Wearable Tech May Detect COVID-19 Infection Before Symptoms

; Abraham Verghese, MD


June 29, 2020

Find the latest COVID-19 news and guidance in Medscape's Coronavirus Resource Center.

This transcript has been edited for clarity.

Abraham Verghese, MD: Welcome, everyone, to this episode of Medicine and the Machine. I'm Abraham Verghese from Stanford University, and it's my great pleasure to co-host this conversation with my friend and colleague, Eric Topol, from Scripps Research Institute. We will be talking about something that has been a great interest of Eric's, well before COVID, as we're seeing a wonderful convergence of wearable technologies and artificial intelligence (AI) that can predict disease in a way that is particularly meaningful during this COVID-19 epidemic. Eric, welcome to this episode. As always, it's nice to be doing this with you.

Eric J. Topol, MD: It's always a pleasure, Abraham. We have a great topic, because we're trying to use technology to assist with this whole COVID-19 affair. We're struggling, no question. Everyone talks about testing, but we know that those tests have issues. One, of course, is the false negatives with the virus antigen test, and there are also logistical issues of doing them at scale. At some point we're hoping for home testing, but we're not there yet; this is at least a few months away. Also, there is the expense; it is a one-off test, so there's a time limitation.

It's interesting that the first thing people turned to was temperature checks. I don't know about at Stanford, but here at Scripps Research, if you want to go to the campus or you're working in the labs, you have to have a temperature check. But that's so silly because, as it turns out, multiple prospective studies about fever and COVID-19 have found that large numbers of people don't have a fever. The big one, published in Nature Medicine, found that about 30% of people had a fever, and in the study that was just published by Color genomics, the rate was 12%. So, many people don't have fever, if you look at it prospectively, and they can be presymptomatic and asymptomatic. We know that 30%-40% of patients with COVID-19 are asymptomatic. And the people who are presymptomatic shed virus and are at least as infectious, if not more, as those with symptoms. So this whole notion of temperature checks is like a placebo.

Verghese: I quite agree with you. We learned too late that we didn't emphasize masks enough and we overemphasized temperature measurements.

I think it's worth reflecting how different this is in terms of where we are in COVID-19 science compared with the HIV era, where we operated for a couple of years not even knowing the cause of the disease and for many more years without any effective therapy, and also the 1918 influenza pandemic. With COVID-19, we sequenced the genome of this virus within weeks. I'm astounded by the science; the only way I can keep up with it, Eric, is by following your tweets, which are at the cutting edge of everything that's coming out.

I wanted to ask you about the wearables, because following on the temperature story, it's clear that this is coming much too late. There's an explosion of wearables that preceded COVID-19, but now, big research consortiums are turning their attention to them and how they can alert us to people being ill before they actually know they're ill.

Topol: Right. The science is moving at a pace that I've never seen — everything, from the structural biology of the virus and the antibodies to the virus from patients, to the design of drugs and vaccines and neutralizing antibodies. The sequence of tracing it temporally and spatially geographically throughout the world has been extraordinary.

But what we're stuck with is, how can we find people in a cluster or an emerging outbreak before it spreads more? Because we know, by the famous Pareto rule or principle, that 80% of transmission comes from 20% of the cases. So if we could find these spreading events early and concentrate on them — because we can't test everybody every day or every week — that would be a good way to get our arms around it. The United States and a few other countries are so far behind. Here is the opportunity to use sensors that get continuous data and would give us an edge. We started one project in March and we now have about 38,000 participants.

There are others that use ring technology. Ours uses a smartwatch or fitness band. It's called DETECT. What we already have is interesting. In the first 30,000 people, we saw a signal correlating not just with symptoms, but positive tests, based on an increase in resting heart rate, more sleep, and fewer steps. Those three sensor measurements were a signal. We had done work previously with flu-like illness, showing that same sort of signal that we can detect rapidly. But folks in Germany read our paper in Lancet Digital Health in January and they developed their own smartwatch app. They have over 500,000 citizens using it. Similarly, China now has 1.3 million people using their app.

