hello um everyone I'm Beth Linus I'm from the recover administrative coordinating Center and the moderator of state's webinar welcome to the recover research reviewer of the R3 seminar the goal of this woman our webinar series is to catalyze a shared understanding of the research within the recovery Consortium the today's topic seminar is leveraging mobile Health platform Technologies to understand paths it's important to note that the seminar series is focused on scientific research and is not intended to provide any clinical guidance I want to start by thanking everyone who submitted questions please submit any questions that arise today using the Q a feature in Zoom after the presentation we'll answer as many questions about today's topic and presentations as possible some questions may also be answered within the Q a and FAQ document for the seminar will be posted with the recording of the seminar on recovercovid.org it will include the answers for questions relevant to the seminar that were submitted in advance or today questions about other scientific topics will be addressed in future webinars and answers to broader questions about recover will be available in the FAQs at recovercoven.org today's speakers will discuss what is known about mobile Health platform Technologies and past the gaps in our knowledge and how recover will contribute to filling these knowledge gaps today we will hear from Dr Vic caterer Paul a principal with care Evolution care Evolution provides secure interoperability solutions for population Health Management Public Health reporting digital clinical trials and consumer engagement their comprehensive platform enables organizations to liberate standardized and aggregate clinical claims and consumer data into single actual standards based repository Dr ketter Paul received his Doctorate in medicine from the University of Michigan at Ann Arbor next we'll hear from Dr Jennifer Braden an epidemiologist at Scripps Research translational Institute she is the pi of the detect study and app-based research study that has enrolled over 40 000 participants and aims to use self-reported and wearable device data to better understand individual and population level changes associated with viral illnesses including coping 19. next we'll hear from Dr Arjun venkatesh an associate professor and section chief Administration in the department of emergency medicine at Yale University and scientists at the Yale New Haven Hospital Center for outcomes research and evaluation Dr venkatesh is co-investigator and site Pi for the CDC Inspire registry and has been funded by several Federal and Foundation sources to study Health outcomes the use of digital tools for patient reported outcome measurement and large data set analyzes of Health System quality and efficiency next we'll hear from Dr Erica spatz a cardiologist and associate professor in this section of cardiovascular medicine at Yale School of Medicine and of epidemiology at the Yale School of Public Health she is the director of the preventative cardiovascular health program at Yale her research focuses on disease prevention when Macario women's cardiovascular health and Health Equity she is part of the long coveted recovery team at Yale and is pro-pi on the CDC Grant entitled Inspire designed to assess the long-term outcomes of adults diagnosed with covid we'll also be joined by a discussant Dr Andrew Waits an nibib programs director supporting a variety of trans NIH and trans agency initiatives with a focus on health informatics digital health and open science Dr Waits leads digital Health strategy for the rapid acceleration of Diagnostics radx Tech program and established the Red X Mars program which is a standardizing test result test result reporting from over-the-counter Diagnostics he also has held key roles in other nihist initiatives including recover and say yes to covet test thank you and I will turn over to Dr ketterable good afternoon thank you Beth um pleasure to be here I think the my portion of the talk is titled opportunity and need for digital tools and phenotyping to help understand pasc um I serve as a principal at care Revolution as Beth mentioned we're a sub awardee for the mobile Health platform for recover as well as do some work for the all of us initiative and some of the work from the neighbor Dr White's uh integration of testing and digital Health uh that is done I do have an important disclosure I come in with a very strong bias favoring the use of digital tools smartphones wearable sensors that enable passive collection of nearly continuous multi-parameter digital data streams over a longitudinal period of time in the next 15-20 minutes I'm going to try to share some of our experiences over the last eight years supporting the innovators and digital and mobile Health space so with that uh going on to the next slide maybe we can focus on just just this basic idea that started I think um back in 2015 or maybe even earlier than that that the traditional gold standard survey based instruments that have been and will remain vital to help Define disease progression and quality of life for patients um have some uh inherent limitations that we are trying to always contend with one they tend to be infrequent and not continuous and they tend to be a bit artificial in the sense that they do not track the lived experience of the patient at home um they require some manner of an in-clinic or discontinuous measurement and this kind of work has led uh I think in the next slide um the venerable Framingham heart study to think about how to embed a digital electronic Framing and cohort inside the the 70 year old cohort that already is capturing tons of data so starting in 2015 the FHS embedded a new electronic cohort where they were handing out some manner of a wearable a blood pressure cuff other kinds of things to supplement the data that we already had available next Slide the kinds of things that this kind of Technology enables us to do is beginning to now show up