CIBC Innovation Banking Podcast

Michael: [00:00:01] We talked to the entrepreneurs who have made their mark thanks to a fearless passion for what they do and for the lessons they've learned, we learn how to be better at running our own businesses, engaging our own clients and exploring new ways of thinking about the innovation economy. On this episode, we discuss an issue top of mind during the worst pandemic in more than 100 years: caring for our elderly parents at home. [00:00:27][25.7]

Naomi: [00:00:28] You know, we see that the signs and, you know, red flags have been raised for many years, you know, in terms of shortages of help. But until a pandemic came along, it wasn't really tested to its fullest. And as soon as it had to be, I mean, the cracks were apparent right away. [00:00:44][15.8]

Michael: [00:00:48] Hello, I'm Michael Hainsworth. The CIBC Innovation Banking Podcast explores the world of startups, growth stage companies and late stage companies that have made a big splash in their industries around the world. [00:01:00][11.7]

Michael: [00:01:10] COVID-19 has shone the spotlight on the importance of providing services in a familiar setting, avoiding crowded clinics in hospitals and the attendant quality of life improvements that come from having personal support workers and others come to the patient, not the other way around. And just like how the coronavirus forced us to advance digitalization efforts in schools and the workplace, the innovation economy at AlayaCare has helped give it the strong standing needed to bring 21st century solutions to problems first revealed by the founder of Modern Nursing, Florence Nightingale. She, too, was a pioneer in data visualization during the Victorian age. We now leap forward 166 years to the modern day nursing movement, where personal support workers armed with smartphones and big data help keep both patients and staff happy and healthy. For insight into the present state of nursing and its future, we turn to the head of research and development at AlayaCare, Naomi Goldapple. She says if there's one thing COVID-19 has done, it's taught us that people do better at home. [00:02:20][69.8]

Naomi: [00:02:22] There's been studies done saying that definitely the elderly fare better when they're cared for at home. But this pandemic really shone the light on the actual facts, the numbers, the sheer chasm between the two different methods, which I think was shocking for many people. You know, we see that the signs and, you know, red flags has been raised for many years, you know, in terms of shortages of help. But until a pandemic came along, it wasn't really tested to its fullest. And as soon as it had to be I mean, the cracks were apparent right away, which, you know, is unfortunate for everybody and very much hoping that on the other side of this pandemic that we have better methods and a better response for dealing with this. [00:03:07][45.3]

Michael: [00:03:08] There's a sense that COVID-19 woke everyone up to the realization that there simply aren't enough people in the home care industry. [00:03:14][6.1]

Naomi: [00:03:15] Absolutely. I mean, it's always been known for people in the home care industry, but for everybody else, you know, you can say the same people that work in home care are, you know, a lot of the same people that would be working in facilities as well. But that general skill, that caregiver skill, there is simply not enough. And especially when, you know, we're closing down borders because of the pandemic, that definitely cuts off some of the immigrants who come in and make up a large percentage of those caregivers. So it's even more restrictive than it was before. [00:03:45][30.3]

Michael: [00:03:46] So when AlayaCare did the research, you learned that the adult daughter is generally the one who is largely responsible for doing the heavy lifting in a home care environment. How did that help you tailor the work that you do? [00:03:59][13.3]

Naomi: [00:04:00] We are one removed from the end client. So we are B to B to C, so our clients are the actual agencies. So the adult daughter is really the agency's client. We have new functionality that really appeals to getting the whole family involved. In fact, a little bit, so it doesn't only fall on the adult daughter. So we have things like a family portal where anybody or everybody who is who is authorized to can hop on and get all the information that they need about the care, about the medication, the next visits, the billing, so it can be more of a, you know, sharing the caring of your elderly loved one with the village as opposed to just, you know, falling on that one person. [00:04:39][39.0]

Michael: [00:04:39] When we talk to people in the wireless industry, they talk about five years worth of growth plans that were consumed within the first two months of COVID-19 rearing its ugly head. How is the global pandemic affected your roadmap? [00:04:51][12.1]

Naomi: [00:04:52] Definitely there's been a lot of switch to whatever can be done virtual will be done virtual. So we had to quickly, when this pandemic happened last February, quickly move up virtual care functionality in our platform, further up on the roadmap than it was before. So the ability to do visits virtually through a video chat or even just by phone, right through the app that definitely moved up. Obviously, you can't do everything virtually. You can't, you know, give baths, you can't help people with certain activities of daily living. But there are certain things like therapy, and a lot of the mental health can definitely be delivered virtually. And there's also I know in my team where we're looking a lot more in terms of remote patient monitoring and how can we take advantage of a lot of the devices that are out there to be able to collect remote data seamlessly so that we can have a pulse on how they're doing at home without being able to visit as much as before? [00:05:47][54.7]

