YouTubeFacebookLinkedInTwitterinstagramrss



Harnessing the power of Big Data and AI for faster, better, and more efficient studies

Click Here to Return to Download Page

Allison Proffitt:

There's a new data source that can help us understand and engage clinical trial participants. Data collected passively through wearables, devices, apps and sensors. Imagine being able to gather last minute data directly from patient monitored adherence to a protocol outside of brick and mortar walls and enroll patients into a study within twenty-four hours. All of these scenarios are possible in the digital era of medicine. Welcome to this clinical informatics news podcast. I'm Allison Proffitt, editor of Clinical Informatics News. Today we'll explore new methods of integrating data from wearables, sensors and apps into a study to develop novel endpoints, predict disease onset and progression, and optimize study recruitment and protocol compliance through digital channels. Helping us dig into this are Jessie Juusola and Luca Foschini, both of Evidation Health. Welcome!

Jessie Juusola:

Hi there!

Dave Johnson:

Hello!

Allison Proffitt

So let's start with introductions. Tell me who you are and what your role is at Evidation.

Jessie Juusola:

My name is Jessie Juusola. I have a PhD in industrial engineering with associates on health policy modeling. Prior to Evidation, I worked in healthcare consulting as well as diagnostics and now at Evidation, I lead our health outcomes research team. At Evidation, what we're really doing is focusing on quantifying health outcomes through digital means and really the intersection of behavioral data with health outcomes data. We have a study platform where we run digital studies and it's really kind of about that direct connection to the study participant and the individual and all the different data types that you can gather and the new ways you can view patients and the new things you can learn about them via digital means. And we'll kinda get into more of that throughout the course of this podcast.

Dave Johnson:

Thank you. This is Luca. I'm a co-founder and Chief Data Scientist at Evidation that takes care of anything that has to do with data. So, data collection, the studies that Jessie was mentioning, what is the device, the app, the data source, really that we want to target for measuring a specific outcome. That is something that my team works together with Jessie's team and our pharmaceutical clients to define what source we should look at. Once the data is collected, it needs to be cleaned and normalized and at that point, it's ready for inferences. So we're trying to understand for this new data screens that comes from what a large is defined the internet of things to understand if there are changes in health outcomes that are measurable. So changes that relate to disease status and progression. So we take care of doing those analysis.

In the spare time, like Jessie was mentioning, we try also to publish a lot of the research that we do and specifically with published work on infrastructure to collect data from internet of things devices and apps at scale and in general research about behavioral psychology to improve uptake and compliance on studies and in general research about the national field of digital epidemiology.

Allison Proffitt:

So how do you run health outcome studies using data from wearables and sensors and apps?

Jessie Juusola:

We run our studies mainly purely virtually. Sometimes they're a bit hybrid where it would start out with an in-person interaction either at a clinic site or sometimes at some other sort of facility site or we've done home visits in the past, etc. But really the crux of it is the virtual participation of the participant. We have a study platform where we can screen, consent, enroll people purely electronically. We then can collect data through lots of different means, some digital, some non-digital. And Luca can speak more to that. But really, it's about bringing the clinical trial into the tech age and so it's no longer that you only can participate in studies via this clinic site and these academic centers that are doing all this research. It's really making research a lot more accessible and the types of data that we are collecting become a much broader set and much more rich data sets.

We have a platform called Achievement, where anyone can kind of go sign up and participate in their health outcomes and wellness and, as part of that, one of the opportunities people get is to participate in studies. So our study platform is kind of built on top of this consumer community that anyone can participate in and then have the option, as well, to then be in the study. Luca, do you want to speak more to the data types?

Dave Johnson:

Definitely. So like Jessie was mentioning, the new data that we're collecting is really just trying to compliment what already is being done and used in medicine for centuries. The only problem with data collected before the internet, before connected device, is that it was very sporadic. So people would have an option to give a data point only when they had an interaction with point-of-care. And that will be maybe between weeks between any two consecutive interactions. Now, with connected devises and apps, we have the chance to really have a continuous feed of data down to the second or even sub-second resolution and that allows us to make inferences about trends and changes much more quickly and accurately than we could do before.

Allison Proffitt:

So you recently announced a large-scale pain study. Can you tell me a little bit more about that?