as we develop uh proxies and and correlates between the digital measurements we're doing on a continuous basis in the natural setting as opposed to sort of the in the zoo measurement as uh I like to call it that tends to be in the clinic and this this kind of longitudinal analysis becomes really important because it enables development of an individual trajectory of each individual as to the parameter of interest rather than taking something on a monthly or quarterly or even less frequent basis as an end point we we start to see what the correlates are of these kinds of measures that are continuously measured using digital devices next these this kind of next slide please uh the kinds of things that we're able to do of course are very profound because we are able to track um very interesting measures such as mood uh things that tend to um defy being able to measure on say uh you know on a continuous basis using patient reported outcomes or survey instruments and things of that nature and some of the work that Dr shijin said and the team is leading with the intern initiative is beginning to show that changes in multi-parameter data such as sleep activity and other markers like heart rate or correlates like heart rate recovery uh heart rate variability that might be associated with sleep and a physical activity that one can discern out of that uh how somebody's mood or depression risk uh may be rising or progressing over time on the next slide we see other folks doing some really interesting work uh looking at potentially things like asthma so the therapeutic modalities the the kinds of disease states that digital tools are able to track uh transcends the more traditional cardiovascular and mood to now things like asthma pulmonary diseases and of course this has a great relevance to pass and covet this particular trial is very interesting because it's trying to focus on the idea of consumer self-efficacy and the role of digital Diaries in being able to not just generate a better marker and if we uh flip to the next slide and there's an animation on this particular slide we're able to see ecological momentary assessments that can be sent out on what is going on today that mobile technology enables us to do and the ability to then take the data gathered from these ecological momentary assessments and co-locate that data with uh possibly collected digital measures on a calendar view enabling both the individual themselves and the researcher and the clinician to have access to a trend line this same technique has been pursued in cardiac rehab on the next Slide by some of the researchers at University of Michigan where during the pandemic we found that trying to have those who have suffered from an MI that might be post stent placement or myocardial infarction cabbage open heart surgery how do we get folks to be able to exercise at home in a safe Manner and digital tools enable us to go beyond just conducting research but also offer tools that enable individuals to safely be able to do something like cardiac rehab while under the monitoring of a exercise physiologist which of course is entering the world of potentially digital Therapeutics but the the Continuum here and the role of mobile Health platforms uh the point behind this sort of landscape analysis of these various initiatives that we've been involved with shows that it's far more um aspirational than just observational data one of the things that I think we're learning uh on the next slide is the actual notion of what a digital phenotype is um in many cases the kinds of attributes that Define a digital phenotype is that it is continuous it is longitudinal it is measured in Vivo in the real world in the lived experience of the participant at home and it gets far beyond the automated surveys or patient reported outcomes or other measures that are survey based there are special sensors on these kinds of devices and many of my colleagues today are going to talk about the advanced work that comes from those special sensors but some that have already been developed and others that we are in the middle of trying to figure out how to develop the new digital biomarkers in addition to this traditional Frontier Technologies for digital biomarkers there's also this notion of trying to understand something as simple as blood pressure measured in clinic versus measured at home which is better and for what use cases how do we get better at these kinds of things in creating an individual trajectory that better approximates the lived experience of participants at home and of course this entire field includes things like air quality index weather patterns and other things that may be going on in the environment and social determinants of Health that are environmental determinants of Health and accelerometers and other kinds of tools that can be used on a smartphone give us yet another arrow in the quiver of being able to do things like tapping tests and other uh approaches Beyond surveys in order to be able to assess the progression of disease or a lack of disease and and what actual quality of life an individual is having at home so some examples of this on the next page um we we start to get into this is how to develop the next generation of digital biomarkers and how that might assist us in task um how the current generation of digital biomarkers like blood pressure are evolving and moving to a world where non-traditional digital biomarkers uh just like um ecological momentary assessments subjective symptom Diaries versus um Pros that are delivered on a monthly basis how this is impacting a variety of disease States uh and and things that we're trying to track on the next page an example of this is the classic Hauser diary in a movement disorder for Parkinson's patients instead of asking a patient at home to look at how they are how their Tremor or dyskinesia is progressing every half hour so we can right size uh their l-dopa or other medication that may be going on we are now able to potentially give