After 18 years in the healthcare industry In March of last year Satish Movva founded CarePredict. The company’s first product, Tempo, is wrist-worn sensor designed to track daily life patterns in the elderly to help determine small changes in their activities that could hint at bigger issues. Tempo detects motion – walking, running, sitting, standing, or lying down – and location; and the software figures out the associated activities. It transmits data and charges wirelessly so there are no cords or plugs. Element14 recently had an opportunity to discuss CarePredict and the technology behind Tempo with Mr. Movva
Element14: Why did you start CarePredict?
Movva: It’s very simple. I am part of what is called a "sandwich" generation. I'm almost 50 years old. I have very young children I'm taking care of. But I also have very elderly parents. My dad's 86, and my mom's 76. I visit them once a week, and every time I visit them, I find that there's something different about them. There's something that's changed. My dad is now shuffling his feet. He's walking slower. Or my mom, I find out when I visited her on Sunday that she's been bedridden for the last two or three days with acute pain in her legs because she's a diabetic and has neuropathy. Well whether it's cultural or what I don't know, they choose not to bother me with their health information.
E14: I don't think that's cultural, I think its universal--simply parents thinking that their kids have got enough on their plate and they're not going to be a burden to them.
Movva: Yeah, that's it right there. They think they don't want to be a bother. That's one aspect, another aspect of it is we're all busy - I'm busy with my children, I'm busy with work - and it's shameful for me to say it, but our parents come towards the end of the list of things we have to think about every day. So I think they partly recognize they're at that age so they weren't calling, they weren't telling me, and when I see them I'm like, ‘But, Dad, why didn't you call me, so I could take you to the doctor on Wednesday when you started shuffling, instead of finding out on Friday and trying to get you to an ER or somewhere else because you're retaining water. That's why you're shuffling.’ Your ankles are swollen. So if I can get them to a doctor, or have the home care nurse visit, or something happened, they can change the medication regimen, put him on the diuretic, get rid of the water, so he doesn't end up in the hospital with cardiac arrest.
So that's where, really, the inspiration came from. I said, 'If there is a way to monitor my parents, continuously, and figure out these little small changes, I can then intervene myself, or I can trigger a clinical intervention by calling a physician or a nurse.’
E14: So how did you go about finding a solution?
Movva: I found that there plenty of monitoring technologies--sensors, carpet switches, toilet switches, refrigerator door switches - stuff that will give an alert when something shakes them, or they move, and from that you can tell there is some sort of activity taking place. But guess what? That entire model falls apart the minute you have two people in the house - because you don't know who is causing the activity. It could be my mom is lying down on the bed with neuropathy in her legs and my dad could be moving around. But I would have no way of knowing because the sensors would only report activity.
E14: The system would not be able to distinguish between who is doing what?
Movva: Correct. So I said, 'There's got to be a better way.' and I started giving it some thought. I figured out what I thought was a neat way of fixing this problem and then, we filed patents in May for the technology, and we are off to the races. Basically my insight was, first that we could make a wearable sensor and we put it on the person and second if I put a FitBit on my mom, and know when she is active and she's not active, that's wonderful but it gives me no context. And by that I mean, you know the Fitbit will tell you that my mom took ten steps yesterday, in ten seconds, and she took 20 steps today, or ten steps today in 20 seconds. But I can't draw a conclusion, because I don't know if the steps were the same; was she going up the stairs, down the stairs, was this up hill, downhill? I can't draw a conclusion - it tells me nothing.
My insight was that if I could have context of this in terms of where in the home they are, while the activities are occurring, then I can draw a conclusion. Our system would be able to say, ‘Mom walked from the bedroom to the bathroom in 10 seconds yesterday. She walked from the bedroom to the bathroom in 20 seconds today.’ Now I can draw a conclusion, saying, ‘Okay, it's the same steps. She's taking longer. She is taking slower steps. She's walking slower.’ That was a key insight into the formation of our company and part of our intellectual property - understanding where in the home they are.
Also, in the universe of 65 year old women the data points say that the normal person should go to the bathroom two times a night. Well that's not really valid because there are so many confounding variables. They could be diabetic. They could have all kinds of other diagnosis that affect how many times they go to the bathroom. So, it didn't make any sense to look at the universe of people within that age on a population basis and draw conclusions.
E14: Were you able to anticipate what technical problems you might have in developing the system that you did, or did things sort of come along so that you solved problems as they emerged?
Movva: We had a lot of discussion right at the beginning, because I wanted to file a non-provisional patent, I needed to solve some problems, at least to not have any fundamental issues. Basically I needed to make what goes in the patent was exactly going to work and was correct and was not subject to change. I was able to think all that through, and then subsequently we solved problems as they came up. But the fundamental vision of the technologies we were going to use that are enshrined in our patent application are still valid and operating today, so that part I got right, right at the onset. That's what we did. I spent the rest of the year trying to refine the business model and building prototypes and whatnot.
