As a visiting professor in the Computer Science Department and the Internet of Things and People (IOTAP) research center at Malmö University, I am pleased to be able to talk about the potential for expanding the research we are doing in IOTAP more broadly into the smart health area. I’ll begin by discussing what I mean by ‘smart health’ and then address some specific areas of interest and risks that must be addressed.
Evolution of ICT in Healthcare
Smart health is along the evolution of how we’re applying computing technologies to the healthcare field. First, we had e-health. We could use computers to keep track of medical records and share them from place to place. Then we had m-health, when we could capture or access medical information on mobile devices. Today I’d say we’re in the era of smart health, because now we can have devices that have built-in intelligence. Now our supporting health systems can do things — can “think” in a way.
What is Smart Health?
When I think of smart health, I do think of providing some kind of health related services. That could be from the hospital or healthcare provider perspective, of course, but my own focus tend to be from the patient perspective — from the user perspective. But all of those are included in in the smart health area. Smart health systems provide health related services using a network — some kind of connection between intelligent agents. These intelligent agents could be computing devices, mobile phones, sensors, Fitbit smart bands, surgical devices, devices that measure your blood chemistry, or devices that measure your brainwaves. Any of these things could be intelligent agents.The human actors — patients or healthcare providers for example — could be intelligent agents in this system. The sensors, devices, computers, applications, and human actors are all intelligent agents that might be connected in the smart health system.
Role of IoT in Smart Health
Access to data
With IoT, the Internet of Things, we have access to a lot more data than we did without IoT-kinds of networks and ecosystems. And that kind of data could be health related. Something I read about recently was that Google’s research arm, Google X, developed a contact lens that can measure your glucose level. They’ve now partnered with Novartis who are going to have this in the market, they say, within five years. So you could have a contact lens that measures your glucose level and transmits that to a mobile device, to your healthcare provider, so you know that you need to give yourself an insulin shot or whatever. And it’s much nicer, for the patient, just to have a contact lens — that you might have to wear anyway — measuring that instead of sticking yourself in the finger ten times a day. So you can see from a patient perspective what an innovation this could be. So, access to data: things that we found it difficult to measure before — things that we couldn’t capture directly — become available to us via our IoT ecosystem.
Connecting multiple agents
Connections to many different devices. That’s what IoT is all about.
And having devices that can decide on their own what action to take are things that we now can see in these kinds of networks. And I’m not talking about having robots take over all of our healthcare, but just having an insulin pump that knows when it’s time to give us a higher dose, or when it’s time to give us a lower dose, based on those measures coming in, perhaps, from our contact lens. These kinds of things can improve our healthcare overall.
IoT Smart Health Opportunities
The kinds of things we can get from IoT:
We’ve seen that we can measure a patient’s heart rate or their blood pressure remotely. The doctor, or a nurse or another healthcare provider, can keep track of that and take action if necessary.
Self-management of chronic conditions
We’ve seen for a while now that we can help patients with chronic conditions, like diabetes or high blood pressure, measure their condition and manage that over time — to see if activities they’re taking are improving their condition, modifying their activities, and so on.
Athletes have used these kind of smart health systems for a long time to try to improve performance. In addition to athletes, we also can see performance measurement in rehabilitation services. So sometimes it’s not just “can I run faster” or “can I have my heart rate come back to to my resting rate faster?” but it might just be “can I dress myself?” or “can I get enough movement to feed myself?”
Behavior modification is another thing we do with these systems. If we want to stop smoking, to start exercising more frequently, to reduce our stress level, or want to get more sleep — all these kinds of things can be supported by these devices. And many of us do that on a regular basis.
Detection & diagnosis
Areas we haven’t seen as often are in detection and diagnosis. We know that doctors might use artificial intelligence systems to help them with a diagnosis, but there now has been a reported case of Fitbit data being used to help diagnose a patient and determine the best treatment. It was somebody who had epileptic seizures; he was taken to the emergency room, and they were trying to figure out the appropriate treatment. Which treatment was best was going to be determined by what the heart rate was over a long period of time. Somebody noticed the patient was wearing a Fitbit and they wondered if they could see this data, and they could. So they could see from the patient’s Fitbit data where his heart rate had been so they could see when the problem started — with the seizure — and how long it lasted. And based on that, they decided which treatment was appropriate. And they ended up leaving the Fitbit on during the surgery, and all the way through, so they could again monitor where was the heart rate and did it go back to where it should be.
