In this interview, we speak to Dr. Amit Shah, an Assistant Professor of Epidemiology with an adjunct appointment in Medicine (Cardiology). Dr. Shah practices general cardiology part-time at the Atlanta VA Medical Center, where he staffs the outpatient clinic and inpatient consult service. He is a Kentucky native, and conducted much of his education and training in the northeast before coming to Emory in 2009. Although much of his research focuses on understanding the cardiovascular pathophysiology of depression and post-traumatic stress disorder, he has a variety of other interests and projects, including: ECG signal processing, cardiac physiology, arrhythmia risk, population-level risk prediction, mobile health technologies, vascular biology, genetics, and risk prediction. Here, we get his thoughts and experiences on wearables and their utility in cardiovascular care.
MA: Tell me a little bit about your background and your research interests?
AS: I’m mainly interested in trying to understand stress as a risk factor for cardiovascular disease. I’m also looking at ways to reduce cardiovascular risk.
In particular though, what I’m interested in is autonomic function - the ability of the body to regulate itself. This is kind of a core metric of health. Things like resting heart rate are a reflection of autonomic nervous system function. Stress and exercise involve the autonomic nervous system. The more tuned up someone’s autonomic nervous system is, the healthier they are. What I like about these wristbands and wearables in general is that we’re getting a very convenient way of doing long term autonomic nervous system monitoring. Then, we’re testing how exercise can improve long term autonomic nervous system performance with interventions like cardiac rehab.
My angle is not just thinking about fitness, prevention, any other behavioral things, but also considering physiology and understanding disease mechanisms. In particular, thinking about exercise, holistically, thinking about all of those things together as ways of reducing someone’s risk profile.
MA: What kind of measurements do you think are good indicators of autonomic nervous response? What wearables have been useful in tracking those indicators?
AS: I think the most basic is the resting heart rate. That in itself is predictive of poor outcomes. People with higher resting heart rates tend to be less healthy. There is a little bit of a finesse when we deal with that, because as people get older and prone to cardiovascular disease, their resting heart rate can change. Additionally, people can be on certain medications that change their heart rate, so it’s not the most reliable metric all by itself. People could have low heart rate and still be very sick. We also look at something called heart rate variability, which is a broad term to characterize how someone’s heart rate changes moment to moment. In particular, or kind of with a little more illustration - thinking about how the heart rate responds to movement, how it responds to breathing, how it responds to hormone changes within the body, how it responds to mental stress. All of these responses give us an indication of someone’s health.
We use well known, published, signal processing metrics to measure heart rate variability. Based on continuous ECG monitoring, we can get a sense of someone’s health. Considering doing it with wristband technologies is a new frontier. The signal to noise ratio is much lower. New wearable technologies have a lot of promise because of how convenient it is to measure such indicators.
MA: Are there any other metrics that you think make sense, any secondary metrics that could also be used for autonomic nervous system monitoring?
AS: Definitely. One of the secondary metrics that relates particularly to exercise is heart rate recovery. Once someone stops exercising, how quickly does the heart rate recover? There are other things that are not necessarily related to wearables. For example, QT interval can reflect autonomic nervous system function. There are some things that you can do with more sophisticated equipment. There are imaging based methods that one can use, as well as methods looking at vascular tone. Thinking about peripheral vasal constriction, as a way to look at sympathetic activation.
One other thing that some wrist bands do is measure sweat levels. So this kind of gets back to like a lie detector test, where if someone is really nervous, they’ll start sweating and set the lie detector off. Some wrist bands do measure that. One of them is the Basis Peak. It measures galvanic skin response, as a way to kind of get at stress.
MA: I guess the other caveat to that is, a lot of these measurements are very individualized, even if you had the data you could ask questions - Is it because of exercise? Or is it because they’re in a room full of strangers and are nervous? Is it because they’ve having internal hormonal changes? You need to have this multimodal interaction to really get an accurate picture of the patient.
AS: Absolutely, it’s a very complicated picture. We get a little bit of help when we consider actigraphy (the science of tracking a person’s activity levels with devices like the Fitbit) , and have that data. With that data we can parse out physical activity and say, “Okay, well, they were being still,” but it still doesn’t give us information on the autonomic effects of eating a steak, or anything that’s not related to mental stress. There are many things that can affect autonomic activity, outside of mental stress, when someone isn’t even moving.
MA: That makes sense. Given that you’re tracking all these indicators, what’s your most favorite variable, and what’s your least favorite variable? And what is desirable for you in a variable?
AS: It will vary depending on the research question. Ideally, what we would have is both high-fidelity movement, and heart rate data at the least, for most types of questions. Of course, the higher the resolution we go with for any of these indicators, we are sacrificing battery life, so it’s really hard to get the perfect combination of things. Something like a Fitbit charge does a pretty good job of getting heart rate fairly frequently. Not enough to ever kind of give industrial strength heart rate variability, or the type of heart rate variability metrics that we would measure from an ECG or Holter monitor, but it’s a fairly decent combination of frequent heart rate metrics, as well as good battery life.