Michael: [00:05:48] I understand virtual care was always on your roadmap, but you managed to build a product within four weeks. [00:05:54][6.2]

Naomi: [00:05:55] It definitely came together very, very quickly because it had to. I mean, obviously, that means that you know, you divert some resources off of some other projects, but that was absolutely necessary so that we wouldn't be cutting off visit capability 100 percent. [00:06:09][14.3]

Michael: [00:06:10] Tell me about creating a constellation map as we deal with the issue of contact tracing and what your app now provides. [00:06:16][6.0]

Naomi: [00:06:17] You know, at the beginning, it's funny how a year later you feel like, oh, that was so first wave, right? Because we're a lot smarter now. But at the beginning, of course, you know, we quickly put in something called screeners so that every time a caregiver would go do a visit, they would make sure that they would screen themselves for any symptoms, but also screen their patient or their client for any symptoms. So if there was anybody that had a positive symptom, then that quickly we would be able to trace who that caregiver has seen in the past two weeks, let's say. And who has that client seen in the past two weeks? We try to, you know, maximize continuity of care so that the people at home get to see the same caregivers and develop relationships. But sometimes they do have different people. So we could very quickly with our system, we could see who visited whom and make a quick map and be able to close off that circle of care so that we would avoid any further exposure. [00:07:09][52.3]

Michael: [00:07:10] What do you say to an entrepreneur who finds themselves in a similar situation where they have to accelerate the development of a project, you know, four weeks is a remarkably short time to build something out? I can only imagine there must be a few hiccups along the way. [00:07:22][12.7]

Naomi: [00:07:23] Absolutely. I mean, you know, a lot of times there's some things that you have to sacrifice. You know, when you do proper product management, you do a lot of your research, you have a lot of clients involved. You make sure that you're testing a lot along the way here. You have to cut some corners, make some assumptions and kind of iron out some of the kinks as you're going along, as opposed to anticipating everything and then waiting until everything is ticked off. But it really is a race against the clock because, what we wanted to avoid we don't want all of our clients to quickly move to an off platform. So taking advantage of Teams and Zoom and all that, we prefer that people are within the platform. Then we keep all the data within the platform and there's a much better, seamless experience in terms of doing the visit virtually. So we wanted to make sure that we would keep them in app. [00:08:08][45.4]

Michael: [00:08:09] On top of developing this virtual care platform within that four week timeframe, you released in March 2020 a COVID-19 screener specifically and I understand that only came to fruition within three days. I can imagine the worst thing to do would be to have in the world of home care, a patient have to go back to the hospital, very critical to be able to incorporate this kind of technology into your app fast. [00:08:34][24.9]

Naomi: [00:08:35] Absolutely. I mean, the beauty is a screener is really another form. So because in our system, you can easily create forms because there's many forms that you have to do that you have to fill out to track a lot of information. So that base functionality was there, but then to make sure that it was in the workflow so that you had to do it. So we're like, how do we force people to fill out these screeners and collect all that data so that we can actually look at the patterns and see which geographic areas are more symptomatic, which demographics? This really helped us to understand where it was spreading and where you have to keep an eye out. [00:09:09][34.7]

Michael: [00:09:12] Artificial intelligence promises to remake the health care industry, put machine learning requires big data pools, the kind of content that home care workers collect, but often isn't analyzed in a timely fashion. But there's another type of big data that AlayaCare sees as critical to supporting patients in the comfort of their own homes. Data about the providers of that health care. It's isolating work, not particularly well paid, and it's hard work. Home Care Pulse reported in twenty eighteen, but the churn rate among support workers was 82 percent. One year later, that fell to sixty four percent. One of GoldApple's big advances involves analyzing the language of the support workers, monitoring their frames of mind through their words to ensure that support workers remain happy and employed. [00:10:01][48.5]

Michael: [00:10:07] What role is artificial intelligence playing in the development at AlayaCare, because you're working with Scale A.I. for monitoring the health of both patients and the workers? [00:10:16][8.9]