this individual a wearable on the next page where they are able to First automate the diary if we go to the next screen um and that allows us to create a proxy uh to say what is the survey-based answer coming off say a wearable um for the Hauser diary so it's been automated but more importantly using the accelerometer on the wearable we're able to actually track the Tremor and dyskinesia so on the next page we show that that we can have a continuous digital measure that allows for Trend visualization of the Tremor versus the dyskinesia and the L on the horizontal axis represents the times when a patient is taking a medication this kind of Highly granular data that happens to be an event which is when the medication is being taken as well as Digital Data coming off a device allows us to be far more granular and precise in in pursuit of the Precision medicine ideal uh as to that particular individual and what else may be going on in their life moving on to the next page Jennifer is going to go through a bit of this in more detail so I'll simply leave it at the you may have seen lots of articles that are talking about the notion of um consumer grade devices like fitbits and apple watches and garments and aura rings and the whoop ring being able to track what is going on with the consumer and whether it's a influenza-like illness including covid or recovery from such illnesses the kinds of multi-parameter data that is coming out of these devices may help inform and supplement other phenotypes that we are tracking to next page um this is a study that of course Jennifer is going to go into in much greater detail uh that we ran with Scripps Research uh where we were able to enroll 40 000 individuals in the course of six to eight weeks in order to be able to gather these kinds of data so the notion of how far and how broadly we can reach the population is also something that's a very profound in in Mobile Health platforms next slide another kind of interesting digital biomarker is the notion of current gen markers like blood pressure only currently done usually in an Ambulatory Care setting to Define who is even a hypertensive and we've been running certain trials that show that you can actually have home base nearly continuous certainly on a relative basis on the next page uh blood pressure measurement that allows us to understand potentially the daily diurnal variation in blood pressure and how it is associated with morning evening afternoon with or without food other things that may be going on during the lived experience of the consumer at home to see if we can have a more precise personalized evolution of how we measure and Define hypertension and how we track response to hypertensive medications anti-hypertensive medications at home to again in in the progression of trying to get the Precision medicine next slide um additional things that I think are very interesting when it comes to brain fog uh things that are hard to measure with Pros uh with potentially validated survey instruments is this entire world of being able to use uh things that accelerometers and touch screens on smartphones enables us to do some example of examples of this are some of the studies going on at the National Cancer Institute right now on the next page that actually try to follow um the the the actual if I may use it the fatigue the potentially the acoustic neurotoxicity and the cognitive function uh with with the use of chemotherapy for a particular patient using simple smartphones you don't even need a sensor because you can do tapping tests or other markers of evaluating cognition next page um this entire field allows us to correlate um how somebody describes in a survey instrument as to their level of fatigue or cognition or attention uh versus a objective measure of those very same things using the kinds of things a smartphone touch enabled device such as a mobile Health platform enables us to do on the next slide we start to get to the final type of innovation made available in this field which is the ability to work with other apps and innovators that are coming out something such as the ability to track nutritions on the next page you'll see uh Innovative tools that allow the tracking of nutrition not by completing a survey of how much we ate or describing the food but rather taking a photograph of the food that that you are ingesting and having a tool that says uh that looks like a banana what we find is that participants and patients are far more engaged in this kind of a interaction and we end up with a better Diary of something like nutrition tracking so if we go to the next page we start to see how this kind of tool may be useful for uh post-acute sqlae of coven one of the things that we are learning is uh we're in the middle of this chicken or the Ed moment when it comes to past we don't really know what the symptomatology of past is we have some pretty good ideas we have some instruments uh Pros that we have designed that describe which symptoms may be associated with task but in some ways that's what recovery is all about is to find out what are the precise uh symptomatology in the lived experience of consumers and one of the things as a mobile Health platform we're able to do is to provide participants a subjective symptom diary that they can refine and configure and personalize to their lived experience with their symptoms and which particular interventions and treatments are impacting them and this is the kind of thing that allows for great greater expressiveness and granularity of symptom data that is available we find that participants are finding this very useful because we're able to create a trend shown on the next page as to what is going on with the symptomatology for self-efficacy and if we go to the next page you'll see that we're able to provide a 30-day calendar by looking at these data all managed within the mobile Health platform technology that is providing value both to the participant enhancing the granularity and richness of what