MA: What do you think is limiting current wearables today? To help you measure the variables that you’re thinking about?
AS: I think that, in general, the problem with most wearables is that they are difficult to validate scientifically. Attempts to validate them from various independent studies has shown performance as not being been particularly impressive. There are still limitations and various outliers. There are some cases where wearables are just doomed to fail. Case in point, if I’m walking around and pushing a stroller, it’s not going to pick up my steps. It’s going to be very hard for it to do that. There’s just no way that we’re really going to get accurate steps in that case. For heart rate, it’s just going to be very difficult to get an accurate heart rate, beat to beat, if I’m moving a lot. There will be some fundamental limitations.
I think that a lack of disclosure on what their (the companies) algorithms are is one thing. I don’t blame them, because if they disclosed the formula and it worked, then everyone would try to do it. Lack of disclosure on when it’s valid, when it’s not valid. It would be great to actually know what the findings are on whatever internal tests have been done. Some companies publish validation studies, but I’ve never been particularly impressed with the type of validation work they’ve done. Typically, it’s going to be a small sample size, in a very small healthy group, in a journal that is relatively easy to publish in that potentially lacks a very rigorous kind of peer review process. Something that’s prone to bias, because it’s, let’s say, sponsored by the company.
So, in an ideal world, we would have some sort of device using an algorithm that has been vetted by a good third party study. That may go against the business model of some companies. I’m just kind of speculating why this is pretty much never done. A lot of companies just make claims that, “Oh, we have the best heart rate monitor,” and they have zero valid third party data that substantiates why their particular algorithm is really good.
MA: Do you see initiatives that are seeking to address those roadblocks? Or do you feel like it’s a lost cause with companies?
AS: I haven’t seen any initiatives. I would hope that people are making them. I think that the challenge is that the hardware is developing rapidly, so the algorithms also have to develop rapidly. I imagine that by the time they made the investment to validate a certain algorithm for a certain hardware, it would take a couple years. By that time, the hardware would be relatively obsolete. Anything that’s been really nicely vetted and researched, has typically been a medical grade device. Something like a Philips Actiwatch, which costs an order of magnitude more than your typical Fitbit, Garmin, or other kind of consumer based product. By the time that they make that investment, they have to make it more of like a medical device to justify the costs.
MA: Makes sense. Given what you said about costs, do you see that common wearables will be established through pull from individual patients kind of already bringing these devices in, or will it be from a hospital push saying, “We trust this device, we want you to wear it, and therefore, you should use it?” And ultimately the patient’s get used to the devices and keep using them after that? How do you see, research aside, the traction of these smart wearables being? What do you think will be the main drivers?
AS: I think that for the most part it will be from people who are looking to manage their personal health. For a hospital to get involved it would first have to be baked into the infrastructure. Second, they would have to have some scientific validation for it. Third, there would have to be more of a partnership in terms of just funding it. To really scale such rollouts, it would have to be something that insurance covers, or something like that. This gets back to having something FDA approved for a certain indication and having more evidence.
MA: We spoke a lot about the tracking piece but you also mentioned that there’s another part, which is, “Okay, now I have the data, so what am I going to do with the data?” What do you see as the opportunities around the data that these wearables generate?
AS: I think for people’s personal health, it’s definitely clear that tracking someone’s activity levels in is some way, is enlightening and motivating at least for a short amount of time. These devices give people useful information and can have some palpable impact. For some people, they’ll say, “Oh well, I get the point, I’m sedentary, and I refuse to exercise,” and then they’ll throw away their wearable. Other people will continue to really like the numbers they generate and be motivated for a more sustained period of time.
Heart rate is an interesting measurement. Some devices incorporate heart rate into their caloric calculations. For someone really looking to manage their weight and count calories, heart rate can be a useful. The other idea is looking at resting heart rate and thinking about just your baseline fitness level. I, personally, have noted that on weeks where I’m particularly doing a lot of activity, my resting heart rate would go down, and that’s pretty satisfying to see. In addition to seeing that I’ve done a lot of activity, seeing that I’ve had an impact on my physiology has been really nice. Some people may find it useful to look at their heart rate data moment to moment, and say, “Oh, wow! At five o’clock I was really mad, and my heart rate went up to 140. Gosh, it’s having a really big impact on my health.” More sophisticated or educated users may get some benefit from looking at more granular data.
There is now a line of wellness related to meditation and biofeedback, based on heart rate variability, that products like the Apple Watch and Fitbit are starting to utilize. It’s kind of neat. I’ve been playing around with it a little bit on the Apple Watch, and it’s kind of nice. I haven’t been doing it for too long, but I think stress reduction could be a potential use.
