this person was not at fault, things like that. And so we've had some wonderful cases there or cases where a resident claims they were on the ground for hours. And we can see, no, they were actually on the ground for minutes. And this isn't somebody that's lying, but they have Alzheimer's disease, right? And so they legitimately believe that. Or we had a case recently where this lady had bruising that looked like fingerprints on her thigh.
And obviously, when in an emergency room doctor, a family sees that, it's really concerning. Like, oh, my god, what is this bruising? And potentially abuse allegations and things like that. And what we could see was that it was actually from the resident herself. And then there's opportunities there to go in and understand, how do we address that unmet need? But it's not that someone's abusing her. She's actually putting so much weight on her thighs when she's getting herself up. And so there's these opportunities here where we can really, again, be a voice for that resident where folks don't need to assume the worst.
We can really understand what's the unmet need here. So from the claim perspective, what we have seen-- to answer your question concretely-- is on the order of a 20% to 30% reduction in the number of claims and the size of claims, but we're still at a small sample size there because we really need of five years of data from a whole bunch of different operators. So we're heading in the right direction but we're not at a point yet where I would say you can expect x, y, z to happen. But a lot of hopeful anecdotal information.
RHONDA DEMENO: So have you been involved in the underwriting process or do you feel that communities that work with your application, there may be some benefits in underwriting?
GEORGE NETSCHER: Yeah, we definitely have had organizations that have been able to get reduced premiums or risk management dollars to help pay for fall management programs, things like that, which is always great to see. I would love for a carrier to partner with us. And because we know-- what I've heard is falls are 40% to 50% of claims. But I've even heard recently that like 70% of claims are coming from fall incidents. And so if we know that and for the first time, through SafelyYou, we have a standard definition of what an on-the-ground event is, that it's not self-reported.
It's determined by the technology. And we score each one in terms of severity. It feels like such an opportunity to use that information in underwriting so we can really show-- not having to wait for five years of data, we can actually show that we put this program in place, we can show we have a reduced number of fall events. Within the first three months, we can show that and then use that to underwrite. I would love that to be us, to use it as a leading indicator to really it should be a win-win-win for the carrier, for the operator, for SafelyYou, for all of us together to help kind move the industry forward as we just use data to make better and better decisions.
RHONDA DEMENO: And we did do a podcast on underwriting that ties perfectly into this conversation and into our falls management. So we hope those that are attending this podcast will listen in to that episode as well because we do discuss underwriting. So we're running out of time, but I'd like to end on future developments. I know we say SafelyYou is more for memory care and those types of residents.
But do you see, in the future, any development for independent living or assisted living or skilled nursing or providers utilizing your technology in those settings? And do you have future developments that you'd like to share today?
GEORGE NETSCHER: I think behind all of this, the key is really data. And I think that's the kind of one takeaway folks should have of, what makes work really well? It's data. And so it's a really fun point. In our trajectory as a company where we've now got so much data in this care setting that no one has ever had before, that it's a really fun time for us that just, every quarter, we get to start releasing the next thing. And there's so much opportunity, both around care, but also just the operations, the business, and things like that. So it's a really exciting time for us to be able to just deliver so much value for this space that is really needed to take work off staff, take stress off staff, and really support in a meaningful way.
RHONDA DEMENO: And you have any final comments, George, that you'd like to share about SafelyYou or any other comments about fall management and artificial intelligence?
GEORGE NETSCHER: Yeah, so we just had our annual summit yesterday and we had some really interesting conversations. I think the world is really bright for the capabilities of AI coming out of it. It can be hard to cut through the noise, so there's, I'm sure, a lot of people that are pitching all sorts of different things. And I would say the right way to vet a vendor is look at the size of the company, look at the amount of data they have. One of the kind of challenges in AI is that it's not that hard to do something with 90% accuracy.
It takes 10 times more work to get to 99. It takes 10 times more work than that to get to 99.9% accuracy. And so I know it can be really hard, especially when you don't have a real understanding of AI and things like that. We're always very happy outside of our specific product to help folks vet out, hey, I'm looking at this thing for marketing or things like that. Definitely, we can be supportive there of helping folks vet what ifs, what can we really expect from AI? It can fail in really hard ways to understand.
I can give an example from our perspective, we might have world-leading accuracy from fall detection, but it's not consistent across the board basically because all of this stuff is trained from data. And so if there's a specific example that we just haven't seen before, then we're going to perform poorly in one specific room until we kind of understand, we see those examples, we train the model, and all of that. And so it can be a little hard for folks to understand, like, I thought your accuracy was incredible.