they are expressing and making it available to the researcher now of course this presents a challenge for the researcher uh to have to come up with new models by which to analyze these data because it is the individual trajectory and lived experience uh in keeping with Precision medicine and personalized medicine that we are having to track as opposed to normative comparisons of the symptoms across a cohort next page um this is the kind of thing that also enables us with the cares act cures act and sync for science some of my colleagues are going to provide how we can have access to real world evidence and data such as electronic health record and claims data that helps establish and create yet another phenotype more easily on a longitudinal study it is not just the data coming from the enrollment site that is important on a four-year basis recover participants may get care from Primary Care Specialists tertiary Care Centers and this model on a mobile Health platform enables a consumer under their control to connect to their patient portal much like they do with mint.com to their financial uh back ends to be able to aggregate all of that data in a single place co-located with their device data and their personal symptom trackers so that they have a single place where the data is being aggregated and shared with the researcher we're able to access some 80 percent of all U.S Healthcare delivered uh by volume on an annual basis with the kinds of endpoints today that are available as part of that sync for science protocol that started at NIH several years ago and has been accelerated with some of the information blocking cares and cures act regulations that enable each participant to have access to their EHR and claims data in an app of their choice such as the mobile Health platform next page so now you know this idea we've talked about EMAs but something that is really challenging to do as a survey instrument or asking folks to maintain paper-based diaries that then have to be transcribed by somebody uh into um into some sort of a canonical um database uh the mobile Health platforms today are able to deliver these In-Place convenient ways to nudge somebody uh to say how are you feeling this morning this evening so the granularity of the information we're able to manage is is much better I guess the final comment is there are things that are highly uh challenging to discover so on the next page we we're talking about in recover it's very challenging to come up with a surveillance system potentially across 40 000 participants for new uh suspected covert infections as we see more re-infections occurring uh so the mobile Health platform could potentially serve as a sensory system allowing and enabling uh convenient uh access by participants at home to say I think I might have a new covet infection uh they raise their hand within their mobile Health platform and then they're able to record that data and on the last slide we show how we can push that to the researcher uh the coordinator at a center to be able to say maybe we need to do uh call this particular participant or their home to see uh if in fact there is an active infection so they can be in the right part of the protocol thank you I think that's all I great thank you so much we will toss it over to Dr Raiden all right thank you for the great background on digital Health Technologies and the application for um improving Healthcare Vic um I'm gonna switch gears and focus a little bit more on covid and wearable devices specifically long covid and so I'm an epidemiologist at Scripps Research translational Institute and I'm a member of the digital health Team and to start out um I'll just go over what is normal and healthy so traditionally when you go to your health care provider they your healthcare provider will rely on population averages to determine whether you're healthy so 60 to 100 beats per minute for your heart rate your resting heart rate they'll recommend seven to nine hours of sleep and they're you often hear the guidelines of 10 000 steps all the little more research is needed for that particular number um however um I um this was from several years ago now before covid I was wearing um an Apple Watch and looking at my own individual data and I noticed that on the left my resting heart rate um was quite um stable around 60 to 63 beats per minute which was within that population average and then on the right I noticed when I got sick with an upper respiratory infection my heart rate jumps rather high up to 77 beats per minute and while this is still within the population Norm or average it was outside of my own individual average which was again about 60 to 63 beats per minute and so um one of my colleagues also was looking at this data set that we had access to which was 200 000 Fitbit users who wore the device for about two years and what he found was that there were times when individuals also had this little spike in their resting heart rate compared to their kind of typical average or pattern and so he found that on average individual's resting heart rate was quite stable it didn't vary for more than three beats per minute from week to week but again when you were able to characterize your own individual normal we can identify these potentially subtle changes that we would not previously pick up on if we utilized the past population average range of 60 to 100 beats per minute and so um prior to covid heading um we also did a study that used that same data set from Fitbit of 200 000 users and we found that if you identify each person's individual average resting heart rate and sleep data you could identify weeks where participants had values that were deviated from their um their average and when you looked at the proportion of individuals each week who had this anomalous data that it could improve predictions for Real Time flu-like illness surveillance and so we looked at five different states and found that the sensor data significantly improved our ability to track infections