Naomi: [00:10:16] We did a year long project with Scale, and they were focused on really two things. More on the efficiency side. Definitely, one of the big problems that we talked about at the beginning is not only, you know, that there aren't enough caregivers, but there's also a big problem with caregiver retention and churn. The churn numbers are very high. A lot of times they're high because they don't have schedules that allow them to feed their families. So definitely the technologies that we're leveraging are optimization technologies. So travel and schedule optimization, because if we can make their schedules more efficient, then they're going to have more hours and they're going to be happier and they're going to stay around longer. So it's really quite simple, but it's harder to put into practice. So we are definitely leveraging optimization technologies to be able to do that. And then in terms of churn, we also want the schedulers and the owners to be able to understand what are the signs and symptoms that somebody might be at risk of churning and then you could have intervention strategies. So this is where we developed a predictive model to be able to understand which caregivers are at risk and be able to have intervention strategies and divert them. And that could be as simple as giving them more hours, readjusting their schedules, giving them a call. There's many, many things. But bubbling up that data for them, there's always squeaky wheels. You know, the people that are going to call every Monday morning and say, I'm not happy with my schedule, but for, you know, five that call, there's going to be another 15, 20 or more who just, you know, quietly become more and more dissatisfied. And so if we bubble that data up, then they can it can be actionable. [00:11:50][93.5]

Michael: [00:11:51] I can imagine for the home care worker, you know, it's a rather isolating experience. You know, we're all talking about working from home and not seeing our colleagues. But these are people who almost never see their colleagues. And in some cases, it's not well paid. It's very hard work. The churn rate issue is huge. But I can imagine it's not a function of them jumping to a competitor. It's them leaving the industry. [00:12:13][22.6]

Naomi: [00:12:14] Not necessarily. A lot of times it's working for a few agencies at the same time. So trying to fill up those gaps during the day. We do have some customers that, you know, I remember we had one of our U.S. customers that said their biggest competitor is Burger King because they can easily just go and, you know, flip burgers for a higher salary than they can. You know, it's a little bit different in Canada, but some of them do do it because it is a profession and it's a calling. But when you are doing it because it pays the bills, then you have to make sure that you have a full schedule. So, you know, one of the things that we saw in our research is that when you start to see if there's too big of a lag between the date that you hire them and when they get their first visit, you're going to lose them if that if that lag is too big because they need to be earning money. So if they see they're not getting hours there, they're going to start to go get ours elsewhere and you start to see their utilization dropping. And those are signs and symptoms that they are not very engaged in your company and you have to find out why. [00:13:08][54.4]

Michael: [00:13:09] So you've built a big data churn model that predicts which workers are at risk of quitting. So you can have intervention strategies to stop that. [00:13:16][7.0]

Naomi: [00:13:16] Yeah, we call it actually a satisfaction score. So basically putting some variables into a model and then coming out with the score. Workers that have higher scores, we assume they're pretty engaged. They're pretty satisfied. You can kind of look at your bottom 10 percent and say, you know what, these ones are at risk. Many different intervention strategies could be as simple as just giving them a call, bringing them back into the fold. As you said, you know, teamwork is so important. And prior to COVID, some of the more successful agencies, they definitely make a point to have a lot of team meetings and start the day off meeting the team, have a lot of trainings, make them feel like they're part of something bigger. And this is definitely been, you know, more of a challenge. [00:13:54][37.7]

Michael: [00:13:55] So you've built into the app the ability to analyze the writing of the care worker when they present their report at the end of a visit to be able to determine just based upon the words they've chosen, whether or not they're likely to quit or move on someone else. [00:14:11][15.6]

Naomi: [00:14:11] We're analyzing the actual text, we definitely discovered that there's a lot of unstructured data within our platform. And our hypothesis at the beginning was can we see by their progress notes whether there's some some signals in the data that tell us that they might be at risk of churning? In fact, when we started that research, we actually came up with something else as we discovered that in that unstructured text, there's a lot of very important clinical information about the status of their actual patients. So we've actually been moving towards and one thing that we're piloting right now is automatic, important flag NLP. So basically the NLP model will read through these intervention notes. So when there is a visit that happens and they will see if there's certain things that are very important, that are urgent that the clinical supervisor needs to know about right away. So things like, you know, was there a fall? Is there any hospitalization, is there any bedsores that are getting worse? Anything to do about COVID? These things are urgent. So once they're written, they want to be red flagged right away. And then there's other things that are important. But they could, you know, kind of get a digest at the end of the day. Our hypothesis here is that by alerting them quicker to what is being written because sometimes they might not get to it, like to actually read through all the notes, if they can get that information quicker, we can actually move the needle on the outcomes. So one of the biggest outcomes is preventing hospitalization. So when somebody comes into your care, out of the hospital, into home care, one of the worst things you want and you want to avoid is for them to go back into the hospital. If we can, by giving them this important information sooner, if we can prevent hospitalizations, prevent falls, concretely, measure those. That's huge. [00:15:55][103.1]