We know that these outbreaks will keep occurring. As we saw in Beijing, they went 55 days with not a single case, and now, suddenly, hundreds of cases. So it's imperative that we come up with a continuous surveillance signal. Unfortunately, people are conflating this surveillance tool — a digital tool tracking things like heart rate or body temperature — with the contact tracing apps, which are quite different, haven't been validated in any way, and are very controversial.

Verghese: Yes. The scale of this is unimaginable — the fact that you could have so many thousands of people in a study in such a short time. It begs two questions: Has this been vetted with the same academic rigor we bring to other trials? And do we get the signal early enough to make a difference in some way?

Topol: Right. We have to validate that. We have only done that for flu-like illness in about 50,000 people using a Fitbit. We saw the signal well before the CDC had detected a flu-like illness. This is a different matter, but you know what? In many ways it's even more conducive to this technology because so many people are asymptomatic. Let's say that at least one third never have symptoms all the way through their infection. But according to different series of lung CT scans in these asymptomatic people, from the Diamond Princess ship and from Korea, it's striking to see that more than half have lung abnormalities, ground glass opacity, and consolidations that are the same as in symptomatic people with COVID-19. So they are having internal hits churning in their bodies. In that case, you would expect things like the heart rate to help pick that up. During that presymptomatic phase, we know that, again, the resting heart rate goes up before any fever is manifest, and a lot of people never have fever, as I already mentioned. So we think the sensors will be able to pick that up.

In the United States, more than 100 million people have a wrist sensor for heart rate. It's widely out there, it doesn't cost anything to wear it, and most people already are wearing it. But, as you say, it has to be fully validated for this specific virus. We hope to find ways to detect infection beyond testing, because testing will always be a one-off. Even when we have a home test, it will cost at least $10, $20, or more each time you do it. And it's valid only for that day. We need something that has a lot more temporal bandwidth so we can see what's going on in people. This measure is not so helpful at the individual level; if your heart rate goes up, you still don't know why. But if COVID-19 is in your neighborhood, if there is a cluster, then that makes it more of a real signal.

Verghese: As someone who uses one of these devices, can I tell my own status, or do I have to wait for the research consortium or the meta-gatherer of all this information to let me know that I'm about to come down with something?

Topol: When you have a surveillance tool like this, you get a signal. Right now, what we would do is offer testing, tracing, isolation. So it is a gateway to precision testing because something may be going on. You don't know for sure. It could be a false-positive signal. In the future, perhaps later this year, we could get home tests to those people who have a signal so they can find out quickly, because hopefully those results would be ready in a matter of minutes.

It's just the beginning. We need to get more precise testing and then leave it to local public health departments, the contact tracers, to activate. We're in the early stages in the United States. We have a limited number of human contact tracers, which is kind of ironic since we have such high unemployment. Why don't we recruit?

This whole digital effort is kind of like the Rodney Dangerfield of science: We get no respect. When I worked with the Rockefeller Foundation National Action Plan, they had a panel of 20-plus experts, like the economist Paul Romer and all of these Nobel laureates. All of the sessions were about how many tests we should do each day: One million, 5 million, 10 million in the United States? And I was like the carnival barker saying, "Well, what about the digital?" Finally in the report, digital was recommended, but it doesn't get the regard it deserves.

The issue is to get people to be citizen scientists. Turns out that a lot of people like to get their data and like to get a notification that something in their neighborhood is showing a potential signal, without inducing anxiety. But I'd like to at least raise awareness. You don't need everyone in the country to be a citizen scientist; you just need enough. We have every state covered but not densely enough yet, so that will be important.