as they were happening and so this is was really exciting at the time because flu-like surveillance in the United States was typically delayed by one to three weeks it just takes a long time for the data to be collected by different Public Health associations and sent to the CDC it takes a long time for when people first get symptoms to when they see care and finally get tested and so the whole process of collecting surveillance data through this traditional clinical in-person system was really delayed which reduced our ability to respond timely to Public Health outbreaks and so our group works a lot with sensors and digital Health devices and um this the Pu Research Center came out with a study a few years ago that found that one in five Americans are actually wearing a smart watch or Fitness tracker and so there are some differences depending on racial background education whether you live in an urban or a rural area but these numbers are continually growing there is a recent study by The Economist that now says that it's one in four so really have great potential to pull this data from many individuals all over the country to better understand both the individual and population changes associated with viral illnesses um like covid so when kovid hit a few years ago um our team launched the detect study which is um built with my data helps platform care Evolution care Evolutions platform and this platform is great because it enabled us to collect data from any adult in the United States who was willing to join our study and share share their sensor data so we're device agnostic we can pull in data from fitbits Apple watches garmins anything that connects to Google fit or apple health kit and um we also have the ability to pull in electronic health records and also collect um symptom data self-reported vaccinations self-reported diagnostic test results and this allows us to examine how sensor data relates to an individual infection and as thick noted we have enrolled over 40 000 participants from all over the United States um so I'll touch real quickly on some of the initial studies that we did our first one um looked at whether we could utilize a sensor data to improve our ability of um kind of identifying which individuals had covid versus another viral infection and so we found that um symptom data was did a decent job of identifying which individuals had covid versus some other sort of viral infections such as flu or rhinovirus or another common colds but when you added in that sensor data you could significantly improve our algorithms and this model was improved on by my colleague Dr Claire he found that you can even use the sensor data to passively detect who had covid and so this machine learning algorithm could be applied to passively collected data from sensors that we could then potentially notify people to get tested or collect that data to better track surveillance trends at a population level we have also looked at kind of the individual variation associated with vaccination so we um looked at different heart rate and step and sleep um fluctuations compared to per a person's Baseline pre-vaccination and found there were some interesting differences based on whether someone got moderna versus Pfizer whether they had previously been vaccinated versus I mean previously infected versus not previously infected and so this gives us some interesting insight into um potentially antibody response if we could relate this data to biomarkers in the future and then finally more recently we published a study that was very similar to our flu-like illness surveillance study where we use sensor data to better understand population level changes of covid-19 in California and the United States and we used very similar models to our previous study and found that the sensor data could also be used to significantly improve our ability to track covid-19 activity at a population level so again this is really important for speeding upper ability to identify outbreaks and improve our response times and so a lot of our sensor work has been validated by many colleagues over across the U.S and also internationally who
have found um that if you I better identify each person's unique Baseline for these different sensor metrics you can then identify subtle changes that may indicate they're coming down with a viral illness infection potentially even before they their symptoms start um so now I'm going to go in a little more detail on one of our long covid studies so the great thing about using wearables is that you can continuously track people's data um for weeks and months after they get an infection and this enabled us to look at people's resting heart rate step and sleep data both before and after they became infected with covet 19. and so on the left hand side we compared symptomatic individuals who tested positive for covid versus those who tested negative and we found that the individuals on average who tested positive had this much higher resting heart rate response during the acute phase and then there was this interesting dip in resting heart rate before it went up again and stayed elevated for on average about two to three months and interestingly sleep and step count went up but returned to Baseline faster and on the right we looked at we grew to People based on their racist heart rate deviation during the second month's post symptom onset when we found that there was a subset of individuals who had this resting heart rate that was elevated and remained elevated for much longer it didn't even go back to Baseline during our follow-up period and so we think that a decensored sensor data can be used to a better track long covid and how people are responding to their infection whether they're returning to baseline or whether they're still experiencing these abnormal fluctuations compared to their Baseline so um we also collected um different symptom variables during the acute phase of infection and we did see that there were certain variables that