E14: The one thing you don't want to do is have unnecessary false positives and disturb a lot of people who might be monitoring what's going on.
Movva: You got it, absolutely because that might happen if you just used benchmarks based on population metrics. So the other key insights we came up with was, ‘Let's just make the damn thing machine learning.’ I'll strap this thing onto my mom - it's a wrist-worn sensor - and let it monitor her for the first seven days of its life on my mom's arm, let it monitor her behavior, so it understands what the baseline is. That was the key insight that nobody's done before.
E!4: I see. They have to wear it, as you said, for a certain period of time so that the system determines what their normal characteristics are. Do they get a message that it's okay, that the trial period is over, or does that just happen automatically?
Movva: Today, it's just automatic. At the end of seven days it's finished. Essentially, it's passive mode of machine learning. From then on it circles automatically into an active mode. Now you can always reset, so let's say the person had a significant change in condition. Maybe it's a diabetic and they had a foot amputation or something happened. When they come back from the hospital, you can reset it to go back into learning mode again, so it can learn the person's new level of activity. Now, if you don't do that, as variations occur, it will use that to influence the baseline, but not drastically. So, the same variation has to appear consistently over a period of time for it to shape the baseline and say this is the new baseline.
After the end of seven days, our system pretty much knows what my mom's patterns of activities are - her daily rhythm. It knows on average my mom wakes up at 7:30 in the morning and then it knows, on average, my mom goes to the bathroom for ten minutes. Then, she goes to the kitchen for about 20 minutes, and then she sits in the living room for two hours. The system knows that's the average pattern for any given particular day.
Now, if my mom wakes up at 10:30 in the morning, like usual, goes to the bathroom, but then comes back and lies down on the bed, it knows it's unusual. It sends me a text message, and gives a very soft message. It says, ‘Hey, Mom might not be feeling well. Why don't you give her a call to see how things are.’ Similarly, if on average Mom goes to the bathroom two times a night but last night she went seven times to the bathroom then it sends me a text message saying, ‘Mom might not be feeling well. Give her a call.’ Again, a soft message. It's not going to tell me she's gone to the bathroom, because that's would raise privacy issues.
E14 What about messages to healthcare professionals?
Movva: The clinical aspect works really well, so if I was selling this to a healthcare system, as we are, the care manager would get an alert saying, ‘Mr. K went into the bathroom seven times last night, potential urinary tract infection, give him a call.’ So the Care Manager can say ‘Hey, what's going on Mr. K? You went to the bathroom about seven times yesterday evening. Are you having pain? Urinary tract infection? Let's bring you in for a urine culture. Let's figure out what's going on.’ The care manager for Mr. K is part of the care providers that surround him and there are no HIPPA issues surrounding it because it’s being used by the registered care provider.
E14: Along those lines You are aiming now at consumers, but if someone was running, for example, an elderly care facility and didn't want to have to dispatch someone to go room to room to check on what's going on on a regular basis, but still be able - from a central location - to monitor the people who are being cared for in a facility, it would seem to me that you system would work there as well.
Movva: Absolutely. In fact, that's very insightful of you because that was originally a market we had not anticipated. Now, since about February of this year we've been deluged to an overwhelming extent from group living facilities calling us and saying, ‘You cracked the problem for us. We need this, like right now because now we can manage the activity patterns and flows of our residents.’
Today in group living facilities they don't have any way other than by observations by humans about the patterns of activities of their residents. When they come down, when they wake up. When they come down to the breakfast room. Do they come down to the breakfast room? At the breakfast room are they actually sitting down and eating socially? Are they just taking their plates and going back up to their rooms and being antisocial? Which is one of the indicators of depression and things like that. So they can figure out tremendous amount of information about the residents so they can provide better services for them. So they know what their activity patterns are. Without a lot of human effort they can figure out what's going on with their residents. They cannot let things slip through the cracks.
E14: I'd like to spend a couple of minutes now, if you don't mind, discussing how the CarePredict Tempo product does what it does. Obviously, the wrist-worn unit has got a number of sensors on it, to determine motion and position and things like that?
Movva: Tempo is worn on the wrist of the dominant arm, that's a basic requirement. The device itself has got a MEMS package, magnetometers, gyroscopes, that kind of stuff, which basically give us activity, body orientation in 3D space, angular rate of moment of the arm, things like that, which allow us to classify activities of daily living.
E14: Is there more than one sensor supplier? Are you willing to discuss who the sensor supplier is?
Movva: I can tell you that. I don't know if it makes any difference, but InvenSense is the company that we are using. The reason we chose InvenSense is they're really the only ones that have all three sensors fused onto one piece of silicon and can give you fused data from all three sensors.