Jessie Juusola:

Absolutely! That's our DiSCover Project. Which stands for digital signals and chronic pain. And what we're doing is purely virtual study targeting enrollment of about 10,000 people. Both people who suffer from chronic pain and those who don't. We are collecting various types of data from them but really with the end goal being able to develop some digital signals for things like chronic pain severity, flare-ups potentially of pain, quality of life, etc. So this will be an observational study that people can enroll in and they'll be kind of different ways that they'll participate in the study. There will be some surveys, different sub-groups will be asked to participate in some different procedures as well such as voice collection. Everyone will be contributing sleep and activity data from trackers and wearables that they're using. And really we'll be tracking these people over the course of a year collecting some self-reported data as well. And then really kind of bringing this rich research data set out of it.

Allison Proffitt:

So you said this project is going to go over the course of a year. Where does it stand now? Where are you in the enrollment process?

Jessie Juusola:

Now we just recently launched this. We are going to be turning up the speed on recruitment soon and we're targeting to enroll in about a couple month's period, which is very fast kind of in the clinical world. Most people think of studies as taking years to enroll and thousands of subjects or hundreds of subjects and we're talking about 10,000 subjects in a couple of months.

Allison Proffitt:

So why do you think you can do that? What is it about this virtual study or this approach that's going to make that possible?

Jessie Juusola:

It's really about that ability to connect directly with the participants. You no longer are limited by site set-up. You know, you're not limited by getting clinical sites on board and getting everyone trained and going through all their various site-based IRB's. Now you really kind of have one core, central experience and you can reach anyone; in this case, anyone in the U.S. So you really just have the ability to set up and ramp up very quickly. You also have a ton of access to more representative populations. You can imagine with traditional clinical studies, you're getting a certain subgroup of people who are in a clinic setting who see the types of physicians who are running research, etc. Here, everyone's online, pretty much everyone these days has mobile access, has a smart phone, or has a computer or something and so we can access really any of those people and so it really allows you to recruit very quickly. And recruit very representative populations.

Allison Proffitt:

So my second how are you going to do this question. You listed a ton of data that you plan on collecting from a bunch of different sources and Luca said down to the second data collection, which is just going to give you an incredible volume. How are you going to manage that?

Luca Foschini:

Yeah. This is one of the challenges and the good thing about Evidation is that we're a tech company that's helping health and in improving clinical processes so really big data and automating decision making that comes from AI permeates everything we do. So the data collection of down to the second resolution streams of data that comes from a thousand patients might seem an overwhelming task in health care and in the clinical world, and in fact, it is in general, but in the tech world, that's not even considered big data to some extent. So there are infrastructure that have been built over the years internally at Evidation that allows us to do the data collection seamlessly and actually ensure the level of [inaudible 00:08:55] is low so ensure compliance during the study.

The next problem after yo have all this data collected is actually how you analyze it. What kind of question do you ask about the data? Here, really, is where AI can show it's power. Simple ideas are tracked to identify factors that underpin [inaudible 00:09:18] so can you, somehow, predict some timing events when an acute pain event is going to happen and how do you aggregate and integrate all these data sources that come from context that's collected maybe through apps and the data sources that are collected through different devices to really understand if there's a change in severity at the patient level.

Allison Proffitt:

You mentioned digital biomarkers and I think Jessie talked about recording people's voices, which I'm assuming might be connected. I think that's fascinating. Can you tell me how the technology is going to discover these? And what you're going to do with them?

Luca Foschini:

Yeah, sure! A digital biomarker is a signal that tells us something about disease state or progression. Biomarkers historically have been hard to discover because you need to collect the data from which you can mine it but I think it's even harder at this point in time when there is so much data coming from different streams and really you can go through it all let's say in just like cycle through all the possible variables that you can compute and see if any of those have the relationship with a change in the outcome. So here is really where AI can help. It can help peer into patterns more effectively and only surfacing the ones that are related to a change in outcomes of your interest and you can see those patterns as digital biomarkers. Digital biomarkers in the era of internet of things will be more related to changes in time. So ability to detect events or short-term changes in outcome as much as they will be about predicting trajectory of chronic conditions. So, progressions versus stability.

Allison Proffitt:

So what do you hope to find in this study? What's going to be the definition of success for this study?