Verghese: Two thoughts come to mind as we talk about this. One is the issue of equity and access. This has been such an earth-shaking moment in our country's history [with the widespread Black Lives Matter protests]. Ever since the first slave ships landed here in the 1600s [there has been structural racism in the country]; now we're beginning to come to terms with our inequalities. With the mass deployment of these devices, how do we ensure that we truly are studying a representative cross-section of this country and that everyone has equal access to what is basically a public health measure? On the flip side is the notion of privacy and our resistance to having Big Brother watching over us too much. How do we go forward with those different issues?

Topol: Those are critical points. On the one hand, the United States has a terrible situation with inequities, and these have been underscored by the pandemic. Already before it came here, we were the only developed country in the world that wasn't providing healthcare to all citizens. Now that the pandemic invaded, we've seen the toll it has taken, and it is profound. The people who have suffered the most are those who are underrepresented minorities and the indigent. It's a horrendous situation.

With respect to the ability to do tracking, fortunately our project and others like it are dedicated to this, and the data will never be used for any other purposes; it's part of the informed consent. But also, it's okay if people don't have a smartphone or a smartwatch or a fitness band, because this is just like the traffic map you look at on Google or Waze. If enough smartphones are sending signals in the area, you don't need a "chirp" from every car to know that there's a big traffic jam or that the traffic is running smoothly. We just need enough to get some coverage. The people who don't have this technology will still derive the benefit of knowing that there's an outbreak potential in their area.

The problem is that people in these underrepresented minorities and of lower socioeconomic status don't have access to testing. They aren't looked after. Many of them are afraid to come in because they could be deported, or who knows what could happen to them. We have a lot of collateral damage from the pandemic here because of our tenuous and, in many ways, pathetic framework of healthcare.

Verghese: It's truly been a story of great scientific advances and success, with miserable communication and leadership. A strange human paradox, if you will.

Technology in the Face of Change

I want to switch gears to something you've been tremendously interested in for some time now, and that is AI. I want to remind our listeners that an AI company, BlueDot in Toronto, was the first to spot the unusual cluster of pneumonia cases in Wuhan. AI has gone on to have incredible applications here — not just epidemiologically but in terms of molecular targeting. I'm learning a lot by reading your Twitter feeds. Talk about how AI has been brought to bear on this problem.

Topol: Analyzing the data, getting in the tens of thousands, if not millions, of people's continuous data, requires deep neural networks to analyze that data. That's one use of AI. As you mentioned, the company BlueDot was onto this well before the US government. They had so many layers of data that they were continuously analyzing in China, not knowing that there would be a pandemic. But if something was going off the track there, they had a baseline and they could see signals through many different means, including things like hospital parking lots starting to change and all sorts of indirect digital indicators that, when you added these different points together, you knew something was definitely wrong. The United States wasn't a part of that.

In California, our governor, Gavin Newsom, was in touch with BlueDot. One of the reasons it was the first state to go into shelter-at-home status was because of that early warning. But that's just one way of knowing where a pandemic is dawning.

We also are seeing several drug development programs being catalyzed by AI, whether by mining lots of data that already exist from SARS-CoV-1 or other coronaviruses and MERS, or simply the structure of the virus or antibodies with the epitopes. All of these are leading to much more accelerated candidate drugs, candidate vaccines, and neutralizing antibodies.

What do you think about combining human expert oversight and the machine to predict who should be in the hospital vs who should stay home? This is an interesting dilemma. There's at least one company, called Current Health, that has a device that measures multiple physiologic parameters. You insert your arm and it measures everything except blood pressure — heart rate, oxygen level, body temperature, and respiratory rate. It gets the measurements through sensors and can transmit that information to the medical facility. The company claims that several health systems are using it to keep people at home, which is a good idea, and keep them from having to come in to the hospital. It provides continuous monitoring with algorithms to analyze that data, with an alert if something is starting to go off track. But it hasn't been validated. What do you think about that?

