"Medical records must become a living record that pulls in data real-time, follows your health, and displays it back to a physician in a useful form."

Q. Dr. Snyder, can you tell us about your role at Stanford?

Dr. Snyder: I am Professor and Chair of Genetics at Stanford University’s School of Medicine. I also run the Center for Genomics and Personalized Medicine, the innovation lab there, and we do a lot with big data in health.

Q. Grace, can you brief us on your role?

Grace: Paddy, I lead Solutions Development for the AWS Disaster Response Program, which focuses on how technology and the cloud can assist organizations that are active across crisis such as the COVID-19 pandemic or across the lifecycle of natural disasters.

Q. Amazon has recently been in the news for several healthcare-related initiatives, and one of them was the Diagnostic Development Initiative. It was targeting COVID-19. Grace, what was the gap or need that AWS was looking to address with this initiative?

Grace: AWS Diagnostic Development Initiative is a 20-million-dollar commitment that we made last year to support customers and accelerate their diagnostic innovations. We provided this support in the form of both cloud computing credits and technical support from AWS experts like our solution architects and our AWS professional services team. These AWS experts helped those organizations that were part of the Diagnostic Development Initiative to use AWS services to either stand up or scale their COVID-19 diagnostics projects. As COVID began rapidly spreading around the world, there were a few reasons diagnostics really bubbled up to the top. First, accurate detection is the tip of the spear for any effective pandemic response strategy. Secondly, diagnostics research has historically been underfunded and largely de-prioritized when compared to vaccine or treatment development. But realistically and thirdly, organizations working on diagnostics also need access to reliable and scalable compute power, which AWS could deliver along with things like analytics and machine learning to help researchers process and analyze some of those large datasets that were being generated and iterate more quickly. So, in the first year of the program, we have been excited and seen some inspirational results from customers like the wearables work that Dr. Synder’s research team is doing at Stanford. It’s really been great to see how these projects are pushing the boundaries of diagnostic innovation. 

Q. Dr. Snyder, Stanford has been one of the early participants in the program. And you have launched a smartwatch-based diagnostic app for the COVID alerts. Can you tell us about the app and some of the results that you have seen so far?

Dr. Snyder: Several years before we found that you could tell when people are getting ill from Lyme disease, as well as respiratory viral infections using a smartwatch. When the pandemic came along a little over a year ago, with Amazon’s help we have really been able to scale this thing up. So, we first showed that we could detect COVID with a smartwatch. It turns out, on average, four days before symptoms and for some people, as much as 10 days before symptoms, we can see when they are getting ill because their heart rate jumps up on a smartwatch. So, we first showed you could do that and recently, we have rolled out this app that alerts people when their heart rate jumps up, which does happen before you get ill with COVID or other things. It can happen with other lifestyle events as well, for example, you drink way too much. But it certainly seems to work for infectious disease about 73% of the time, according to our latest work. It is a simple app that you download on your smartphone that integrates with the smartwatch. It works for Fitbit; Apple Watch and we are trying to work it for other watches as well. Basically, they are following your heart rate, will transfer the information over to the phone. We use the cloud to pull any information, and then we compute using our algorithm. When we see a jump up in heart rate or other abnormalities, it will send off a signal which pushes back to your smartphone and it’ll set these alerts. Right now, we have just launched the second phase of the study where we are sending the alerts. As I said, it sometimes picks infections, and it does pick up COVID infections as well as asymptomatic cases. We think this is going to be very powerful. It absolutely requires the cloud for this to work because you need to be able to access people all around the world. The study is global, and you can compute everywhere. You keep the costs down actually by running some of the computing areas that are less busy and then distribute the load, so to speak, in the more cost-effective fashion. That’s probably the only way you could do a project like this that uses the cloud and it’s totally scalable. 50 million people in the U.S. wear a smartwatch and right now they could all have an alerting function for COVID-19 if they tuned in to this program. 

Q. This is a global program. So, I imagine that your application has been downloaded globally by people onto the smartwatches – Apple or Fitbit. You mentioned that the elevated heart rate could be a result of various potential activities and not necessarily just COVID. I imagine that the algorithm in some way adjusts for different likely causes and then combines it with other kinds of wearables and so on. What has been the predictive power of these algorithms and how do they hold up across populations or across regions?

Dr. Snyder: We’re going to need more data to answer the last question because the numbers are so small. That’s why we want to have more people join the study. We’ve had several thousand people signed up. We’ve had something like 70 positive cases so far. So, we’ve picked up seventy-three percent of them from different parts of the country, and we’re still improving the algorithms. We want to get that Seventy-three percent up to ninety-five percent or better. We can do that as we pull in more different data types focused on resting heart rate steps and sleep. We pull in different kinds of data; we can improve the algorithms, so we are trying to get as many people signed up as possible. We can detect COVID from different ethnic groups. I’m optimistic it should work for everyone because when people get sick, their heart rate jumps up.