Jessie Juusola:

It's really around the concept that Luca spoke about and being able to find a digital signal in the data and really develop some sort of digital biomarker for something like chronic pain severity, flare-up, quality of life. Hopefully there's many coming out of that but at least something will be a success. And I think even at a more basic level, even just proving that you can run this really large-scale study, that you do enroll these 10,000 people quickly and collect rich, meaningful data sets from them. I would say even from the starting point that that is success and then all of the interesting research findings that come out of it and the potential that those give us for kind of down the road validating research and those sorts of things will kind of be cherries on top.

Allison Proffitt:

So, let's take a step back. We've talked about this particular pain study, but how do you use digital tools to manage protocol compliance in any of the studies that you're running?

Luca Foschini:

You can use AI in a very loose sense. AI just means automatic decision making in a way. And so you can think about all the tools that already exist out there to manage compliance, to reach out to participants when the data is not being provided and make them be powered by intelligent systems. So, when you have a continuous data feed from your participants, for instance coming from a wearable device, and you're able to detect that they're missing that so the person most likely is not charging or is not wearing the device in this specific example, then you can have an automatic system that can incentivize the person to go back to collect data in a variety of ways that's more than just maybe sending a notification or a text message or really alerting the clinical specialist that none of these work.

Jessie Juusola:

I think there is even at a more basic level just the fact that there is data coming in and you know what is going on in this person's life on a daily basis. Even before you get into the opportunities that are available that Luca was talking about, you now know if someone is complying with the protocol. If you think about kind of more traditional studies, you don't necessarily know what's going on at home. You typically don't know what's going on at home. You have this snapshot into when a person comes into their clinic but that maybe they don't even come to that clinic visit and they can tell you what they were doing at home, but here now we have data coming in that is a much more objective indicator of what is going on at home.

So, there's then a lot of ways that you can use that. So you can do some of the more manual outreach even potentially. You can, as Luca was speaking to, develop some algorithms that will do some more high-tech types of incentivization and outreach. But really, it gives you a lot more insight into what is going on with a person who is participating in a protocol and allows you to be able to better manage that and really get that rich data set in the end where you know you have the right data at the right time points from all the different people in the study.

Allison Proffitt:

What other areas can these new technologies improve? Other than just speeding recruitment and having a better grasp on protocol compliance?

Jessie Juusola:

The one thing I spoke about earlier, is more representative populations. I think that's a really important thing. There's been a lot of areas in medicine where we know the samples have been very biased in a lot of ways. In cardiovascular, for example, it's been a lot of white men. That's not necessarily the population that's affected by cardiovascular issues. And so I think one big thing is being able to really access the people who are impacted by all these various different issues of health care. Of which, as we know, there are many. I think additionally, there's a lot of cost efficiencies that can come out of this.

Part of the speeding up recruitment is what I spoke about with not needing to set up all these sites around the country and as we know, R & D budgets and clinical study budgets are huge. And a lot of that is a lot of the groundwork and leg work that goes into these brick and mortar settings. If you can really have these more centralized, virtual experiences, everything just becomes a lot more efficient. You have data coming in through all these different means, but most of them very cost efficient. And so it just really brings the efficiency of research into a whole new realm.

Luca Foschini:

In addition to that, one benefit really just comes from the sheer amount of data that you get longitudinally. When you have so much frequency, so much [inaudible 00:15:10] data over time, then you're able to understand your individual variability and your individual changes of outcomes much more accurately, which means that you're able to detect changes on an individual basis versus their own baseline much more accurately, which means, in turn, that you can have smaller or shorter studies because you're able to detect signals earlier and with smaller cohorts.

Allison Proffitt:

I know that lots of groups are wanting to use devises like Fitbits or Apple watches in studies. What advice do you have for them as they try to make that work?

Jessie Juusola:

I would say we have a lot advice, but I think a big part of it is work with the right partners. There's a lot of intricate issues that are at play here. There's a lot of nuances. So work with someone who knows what they're doing and has done this before. I think that can make it a much more fruitful and valuable experience. Also, I would say go in to these types of studies with a somewhat focused research question. It doesn't have to be incredibly focused, but at least in terms of kind of what data streams are valuable that you want to collect. You don't want to try to boil the ocean. Yes, you have a lot of thing available to you, but you also need to think about, again, keeping this efficient, keeping the participant burden manageable. You don't want to ask people to be wearing ten different things at a time, for example.