Michael: [00:15:59] Goldapple recognizes the importance of artificial intelligence. Many within the health care space do. But artificial intelligence can be poorly implemented and downright dangerous if done wrong. Goldapple believes the secret to making machine learning an effective tool is in its implementation. [00:16:15][16.3]

Michael: [00:16:19] Let's talk a little bit more about the use of artificial intelligence, machine learning and the big data that feeds all of that, because A.I. is more than just an API that a developer can drag and drop into an application. What unique lessons did you have to learn when it comes to the science and I suppose the art of artificial intelligence? [00:16:40][20.5]

Naomi: [00:16:43] So one of the bigger ones is, you know, the actual machine learning model is, you know, about five percent of the entire process, right? So it's what people talk about the most, but there's a lot of kind of dirty data extraction and cleaning that has to happen at the beginning. And even before that, you need a data pipeline. Right. So you need to have a good flow of data coming in that's going to constantly be retraining your system. That's going to be constantly be feeding a model so that you can have near real time predictions. There's all of that that happens before and then there's actually putting this into production. So there's a lot of data engineering that has to happen to be able to make this seamlessly be part of your of your application. So whereas before, you know, we talked a lot about the actual machine learning experts that can tweak algorithms, that can tune algorithms, that's great. But now when you look at a data scientist, they really need the skills, right from, you know, extraction, cleaning all the way into putting it into production. [00:17:41][58.4]

Michael: [00:17:42] Right. The idea of garbage in, garbage out. If the data that's going into your model is not cleaned and efficient, and effective, then the output of that machine learning algorithm is going to send you the wrong way. How do you audit the outputs of artificial intelligence to ensure that you're not sending your company down the wrong path? [00:18:03][20.8]

Naomi: [00:18:04] Absolutely. So there's also there's something called feedback loops. We make sure that in training you make sure that you have a certain level of accuracy for your machine learning model. And then you have to make sure that you're constantly measuring that level of accuracy, make sure that it's not drifting too far away. And you have to make sure that the data that is feeding this model that was trained, that that data doesn't start drifting from your original data distribution that you had, because if it does, then the model needs to be retrained. So that's why this whole data engineering, you need to engineer it so that there's a feedback loop, so that you see whether your prediction was correct and that feedback goes in and retrains to make sure that you're always going to maintain your level of accuracy. So there's a lot more kind of engineering that goes in than one would think to make sure that that whole process is really seamless and is constantly updating, which is very different than just building regular technology where, you know, you put in code and it's just always going to do that until, you know, for the next 30, 40 years. Here, it's a bit of a living, breathing animal where it's going to constantly learn. And you have to make sure that it keeps learning the right things, you keep giving it the right reinforcement and that you maintain those levels of accuracy. [00:19:14][69.8]

Michael: [00:19:15] So then help an executive who recognizes their company may need to pivot to machine learning and artificial intelligence. I can imagine the corner office says well we've got an I.T. department will just get them to deal with it. [00:19:27][12.0]

Naomi: [00:19:27] It's changed a lot in the past couple of years and there's definitely a lot of auto ML available. I mean, there's a lot of technology. The Googles, the AWSs is they're offering up all these auto machine learning tools that your IT department with a little bit of training can start to put into production. [00:19:42][15.1]

Michael: [00:19:43] But it strikes me that it's critical that we recognize that your I.T. guy who's going to fix your printer is not the person you want to be dealing with artificial intelligence. [00:19:52][9.2]

Naomi: [00:19:53] Absolutely not. Not the one who fixes your printer, but people who are developers. It can definitely start to learn these skills. You know, the important thing is what you want to do is not just say we're using ML, but you want to actually create value. And that's really the harder thing to do because you can have you know, we just talked about having a prediction score. The idea you have to think about is who is going to be consuming this prediction score and how is it actionable? What does it mean? So if I see a risk level that is, you know, on a score of one to five and I see it's out of three, what do I do? Right. So you have to think about how is this actually actionable? You have to be bringing your product people into the mix. You have to be talking to the customer to understand what value this is supposed to be producing, because there is a lot of people like, yes, yes, this is based on machine learning and they have something in their system that has machine learning, but it's not necessarily creating the value that's needed, understanding not only how the models work, but actually understanding how it's going to be applied. The whole applied part is extremely important. So there's a big difference between actually developing algorithms and actually taking, you know, kind of open source algorithms and putting them into practice. It's all really in the execution. That's a lot more critical. [00:21:10][76.7]