Verghese: I've been humbled by one of the early AI studies that I'm sure you will recall, that showed that we were miserable about predicting 30-day mortality in our patients, that we were really terrible about prognosticating and recommending chemotherapy because we had no good sense of how much longer a patient had to live. The machine clearly could do a lot better than us. So I'm in favor of this. With any data like this, we need to have continuous clinical input and monitoring and to make the data available for validation by other groups rather than being blindly applied, which can result in amplification of all kinds of injustices and mistakes.

We're talking about AI in the face of the great dearth of the kinds of social intelligence, the kinds of administrative intelligence, that are required for a crisis like this. To politicize this virus in quite this way is extraordinary. On the other hand, I've been rereading a lot of the plague literature — Camus' The Plague, the Decameron, and any number of plague narratives — and we have to understand that this is very much the nature of human beings. We have a cognitive intelligence side and an emotional side.

In many ways, that is the problem: wedding science with humans' abilities to undermine all good science and all good public health information. I wonder if we could get AI to figure out how to help us with the sort of moral, social choices that are undermining everything that medicine can do.

Topol: That is a challenge, because that's the part that we as humans should be doing much better at and we're not. Whether that support could be bolstered through algorithmic means remains to be seen. I have been struck by the complexity that isn't adequately appreciated . For example, you mentioned not respecting masks in the beginning. And still, 1 out of 2 Americans are not wearing a mask, which is shocking. We can't even get that straight. What is the big deal? Any sort of face covering will help reduce the spread. So, we don't have a social contract.

It's like a vaccine that only works if everyone is taking it. We have an anti-mask movement like the anti-vax movement. I don't believe any algorithms will solve that one.

Then we also have these crazy changes in manifestations. This was thought to be a lung disease, and then we realized, oh, you know what, actually this virus likes ACE2 wherever it is, and it will attack the heart and simulate a heart attack. It will attack the kidneys even on an isolated basis, or the brain. Anywhere there's ACE2, in fact — even new-onset diabetes because the pancreatic islet cells have rich ACE2 expression. So the complexity hasn't been fully appreciated.

The other thing of great concern is chronicity. So many people get out of the hospital, perhaps after a long intubation, and they are well enough to go home. But they have chronic problems, chronic fatigue, high difficulty breathing. We don't understand or even know the natural history of that. Patients ask, "Is this ever going to go away?" It's really a concern that we don't understand this.

There's also this peculiar immune secondary wave we see — for example, in children who have a Kawasaki-like disease. It could be after they seem to have recovered, but they get a pan-body, multisystem, inflammatory organ involvement. I just think we haven't yet appreciated the full hit. We have tried to simplify it, and that hasn't gotten us too far.

The asymptomatic state is another one that's like the iceberg below the surface. The science community was in denial. They said, "Oh, well, it can't be 30% or 40%; it's just not possible. We've never seen a virus that has 30% or 40% asymptomatic cases but then can also be a killer." But it is easily 30%.

That denialism shouldn't be the norm when you are simply seeing something that's never been seen before, but we tend to get entrenched in our priors. Every day we're learning new things about this horrendous virus and its toll in so many different dimensions.

A Cohesive Effort

Verghese: The good news in the midst of so much misery is that we've seen a jumpstart of science. We've seen a jumpstart of AI. We've seen a tremendous boost to wearables and this new application, and big consortiums coming together. Google has bought Fitbit and so on. I believe that is the silver lining of this very, very dark cloud. The challenge remains to communicate the elegance of the science, the pure solutions, to the public in a way that is not politicized but penetrates and gets through. That's on us in science to find clearer ways to communicate.

I'd like to think that people will put their health before anything else, although right now it doesn't quite seem that way. As you look ahead to the frontier of digital technologies — wearables, AI — what are some things that you're looking to? Where do you see light at the end of this tunnel?

Topol: Before going into the positive things, I want to mention that we've had a lot of hope for contact tracing, but we don't have enough people doing this. It's an arduous task to contact people and tell them they've been exposed and to go forward and backward tracing contacts. It's quite laborious. The idea was that we would have apps that would be picking up smartphone signals and alerting people. But we don't know if that's going to work. It's been tested in Singapore, Norway, and other countries, but there hasn't been enough participation. You need to get 60%-80% of the population to use it. Plus, there are issues about privacy, ethical issues, false positives, and all sorts of other barriers. So it hasn't been validated. It may never fly, and that remains to be resolved.