So I think that just being very thoughtful about what you are collecting and putting thought into the design so that you are collecting a set of data streams that you believe are very meaningful and then you will have a very large, rich data set to work with. But you won't be trying to boil the ocean.

Luca Foschini:

Yeah, absolutely. I couldn't agree more. I think that the really important, key point there is go in the study with some sort of idea of a context of use. What is the specific question that you have, the specific population you want to target and what are the implications of your findings really that you want to see in the long term? That will inform you what kind of data you want to collect and, at that point, you realize that really the devil is in the details. Heart rate, many devices collect heart rate, but heart rate at what frequency? Heart rate at what accuracy?

Move away from blanket statements of this device is not validated or this is not right or this is much better than the other one. Really it all applies to a specific context of views that you have at hand. And the other part of it is, like Jessie was mentioning, the patient, the participant. Like really try to build the experience of the study around the participant that you have. So think about what their daily life is going to look like. Are they likely to charge a device three times a day or do they need something that they only need to charge once a week or maybe even never. Just, you know, set it and forget it, kind of the device. Those are the things that we run into more often that they actually become a very big problem if you understand them too late in the study design because at that point you can't back off.

Allison Proffitt:

Jessie, you mentioned that companies, pharma companies, should be finding the right partners. Can you tell me what they should be looking for? What you think they should be looking for when they're looking for a partner to run a digital study?

Jessie Juusola:

Yeah, absolutely. So I think that when we're looking at studies, you need to be running these in the right way. There are research best practices that you want to make sure are followed so I don't think it's any partner that can do this. You want to work with a partner who's really credible. Who does robust research and is doing this in the right way. And, in large part, based on the participant experience, that they're designing it with that in view, that they are following best research practices, etc. And also, you know, someone who's able to look at it from the viewpoint of let's really crystallize the question you're trying to ask and put that context around it and understand all the different variables at play.

You know, again, I think that a lot of people are very excited about digital health right now, as they should be, and a lot of people think, "Oh, I could just go put a Fitbit on anyone." But the reality is, there's a lot of different factors at play. Some of those, which Luca has touched on. Some of those just being the infrastructure that's really necessary to run a study well.

You need to get the right data from the right people at the right time in order to really have a robust data set to work with. And so there has to be some amount of structure and infrastructure and capability and experience in place there in order to do that well. You need to have that direct line, that direct connection to the participant. You need to have the experience with IRBs. You need to have the ability to monitor things like protocol compliance, data quality and make sure that that data set that you are getting is really high quality there.

Luca Foschini:

Yeah, I think that the hardest part is really to find a partner that understands both pieces of the equation; the tech piece and the clinical piece. And in the specific case of real world studies where really the device that you're going to use, the apps that you're going to use, are going to be given to participants that will be outside your view, your control. The two parts are really intimately entangled. So, people that do the clinical part and run the study, they might have to deal with malfunctioning devices and might have to understand something about the API behind the device. At the same time, the people that do the technology and the data analysis really need to understand, again, what we were mentioning before, what are the daily routines of the participants and how those interact with the clinical practice of the study itself.

So, it's really hard to outsource one of the two. You can be a clinical person and then outsource the tech and you can be a tech person and outsource the clinical. And I might be biased here, but I think that at Evidation, we really have started to ... like the inception of Evidation is that in putting together clinical and tech by design from the very beginning and so I'm really happy to be working with someone like Jessie that can compliment me in everything that I don't understand about studies.

Jessie Juusola:

Yeah and that really goes into play importantly during the design phase. And then also during the analysis phase. Really coming at it with expertise in both sides is what allows you to design the right protocol. And then run the analysis and really find the meaningful pieces out of that.

Allison Proffitt:

So you've both been doing this for a while; have some virtual studies under your respective belts. What has been the most fascinating learning that you've both had from the studies that you've run and analyzed?

Luca Foschini:

For me, every day, has an opportunity for fascinating learning. Maybe my favorite one would be if I were to think about in an intervention to increase flu vaccination that we ran last year. Wellness platform a completely virtual. We sent out very simple push notification messages trying to get people to get a flu shot [inaudible 00:21:50]. And this was deployed to 8,000 people, completely virtually, again. Something as simple as a hundred character push notification that incentivized them to go get a flu shot and we did see a detectable signal. The people that got the message actually had an increased likely to vaccinate. That made me really feel that you can make a change with so little in technology with less than a hundred characters and that's fascinating to me as a technologist.