They are a well-known manufacturer - they're in Apple, they're in Samsung. That's why picked them and their packaging is very small. I think their current packaging is three millimeter by three millimeter by one millimeter. They offer an I2C bus I mean there are really cool things in that product we really liked it. So that's why we picked that. It's got an additional ARM processor in there that can do classification of the motion and other stuff on-board.
Then the other thing is the room location stuff. We are using line-of-sight to figure out which room we are in. So instead of using RF--Bluetooth and things like that-- to act as beacons, which can seep through walls and whatnot, and give you false readings, we are using a line-of-sight model, so we know what room the person is in, without going into too much detail.
E14 Let me make sure I understand. You are not using Bluetooth, but, rather, line of sight which I take to mean IR.
Movva: We are not using RF. Let’s just say we use line of sight. There are room beacons that you have to place, peel-and-stick on the wall of each room of the house. One beacon per room or area, like a foyer, a hall, or whatever. As long as the Tempo device is within that room, it can detect the nature of that room, the type of that room, whether it's a bedroom, a bathroom, or a kitchen, a foyer, or whatever.
E14: What about power consumption?
Movva: We have close to 30 days of battery power per charge and we are trying to get to a 60 to 90 day charge on it. So, it’s really low power type stuff. We're doing a lot of low power algorithms and what not. And we have another third device in this is the whole series which is the communications hub, a very simple short range 3G cellular gateway essentially. So, basically the Tempo device, as the person is walking around the house, when they come near the hub, it uploads it's data to the hub, and the hub just shoots it over to the cloud, where all the predictive, analytic software, and bench marking, and all that goes on.
E14: Okay. If you're using 3G from the hub, then there's got to be a supplier to that and there's a cost factor involved. Does that get passed over to the user?
Movva: That's included in our monthly fee.
E14: Oh, okay. What if the user said, "My house is already wired for internet access. Is there a way that you can connect with a router that I might already have?" Is that possible too?
Movva: It is because our communication hub has an Ethernet board and we are considering putting a WiFi client access point in there. But, you've got to remember that the demographic we are going after are 65 plus seniors. We want to make this as simple as possible. If my mom and dad can't install this on their own, it's not shipping. So that's basically the red line I have in my mind and the way I thought about it is yes it'll save me a few dollars a month by using their own built-in internet in their homes, but the cost of putting a user interface on the hub so they can program the SSID, all that was just too much and the support costs on the back end will negate any cost reduction you might foresee from not going cellular.
E14: I understand that completely. But the assumption behind that is elderly people are going to be buying and installing this, as opposed to their children buying and installing it for them. What do you anticipate will be the percentage breakdown?
Movva: We are marketing, on the consumer side, to the adult children of seniors - people like me who want to take care of their parents. While we are marketing to them, we do expect that when the device shows up at the home, not all the adult children are in the same town, or neighborhood to come over and install it, so it's more than likely their friends would, or they would themselves. We are designing it for somebody who is 65 or older to install this themselves. Basically, when they get the kit, it's already been preconfigured, they have to plug in the communications hub and they have to wear the wearable sensor. They have to peel and stick these sensors - these beacons, which are little tiny things, on the wall of each room of the house. And then they walk around each of the rooms. The thing vibrates to tell them it's reading everything and it's receiving everything. That's it.
E14: Has your experience so far with the system presented any surprises or has it gone pretty much as you anticipated?
Movva: Well so far we are only in limited user trials and so far things have gone as expected. The one insight that we got was that with the comm hub and this device we're essentially created a footprint, a platform, in the home. So we should consider putting in some additional sensor packages and things like that. You know like put some temperature sensors and barometer sensors, things like that because these could be valuable data points in the future to understand data about seniors living in their home.
Keep in mind that we’re adding about 10,000 seniors a day turning 65. All those baby-boomers, as they are aging and heading into our healthcare stream, need to be taken care of and cared for, but we don't have enough caregivers in the country to do that, to accomplish that. The only way we can do that is by using machines to monitor them, with predictive analysis and whatnot, and raise alarms when they need attention, so that the human caregivers who are timeshared across so many different people, can know to pay attention, fix the problem, and move on to the next person that needs attention.
All the highly developed nations with aging populations are struggling with the same issue, they are the ones that are most interested in our platform and are calling us most consistently, from the Scandinavian countries, Germany, UK, South Korea, Japan, South America. These are the countries that we're getting tremendous interest from because they have this aging population and not enough care givers to provide care.
E!4: What's the status of availability of the product? When will you be in full production? When will your distribution be more mature than it is now? What's the timeline?
Movva: So today we are still in limited user trials, beta level stuff. We will have production shipping by the end of September. Enough to where people can expect to get the first batch of our devices.
E14: Thank you very much. I appreciate your time. It sounds like you're going to have a quite successful product and system here.
Movva: Thank you for the opportunity to speak to you and share our story.