um during the acute phase that were associated with this longer term um resting heart rate fluctuation and so um variables such as shortness of breath during the acute phase were associated with a higher frequency of being in that resting heart rate group that was extremely elevated and so um some of our colleagues um at the Robert Cox Institute have actually replicated our findings and they have also shown that um resting heart rate on average remains elevated for several months um post-infection they have also looked compared vaccinated and unvaccinated individuals and found some slight differences with resting heart rate returning to Baseline quicker in the vaccinated group and so as I mentioned early on we are also interested in aggregating this data at a population level to better understand um kind of Trends in a in a state or a county so that we um these can be used to inform Public Health policies um resources and one part of the piece is sensor data that can be combined with many other data streams such as rapid test results Google search data movement of individuals vaccination and many other in Wastewater surveillance and one piece is also looking at recovery from long covid since that's a growing population of individuals across the U.S and globally who are now suffering from long covet and wearables can potentially be used to track when people are moving into the Recovery Group and I'll quickly mentioned that one of my colleagues is also working on a long covet and pacing study so we we think that wearables not only can be used to better track and quantify long covet but can also be used to potentially help manage symptoms and so she's working on running a study that will evaluate pacing and management of symptoms so quickly um I'll just touch on some of the possibilities and challenges of wearable sensors for long covid so one of the great benefits is that so many individuals now where these devices one of the challenges is they are still quite expensive so there are we're working on different ways to get around that providing devices to individuals who might not have one or also looking at low-cost Solutions such as sensors in a smartwatch camera that can also calculate someone's heart rate potentially respiration rate and also activity level the great benefit as Vic mentioned is this is continuous data from the comfort of a person's home so it really gives us a unique view into what their individual Baseline and Trends are over time also it allows people to participate for many any time and from anywhere which allows us to include people from rural areas or places that might not have been included in traditionally in clinical trials due to the distance to the site um and also as these devices um evolve over a time we're seeing new metrics that are more widely adopted into the um common commonly used fitness trackers and so these new metrics will further improve our ability to track infectious infectious diseases and differentiate changes associated with the viral infection versus changes associated with stress or um dehydration or alcohol consumption or other causes um one of the biggest challenges is the long-term engagement and so something that we're going back to to improve that is providing useful data back to the participant really working on establishing trust and also um always increasing diversity and representation so that these Health Solutions can be utilized by everybody and finally I think one of the greatest benefit potentially of these devices is they provide objective data to support some of the symptom data that many participants have experienced but have potentially been ignored or not really recognized by the medical fields and so being able to quantify what people are experiencing through resting heart rate activity and sleep data can really give us a better idea of what each individual is experiencing I think in the future the ability to predict which individuals are more likely to develop long covid versus not as an interesting application as well as evaluating different treatments and symptom management also looking at different subtypes we know that some individuals go on to have this prolonged tachycardia some individuals experience a relative bradycardia or lower resting heart rate so there's a lot that we still have to learn by using the sensor data and finally I'll mention that none of this would be possible without our huge payment at Scripps who really pivoted to work on this study during the pandemic oh thank you thank you Dr Raven next we'll hear from Dr venkatesh and Dr Erica's pets hi everyone um Sergeant venkatesh here from Yale along with my colleague Dr Eric espasmio um and we're sort of excited to tell you about a project that is uh really it's in some ways uh Midway point we recently completed enrollment but the findings and the early results are just beginning to come out and so what we want to speak to is the aspect of our project that comes from digital Health platforms within the Inspire study and go to the next slide uh this study the Inspire study is a really a data registry amongst individuals with uh coveted infection that's supported by the CDC uh just to sort of be clear this is our presentation today it's really from us as researchers at Yale and not official sort of CDC or HHS uh statements uh nor if necessarily from the official officially from the study project itself next time the Inspire study Consortium was started really early in the pandemic I think the CDC was thoughtful about recognizing that there would be a need to prospectively follow individuals and understand long-term outcomes there was at the maybe at that time very few and early anecdotal reports of things called long covet or prolonged symptoms and at the time a set of eight centers as you see here on the slide were able to pull together with Rush University serving as the administrative Corps for this work the University of Washington providing uh the clinical core function and ourselves as Yale as the