And it's like, it is, but for this specific family we've had three missed falls in this room, as an example. So you'll see that same kind of thing across the board when you're deploying AI solutions that, if the data set doesn't have diversity in it, it may very well be that the AI performs really well in this specific setting or we're using generating marketing material as an example again. It might perform really well in another industry. But then when we're trying to use it here, it actually doesn't work that well and it feels kind of clunky and things like that because we haven't trained it on data in this space.
So I think that can be one of the hard things, is that if anybody has used ChatGPT, the answers can feel incredible. But then you get an answer that it's like, this is very wrong and it's very confidently wrong. So we have a lot of things that seem really convincing. But if you're actually an expert in the space, you're like, this is not right. And then that undermines trust. And I think the whole field has a long way to go but there's important work happening. And how do we build trust over time and be able to have better answers to when we're really confident in the answer versus when we're not really sure and you should double check this answer and things like that?
RHONDA DEMENO: Yes, I agree with you. And I'm really excited to see what the future holds in regard to AI. But we really want to note that AI at this point should complement human caregivers. It should augment the care and not be just the one and only solution. You mentioned privacy in your earlier points of discussion. With the emergence of artificial intelligence, there are a lot of privacy concerns.
Can you elaborate on privacy protection and what SafelyYou has done to secure personal health care information and to protect this information?
GEORGE NETSCHER: Yeah, absolutely. So from my vantage point, it's really about transparency and choice, basically. So you should be transparent about use of data and all of that and give folks a choice to if these are the implications of using your system and here's the benefits I get, then it's my choice of whether I want that system. And so as an example, for anybody that-- our program is fully opt in. And for any family or responsible party that opts in for the program, they sign a thing that says, basically, this is how the data is going to be used.
And I'm opting in for this program that is going to have these benefits but it also is not going to have perfect accuracy and it may fail and things like that. And so we think, basically, it's really important to give folks choice then for us to do all the right things to make sure that we're protecting the data on the background-- in the background. So we are not only fully HIPAA compliant, we are also fully SOC 2 compliant, which touches more on the security side of things.
That's something that no other vendor does in this space that we take a lot of pride in that we take security that seriously that we went and became fully SOC 2 compliant. So think it, again, kind of-- it comes down to being transparent, providing the right protections, giving folks the right choices. And ultimately, you can have-- the reason we get the opt-in rate we do is because there is a balance between privacy and safety, right?
And the ability to only keep data, only keep that video when somebody would want to be able to tell us is at the core of everything we believe. I keep bringing it back to my mom and what would my mom want? And my mom would not want us to be able to see all the time, right? Like, she might have a boyfriend over and I want her to be able to do that. And I can tell you for sure, I do not want to be able to see it, you know? So it's really about, how do we empower the right trade offs for folks? And so when would she want to be able to tell us?
She would want to be able to tell us if she hit her head or not. She would want to be able to tell us in kind of certain other specific situations. And so it's all about that, how do we empower giving residents a voice? And when we think about some of the stuff I've talked about around being able to tell the difference between staff and residents, as an example, a very similar thing there where what we have today in the industry is if you have somebody who's cognitively healthy, you can have a conversation with them about the level of care and say, hey, I see you may be having trouble showering yourself.
Would you like assistance showering? But that's not the case for somebody with Alzheimer's or dementia because they don't have that voice. And so you end up going to a family member instead and saying, hey, we can see your mom is having trouble showering. And the family member may say something like, what are you talking about? I was just with my mom. She's not having trouble and you're just trying to charge me more. And so because we don't have that kind of transparency, and the individual doesn't have a voice, it creates all these challenges in the industry. There's such an opportunity here to kind of use AI to really give a voice to folks like my mom.
RHONDA DEMENO: Thank you very much, George, for attending our call and speaking to our audience today. We appreciate you. I know I've learned an awful lot today.
GEORGE NETSCHER: Thanks for having me. This is important work.
RHONDA DEMENO: For those of you that are attending our podcast, if you want to learn more information about George and SafelyYou, please go to our podcast page. This concludes our fourth episode of our fall management series. Again, we hope you listen to our next episode number five that will focus on event reporting, conducting root cause analysis and disclosure. Thank you all very much for attending our call today.
PRESENTER: Thank you for joining us for this WTW podcast featuring the latest perspectives on the intersection of people, capital, and risk. For more information, visit the Insights section of wtwco.com. WTW hopes you found the general information provided in this podcast informative and helpful. The information contained herein is not intended to constitute legal or other professional advice and should not be relied upon in lieu of consultation with your own legal advisors.
In the event you would like more information regarding your insurance coverage, please do not hesitate to reach out to us. In North America, WTW offers insurance products through licensed entities, including Willis Towers Watson Northeast Incorporated in the United States and Willis Canada Incorporated in Canada.