Q. If you do get a million people signed up, what’s the end goal here?

Dr. Snyder: My end goal is to put a smartwatch on everyone on the planet, seven billion people, so they have a health monitor for every single person. That cannot happen today, but that is the long run. The only way to do it is to be following your health in real time, not doing PCR two days later when they get symptoms. You want to be following people while they are healthy in real time, seeing when you see and detect an abnormality and catch and push it back to them as quickly as possible so they can act on it. In the case of a pandemic, if they get one of these alerts, we want it to be as sensitive as possible and as specific as possible, we want them to ultimately self-isolate or get checked right away, before they spread it around to one hundred other people. 

Q. Grace, one of the outcomes of these programs is that you are going to get a lot of data about patients, about consumers and so on. Do you have any plans to harness insights from this data in any way, let’s say, for public health in this case?

Grace: No, that is not the programs intent. AWS is vigilant about our customer’s privacy and data security. Our technology and program policies are really designed with that security and privacy in mind. So, for customers like Dr. Snyder at Stanford and others retain ownership and control of any data and content that they store on AWS, along with the ability to encrypt it, protect it, move it or delete it in alignment with their security policies.

Q. Dr. Snyder, how do the wearable device integrate with Epic or Cerner and make it a part of the longitudinal patient record for diagnostic and treatment on an ongoing basis?

Dr. Snyder: Right now, we’re in the research phase and testing these algorithms, seeing how well it works, and optimizing them. You would have to have a follow up test for that to go into in the record. That’s where we stand now, but in the future, these things will get better validated and they’ll have to get FDA approved, which is not hard to do for simple devices like thermometers. And that will be the case for smartwatches. I think they will be able to get validated and you’ll be able to pull information from them and aspects of that will be in the medical record. Now, my own view is the whole medical record needs to change. Right now, it is not useful to most doctors. It’s hard for them to access information from the record. I’d like to see the record become a living record, meaning it pulls in your data in real time, follows your health, and then can displays it back to a physician in a very useful form in which they can see how is your cardiovascular health, how is your metabolic health, how is your other forms of health? So, I think we should transform the whole medical records system to make it in a useful fashion. An example of this is when they measure your heart rate in a doctor’s office, it’s all over the map and it depends on whether you drove by bike there, what stress is going on, all sorts of things. But you can pull a pretty accurate heart rate right first thing in the morning from someone and get a much better picture of their health. Imagine incorporating that kind of information into a health record for a physician to be able to see what is called a longitudinal record so they can really follow what is going on.

Q. Dr. Snyder, what about patient privacy? In the research phases I guess you are putting some guardrails on what happens to the data, how are you protecting patient data and so on. Can you talk about that?

Dr. Snyder: That is a big concern. So as Grace said, we encrypt everything as it comes. It gets encrypted as we compute it, and we compute encrypted data. As these alerts go out, they get pushed back so that everything is stored. One thing that is important is we do try and pull the data and share it in an anonymized fashion and Amazon has been fantastic for helping. People use the term data lake, but I want to make it a data ocean where we have all these data for people to be able to access again in an anonymous fashion so that we can improve our algorithms and be able to detect disease much better. I think this kind of platform is going to be powerful well beyond the pandemic, meaning you can pull other kinds of information from your smartwatch. You can pull other kinds of health measurements from a smartwatch like dehydration. So, by having data that is accessible, researchers can improve this health monitoring system, I think we can really transform the way people’s health is followed. So, I like to think healthcare instead of sick care, so we can then follow people and better manage their health.

Q. AWS is offering millions in credits to developers worldwide as a part of this program. What has been the response so far? What kind of research projects are we likely to see with a focus on healthcare?

Grace: In the first year of the Diagnostic Development Initiative, we supported around eighty-seven organizations in 70 countries. The organizations included customers that are startups, non-profits, research organizations and businesses. We provided cloud computing credits and technical support to really work backwards from the needs of these researchers to understand how technology could help accelerate or scale their work. In addition to the work that Dr. Snyder’s team has been doing around wearables at Stanford, we’ve also seen organizations focusing on looking at uncovering clues about how COVID-19 presents in individuals and what are some of the impacts or what are some of the outcomes that they’re seeing based on characteristics of their immune response networks been done by the Institute for Systems Biology. Our biology team uses machine learning to try to quantify the silent spread of COVID-19 for those with symptoms. Organizations look at using smartphone cameras to provide accurate and reliable diagnostics within 30 minutes of doing a test. One of the things we are doing this year is broadening the scope of the Diagnostic Development Initiative to cover not just diagnostics but also three new areas. First, early disease detection to help identify outbreaks and trends at both the individual and the community level. Also, prognosis to better understand disease trajectory. And then last for public health genomics to bolster genome sequencing worldwide, which is becoming more important as different variants of COVID-19 emerge.