These devices that are sold, and you’ll notice they say very clearly “not a medical device”, because the manufacturers don’t want people to think they’re accurate enough to be medical devices, but we’re starting to see a need for this — a use for this.
An example not as complicated as the epilepsy one: How often do older people go to the emergency room or the doctor and say “oh, I feel so dizzy.” Well, a doctor never knows what to do with that because there are so many things that could cause it. But if you could just look at one bit of data, like the blood pressure, you can see “ah, the blood pressure was really high — that could make you feel dizzy!” If you could get that data from somebody’s device, you’d have a pretty good idea of what was causing the dizziness. But dizziness could also be caused by very low blood pressure, and the treatment is going to be different depending on which it is. So the blood pressure data from the patient’s own device could be accurate enough to point the healthcare provider in the right direction. This is a very simple example of how we could use this data, even if it’s not perfectly accurate, in the diagnosis process. I think we’re going to see more and more of that, even as device manufacturers keep saying, ”Not a medical device” because they don’t want to be sued. I think we will see more physicians and healthcare providers saying “Just let me see if that information is useful to me.” So we’ll see more in detection and diagnosis, and I think eventually into treatment as well.
Challenges to Utilizing IoT for Smart Health
Platform & data heterogeneity
One issue, and some people say this is the number one problem of expanding the connected health system, is that we have so many different platforms and so many different types of data that it’s very difficult to integrate them. In the U.S. just with the electronic health records, we have at least five major vendors of electronic health record systems that are being used — and they don’t share data. They have different standards; somewhat different ways of storing things. If we can’t share that data, we can’t pool it and we can’t do really sophisticated research on it. So these different data platforms are a problem for trying to do some of that big data analytics kind of thinking with our healthcare data.
Data integrity & accuracy
We also have problems with data accuracy. We know that the measures we’re getting from our personal wearable devices tend not to be perfectly accurate. Studies have compared Fitbit and other devices with “reality”, with actual things, and found they’re not always right, for example when someone has a stride that’s different than the norm: somebody takes itty-bitty steps or great big steps, not normal-sized steps. Or somebody who’s older than the norm. So it’s not always always as accurate. And that’s something we need to work on, especially if that data is going to be used in medical situations.
Privacy & Security
Healthcare data is ten times more valuable to thieves than your financial data. It’s more persistent and you can get a lot more money for it on the marketplace. So if you’re thinking about selling any of your data, go with your healthcare data rather than your banking data. Because that’s what people really want.
Because that data is so valuable, people are trying to get it all the time. A group of white hat hackers were hired by the Mayo Clinic to look at their healthcare system — it’s one of the big research-based healthcare systems in the U.S. They hired these white hat hackers to come in and look at all their devices and see how safe they were. What they found was that one hundred percent of the devices had had some type of intrusion. So it wasn’t a few. It was all the devices. Another group later on looked at many different hospitals and found, again, a hundred percent of the hospitals had had some kind of intrusion into their network. So hackers often try to get in through users — we know they’re the weak link — into the network and then get to these devices, and they get to these devices because they are not well protected.
They found that very often the password used on these devices were the ones installed by the vendor. And sometimes the passwords couldn’t even be changed, which seems incredible. So, once somebody knew the password for that model, they could use it anywhere they found that device. Luckily they didn’t find any instance of hackers trying to go into one of these devices and kill somebody — which you definitely could. Instead they were coming into these devices that were not well protected, and then getting onto the network and going after that valuable patient data — and pulling it back out through these devices.
Anything that’s connected to the Internet can become a potential portal for hackers to get in and get data out. So, these medical devices that now we’re saying are going to be part of our Internet of Things healthcare ecosystems are a real risk for us. The vendors often have been notified about the problems, and they don’t really want to solve it. They think “well, let the users deal with it at their location — let their network protection be better.” I don’t think that’s a great solution, but there’s not a whole lot we can do about it. It’s something to be aware of as we research and develop these systems.