Jessie Juusola:

From my perspective, there's a lot of the studies we've run that are kind of around the digital biomarker concept that we've spoken about a lot. Others are actually testing different digital interventions. And actually testing if those have impact on different health outcomes. In terms of some of the fascinating learnings for me, some of them have come from the digital tools that we have actually validated in terms of impact on outcomes. So that's another type of study we run in addition to the digital biomarker type of study that we spoke to. But this would be kind of the concept of actually deploying an intervention and testing the impact on health outcomes.

And we all know there's a lot of investment in digital tools, there's a lot of digital tools out there, but there isn't right now a lot of evidence behind necessarily a lot of these tools saying whether they work or not. It makes me feel good about the way that digital health is going when I see what some of these tools actually being validated and seeing that they do have impact. I think there's a ton of potential to make different tools more accessible to people to make health care part of your daily routine versus just something you do every so often when you see the doctor but really to use technology to bring in your daily life, bring disease management to your daily life, bring preventive measures into your daily life.

And so in some of the studies that we've been running, the results are coming out. We've presented some results at the American Diabetes Association this past year, for example. We have a few different papers coming out very shortly in journals. And what you see is that a lot of these digital tools are impactful and they really do help people manage their disease. And I think it's really important to develop that data and make sure that that data is disseminated so that these tools can become more widely available and really improve healthcare for everyone. And make these tools very accessible and make disease management and preventive measures very accessible to people across the country.

Allison Proffitt:

So as virtual studies become more frequent, how do you see them evolving and what will tech's role be in that evolution?

Jessie Juusola:

From a fairly simply standpoint, I think that there just will be more and more things we're able to do. You know, at Evidation, we are a tech company and so in our product is our study platform and that continues to evolve so things we can do today are pretty impressive I think and we can do a lot but there will be even more and more down the road, and we're continuously building the study platform, making it a better experience for participants, improving the capabilities, improving the types of studies, and the study procedures that can be run through the study platform. And that will just continue to grow and evolve over time and allow us to run even more complex studies in a very efficient manner.

Luca Foschini:

I think that virtual studies will become more and more like products. They will become like an experience that really takes the participant for possibly what is a very long time. If you think about some of the longest longitudinal studies, some of them like the Framingham study goes for lifetime, right? So, at that point you start having problems that are very similar to what people in technology have when they're building a product. You need to be engaging your participants. You need to make sure that they don't drop off your study.

So I think there will be a component of more product people coming into the design of study and almost orthogonal dimension. I also think that we're going to see more of fast iteration during the study especially interventional ones. If you come from technology and you want to test whether an intervention like an app works, you do A/B testing or you do even more refined strategy in which you adapt the size of the arms as you go. Of course this is not as easy to do in the clinical world for safety and scientific reasons. But there is still an opportunity to push the envelope there and I think we'll see more of this adaptive design becoming real in the next few years.

Allison Proffitt:

Thank you both for joining me. I really enjoyed our conversation. Is there any last note that you want to leave us on?

Jessie Juusola:

From my perspective, I just want to make it very clear how important adopting this type of research really is. There is a lot of value in digital health and in the digital tools that are available, but we need to actually realize it. We need to utilize it to better understand health outcomes, to develop digital biomarkers. We need to prove that the interventions, the new tools that are out there, are working and who do they work for. I think as we think about personalized medicine, we often think specifically about genetics, but it's so much bigger than that. There's so many characteristics to any given person and digital tools can help us to identify and characterize those things. And then better match interventions and tools to people.

Luca Foschini:

Absolutely. I think, like Jessie was saying, there's a huge opportunity there. I think people are starting to realize that the signal is there and can help us explain things at the more individualized level. So finally moving towards personalized medicine that we've been talking about for more than 15 years now. It's finally bound to happen. What's really important there is to solve all the issues on the way so that these tools are now standardized, they changed too quickly and all the things that we hear if you read any of the standard literature in clinical studies. So I really think it's important there that the regulators step in just like they're already doing and be open to dialogue with technology partners to make sure that this new way of doing studies fits into the traditional context of clinical studies.

Allison Proffitt:

Jessie and Luca, again, thank you so much and thank you for joining us for this Clinical Informatics News podcast. To get more podcast information, visit us at http://www.ClinicalInformaticsNews.com