MA: I think so too. As part of my job, I get asked this question a lot, “Oh, we want our patients to use a connective blood pressure cuff, or connected weighing scale, or a Fitbit”. I wonder, is it actually worth the technical expertise that you sometimes need because these devices aren’t seamless in the way they connect with phones yet. Each of them follows different protocols, each of them has different issues. Despite that, some physicians are starting to feel like they need that data. I’m curious, as a physician, why would that be needed, as opposed to self-reporting of information? What are your thoughts around that?
AS: I think that any type of diary keeping can be helpful to both the patient and the clinician. Back in the “old days,” we used to ask them to write their measurements down. Measurements like weight and blood pressure would be written in a journal and brought to the clinic session. Personally, I would find it helpful to see how things are at home, how things had changed over, say, a holiday break. You can relate certain events to trends in their blood pressure. Sometimes, blood pressure at the clinic is not particularly helpful because of the white coat effect. It is nice to get some sort of diary from home and these Bluetooth connected devices just make it more convenient. They add a little sexiness, along with not to write everything down. People can show their flashy app on their phone, and be like, “Oh, look!”, you can scroll through and email their doctors what they have. Wearables add a level of convenience, perhaps, but that central thread of just having a diary to reflect upon, I think that is the most important thing.
MA: That makes sense. You mentioned in the beginning of this interview that another service would be behavioral modification. How do you see behavioral modification happening on top of the data that is generated from wearables?
AS: I think the main thing is just getting feedback. How healthy or unhealthy a patient is. Then, with that additional insight, someone realizing, “Oh, wow! I’m sedentary.” Or, “Oh wow! My heart rates really fast. I need to get my act together.” I think any kind of feedback on a modifiable health factor can be useful for behavioral change. The other thing is that some wearables monitor sleep. If someone is able to quantify that they’re not sleeping much, maybe that can motivate them to sleep more.
MA: So the trend seems to be kind of closing that feedback loop, based off the data that the wearable generates? That’s pretty cool. The question is, you’re doing work on the autonomic response, how do you see therapies being built around controlling that response? What do you feel is the next generation of therapies that can be used, utilizing all these technological wonders that we have today?
AS: I think, in general, I tend to classify therapies kind of as like a pharmacologic, or non-pharmacologic. From a pharmacologic perspective, there are some studies looking at, for example, drugs like beta blockers. How do they improve autonomic nervous system function in the long term. One possibility, is let’s say someone has high blood pressure, and their autonomic nervous system is particularly a problem. We would preferentially prescribe them a beta blocker if their heart rate variability is particularly bad, as opposed to, let’s say like a diuretic, which is what we normally do.
The other thing that is more of home run, is really just to push wellness interventions related to exercise and stress reduction, probably as two of the main things. The other thing is smoking cessation, if it’s applicable, and weight loss, which kind of goes along with everything else. Certain diets, such as the Mediterranean Diet, can improve autonomic nervous system function. Or at least there’s a relationship there, between diet and autonomic nervous system functions. In general, healthy lifestyle, holistic interventions, are probably going to be the way to go.
MA: Definitely a lot to think about! What else do you think people need to know?
AS: We didn’t really get into wearables themselves, which wearables there are. I think, just a brief thing to say with the disclaimer that we really don’t know what the validity of a lot of these wearables are, especially with new ones coming out like every year, every other year, kind of like in the way that we’re release new iPhones, or new Samsung Gears. One thing I’ve noticed is that there has been a recent surge of very cheap wearables. My head was spinning looking at how many of them there are, and how similar they all look. But they’re these generic wearables, that pretty much try to do exactly what the Fitbit Charge does, and on paper, have similar specs. I haven’t tried any of them yet, but I’m tempted to because at 30-40 bucks, they are definitely much easier to scale up if they actually offer a generally similar level of information. It doesn’t have to be 100% perfect on steps. I think as long as it consistently measures steps it could be worth something.
I’m curious to see how the field will change, now that these really ultra-cheap versions are coming out. Also, there are versions that claim to measure blood pressure, and I’m curious to see where the future lays with some of these things, and whether it’s going to lead to lawsuits, or whether there’s going to be some substantiation in the literature. Most recently I’ve seen blood pressure and pulse oximetry, which are neat additions. At least, they claim to measure those. So, we’ll see.
MA: Do you see any concerns around, “Oh, these are like commercial grade devices, and there’s a danger of people trusting them too much, as opposed to going by whatever the FDA has regulated and things that’s much more proven?”
AS: Yes. People may have false alarms of too low pulse oximetry or too high of blood pressure, leading to the over-utilization of health resources. The inverse being the under-utilization of medical resources should the devices give them false reassurance with regards to blood pressure. There could definitely be a concern, and I think that people should just think of them more as kind of having more of an entertainment value, as opposed to a real diagnostic value, since they haven’t been cleared. As long as it’s read with a grain of salt, I think it’s fine to at least try them out.
MA: I hope you get to try them out as well, we’re seeing that they’re becoming more commonplace.
AS: Yes, we’ll see how it goes.