Michael: [00:21:11] How unique are you in the industry here, because when it comes to home care in this space, traditionally not a very sophisticated industry and I don't know of any other company in your environment, in your world that has an R&D arm. [00:21:27][16.0]

Naomi: [00:21:28] This is true. And so this is definitely making us unique. I have found over the past year that there's a lot more interest in our customer base. Right. To actually take advantage of some of these more emerging technologies. It's interesting, I just had a call yesterday with a relatively sophisticated home care agency who's quite proud of their setup. They were saying that, you know, if they want to use any artificial intelligence, they can you know, they can use what's available today. One of the things that makes us unique is that in AlayaCare, we have specific home care data sets. So if you take a natural language programing model that's been trained, you know, just on general Wikipedia, it's not going to perform as well to one that's been trained on a home care data set. But we recognize that a lot of the problems that that exist in the home care industry are problems that bigger industries, more sophisticated industries have tackled, basically, you know, travel and schedule optimization. These technologies, these algorithms exist, but it's actually in the execution. What makes home care unique? This whole continuity of care makes it very unique. Having the different workforces makes it unique. Being able to have specific schedules where they have to, you know, fit within certain timelines makes it unique. So we have to leverage these technologies, but with a very specific lens for this for this unique application. [00:22:50][82.7]

Michael: [00:22:52] It's absolutely fascinating to me that the idea that you're collecting all this big data. If data is sort of oil 2.0 in the 21st century, that gives AlayaCare a remarkable amount of digital currency as the case may be. [00:23:08][15.7]

Naomi: [00:23:08] Absolutely. And we're really capitalizing on that right now. We've actually just created a new VP Data, and this person is going to be responsible for actually leveraging all the data for all of our platforms. So AlayaCare has made a couple of acquisitions over the over the past year. And so that means we have access to a lot more data than we than we did the year before. But it's important some of these acquisitions or mostly on premise. So the actual installations of the software is on premise. So that means it's not in the cloud. So it's harder to get to. But this data is extremely powerful. So we are having, you know, very specific strategies on how do we collect all the data from our different platforms and kind of put it in one big data lake so that we can leverage, like even bigger, even bigger data sources. So we're actually super excited about that. [00:24:00][52.0]

Michael: [00:24:01] Could you imagine going back in time and telling Florence Nightingale that you hired a chief data officer? [00:24:06][5.2]

Naomi: [00:24:07] Yeah. And that we're we're putting data into a data lake. Yeah. No, they wouldn't find it as pertinent. Definitely. And this is this is still a challenge because we're on the clinical side. Definitely. We have some of our agencies. Things are still on paper. Right. And they want to take advantage of these exciting opportunities. They're like, we love these risk models for rehospitalization. That would be amazing. Like, but everything is still in paper. So we're helping our clients by almost holding their hands and telling them, OK, this is what you need to digitize. Now, let's make sure that everything is captured in the system properly so that in six months when we have that data collected, we can then start to leverage it and build and build systems. But if it's still on paper, we can't help you right yet. So, you know, it depends. We have some that are very sophisticated and some that are right. You know, you feel like haven't gotten that much further than from Florence Nightingale. So so you have to you have to hold their hand along the way. [00:25:03][55.7]

Michael: [00:25:04] Naomi, fascinating stuff. Thank you so much for your time and insight. [00:25:07][2.7]

Naomi: [00:25:08] Oh, you're very welcome. My pleasure. [00:25:09][1.0]

Michael: [00:25:11] So much attention during COVID-19 has been focused on the front line workers in hospitals, yep, those personal support workers visiting the sick and elderly are on the front lines of a new frontier in medical care. And keeping them happy, engaged and on the job is proving to be a hard task by leveraging technologies like smartphones, A.I. and natural language processing. We have a greater chance of providing greater care for those who need it most. [00:25:37][26.0]

Michael: [00:25:43] This has been the CIBC Innovation Banking podcast where we learn the secrets to innovation, economy, success from the entrepreneurs who are paving the way for the future. [00:25:51][8.3]

Michael: [00:25:52] If you haven't already subscribe on Apple Podcasts, rate the show and tell us what you think with a review. I'm Michael Hainsworth. Thanks for listening. [00:25:52][0.0]

[1475.7]