Related to privacy is the idea of trust, because some patients are aware of these security issues. Others just in general don’t trust any network to keep their data safe. So, when we ask patients to share their healthcare data — with a larger system or with us doing research — a lot of times they don’t want to because they know or they suspect it’s not secure. And they don’t want it to be shared with the world. We need to get over that issue because we need patients to be willing to share their data so that we can build these systems to help make them healthier, and also to learn what works so we can overall develop better healthcare solutions. It’s an issue that those who develop these kinds of systems have to think about: how we make sure people trust us, because they should trust us, because our networks are safe and secure.
Basic IoT Smart Health System Functions
To think about how we could study issues in smart health, I tried to identify the basic operations that a smart health system does. For most smart health systems that we see — from a consumer perspective — there’s something we’re measuring that’s going to be compared to a baseline (a target, a goal, a norm). If you’re trying to be more active, and you use Fitbit, it has as a default that “ten thousand steps” would be your goal for a day. Is that a good goal for everybody? For some of you, ten thousand steps is nothing. You need something way higher. For other people ten thousand steps might be too much. Having these goals set by the vendor isn’t always ideal. It’s better if the patient can have some interaction with that, and on Fitbit you can modify that. It might also be a good idea to have your healthcare provider involved in that for some situations to say “this is appropriate for you” or “this isn’t.”
The sensor collects data about that thing that we’re measuring. It stores and reports the results and then usually provide some kind of feedback. Whether it’s telling you “good job,” giving you a little trophy, comparing you to other people, so you know how you’re doing. Different systems do different things.
Better IoT Smart Health System Functions
To me this is a starting point, but where I see we can do interesting things is looking at how we can personalize these aspects of it. So personalized goal setting; personalizing the kind of feedback that’s coming out of these systems. Deciding what the system should do, and what people should do — what kind of things should be done automatically by the system, what kind of things are left best to the user; where should the user be involved. We’ve heard earlier today that really the advancement is going towards more and more automation, where the system is doing more rather than having the person do more. And these two do conflict a little bit: personalization is almost the opposite of saying “just let the system do everything.”
Consumer Smart Health: User-Centered Design
Setting meaningful goals
Studies have found that people are more likely to reach a goal if it’s something that they’ve been involved in setting and they’ve said to themselves or to others why that goal is important. You know, if I just say “I want to walk twenty thousand steps every day,” I’m probably not going to do it. But if I say “I want to be healthier, I want to reduce my blood pressure so I can live longer and do more research,” then it becomes meaningful and I’m more likely to do it. So for any kind of goal we’re asking somebody to set, if they can do it in a way that has them reflect on why they want to do it, why it means something to them, they’re going to be more successful. Also we tend to set these numeric goals that are relatively short term when maybe a better goal setting method would be to think long term: Where is it we want to end up? What kind of person do we want to be? There’s a lot we can do in goal setting to increase the way these systems interact with the user and allow us to set more meaningful goals.
Customizing feedback & coaching
People respond differently to coaching. Who likes it when their system tells them “good job”? Some people like that, they feel better about themselves and they work harder. Then there are others (and I might put myself in that category) who when they get that kind of “good job” feedback, hear it sarcastically. So maybe it’s not motivating to those users. We can’t build these systems the same for everybody; we need to find ways to customize them.
Aligning multiple goals — holistic perspective
Also we tend to think of these things as helping us achieve one goal. Who has one goal? Usually we have many goals. Even with our health we might have many goals. We have other goals for our inter-personal relationships; we have other goals for our travel. Wouldn’t it be nice if they could all come together somehow? And we could find ways to achieve many of our goals by doing one activity?
Recruiting and retaining appropriate design study participants
Then from the researcher perspective: When we want to study these things, I think we need help in figuring out who should we have in our studies and how can we get them to participate. Because if you just ask for volunteers to participate in your study, are you going to get a broad sampling of the population? No, you’re going to get the people who feel important when they participate in studies (or those with particularly strong interest in the topic). I think there is work we can do in how do we recruit people to be subjects in our studies. And how do we keep them engaged and involved? Those are some research areas I find really interesting.
Some specific areas where we are either starting to work or I see potential for collaborations are in the areas of bio-sensors, device and application development — particularly in the area of identifying optimal design choices and determining appropriate behavior modification methods — and in the reporting and visualization of data.
In summary, this is a rich and complex research area that we are well-positioned to study. I would be happy to discuss this in more detail with anyone who is interested. Please feel free to contact me at nancy.russo(at)mah.se.