On the positive front, I think we will get through this. It's going to be a long haul, though. There probably will be a vaccine. What's fascinating about this virus, as evil as it is, is that it doesn't evolve much. It sits tight. Mutations are innocent, as best as we can tell, and they're slow. So the structure is quite amenable to a vaccine. There are many vaccine trials, even one now that's going into phase 3. It is possible that we'll see a vaccine in the early part of next year, which would be beyond any record in humankind, to have developed an effective and safe vaccine in such a short time. It probably will require at least yearly repeat injections, perhaps even more frequently.

And then the question we have is, can we override the anti-vaccine issues in this country, which are unlike anything in the rest of the world? We have a very strong anti-vax movement through social media. We have to get through it. We have to vaccinate 70% or 80% of the population to begin to get artificial herd immunity, if you will, because we're sure never going to get it naturally. We're at a 4% infection rate now, and look how many deaths and other issues we've seen already. We'll get out of it; the vaccine is going to be the ultimate way.

But before that, dexamethasone looks like it helps people on a ventilator, remdesivir helps some people recover more quickly, and we'll see other drugs that will help to modulate different phases of this illness. And neutralizing antibodies are now in clinical trials. Temporally, that's like having a vaccine because this is multistructural, taking down the virus in many different epitopes. I'm optimistic that we'll see the fatality rate drop considerably over the months ahead. Add in the vaccine, and then someday we'll be able to test at home, have the results on your smartphone with a bar code, and when you go into work, you just show your bar code and you're good to go. Testing is essential to getting back toward normal in the months ahead.

I'm optimistic, but we have to be patient. Unfortunately, the states that have been opening prematurely with no testing, tracing, isolation, or infrastructure — and even at times opening when their cases were not trending downward at all — these are bad experiments. I understand that people don't want to be locked down and we want to get the economy back. But we have to be patient, to go the distance. I still haven't given up hope. Just look at how much we're learning every day about better treatments and better ways to mitigate.

Verghese: That's wonderful. I am reminded of Camus' The Plague, where he says, even with as many plagues and pestilence as there have been in history, that everybody's always surprised when the plague arrives. It seems that we have to relearn the lessons with every plague. No one ever seems to carry the wisdom from the previous plague much further.

But I've heard some great ideas today, Eric. I especially agree with the idea of a federal program to employ more people who are unemployed right now as contact tracers. That would be a very valuable function to combine with testing. Otherwise, testing by itself does very little. I'm excited by the optimism I'm hearing in terms of the science and the way it develops. Thank you for taking the time to go over all of these things with us. I'd love to get another update in a few weeks, because that's how fast this thing is moving. I'm sure we'll have new things to talk about in a very short time.

Topol: Thanks, Abraham. It's always great to have a conversation with you. It's great that you brought out how the science is marching along at high velocity. That's something we're good at. We weren't so good at preventing the pandemic from landing in this country, but the science community has come together worldwide like never before, and even switching into COVID gear from other areas, whether it's life science, social science, or whatever. That cohesive solidarity within the science community is terrific. I hope that will be enduring, not just through this pandemic but for a long time thereafter.

Verghese: Thanks to our audience for joining us for this episode of Medicine and the Machine. We look forward to your reactions and comments, and I look forward to chatting with Eric, my co-host, again very soon.

Eric J. Topol, MD, is one of the top 10 most cited researchers in medicine and frequently writes about technology in healthcare, including in his latest book, Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again.

Abraham Verghese, MD, is a critically acclaimed best-selling author and a physician with an international reputation for his focus on healing in an era when technology often overwhelms the human side of medicine.

Follow Medscape on Facebook, Twitter, Instagram, and YouTube


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.