primary analytics center for the study and then additional sites at UT Health UT Southwestern UCLA UCSF and Jefferson are supporting the enrollment of people really across the country next slide the study primary objective was to compare disease trajectories symptoms patient report outcomes clinical outcomes amongst adults with covid-19 compared to a control population we had originally planned to enroll approximately 4 000 individuals we were fortunate to receive CDC support given the early success and enrollment to increase that number to 6 000 participants and actually just completed that enrollment Target a few months ago the study is designed to provide up to 18 months of follow-up for all of those enrolled and collecting a variety of outcomes through a variety of data sources what I think makes the registry unique has been an early focus and an early sort of attention to Patient reported outcomes largely using existing patient reported outcomes to understand the long-term sequelae of a coveted infection and Eric I'll speak to that in a little bit the other thing that's unique about this study is this is where the use of the digital Health platform comes in is our hope to link those patient reported outcomes to other data that may be available in real world electronic health records and the goal of this is to really determine the risk of long coveted adults presenting with symptoms of covet infection compared to people with a negative test I think one of the interesting things about our study uh that often gets missed is that it's not only perspective but we're enrolling people prospectively that had symptoms at the time of enrollment uh at a ratio of three to one three positives to one negative amongst those who had a positive test for covet versus those with a negative test next slide the digital Health platform really is being used in this study to link across a variety of research work functions there is a part of the digital Health platform That's essential for screening and enrollment we have a web-based approach to recruitment where we can direct anybody to that website really not just within the eight geographies of those enrolling centers but Across the Nation screening questions can be answered by anybody via that web page they can provide electronic consent they can sign up for the study they can connect to the digital Health platform and they can complete the initial Baseline survey necessary to be enrolled in the study and for some who are you know the most electronically Savvy that can be done entirely without any additional contact that said we have a tremendous amount of research support and resources that we've put in to make sure that we're purposely capturing a broader population that's more generalizable and not just the those who may be the most tech savvy out there we can then use the digital platform to collect a variety of survey data some of these include socio-demographics and data around employment and finances as well as standardized measures of patient reported outcomes across a range of domains the same platform is then also used to connect each participant's various electronic health records to the study the vast majority of people connect what is data from a hospital or a health system often that may be the primary hospital or health system in which they receive care for covet or testing for covid at the time of enrollment that said one thing we do notice is that particularly amongst younger populations there may be people who don't have a primary health system don't primarily have a lot of formal health system use and therefore may have a pharmacy portal or something else that has actually richer data available for linkage the reason this is important is that if we were to limit ourselves to studies around Long covid people that are existing within ehrs we'd be really prone to studying those who access care seek care in traditional hospital or ambulatory Healthcare settings and who have the ability to access care in those settings and so our hope has been to use the digital platform here as a way to broaden access to these research to the research study and also be able to capture populations that may otherwise be absent from prior retrospective work next slide how have we gone about recruiting and screening one of the things that was really fascinating about doing this project is we had to flip a lot of our conventional approaches to research on their head when we kicked off the study this was a study where the discussions with the CDC about beginning the study occurred in 2020 and the build of the actual recruitment modules and the ability to do the study were occurring by late 2020. and so we did a lot of different things we partnered with Community leaders and health departments and several geographies to help identify individuals who tested positive or were being tested tested for covid at some sites they were able to work with their health system to get an electronic health record review of anybody who was tested not just necessarily within the hospital but potentially at drive up or ambulatory testing sites that are run by Health Systems we did digital advertisements via a variety of venues both social media and not social media we recruited in many different settings and we found that recruiting particularly in ambulatory settings and testing sites became a higher yield way to identify individual those who are interested in the study and broaden the population as an evolution from our very early enrollments that were largely at a hospital-based setting we provided a lot of telephone support and that text and phone-based support along the way for these interactions some people received a very high touch approach to recruitment versus those who required less touch and that was largely driven based on sort of not just their sort of experience with using digital Health platforms their familiarity with these types of tools but also for them to learn about the study themselves Our Hope and Ethos of this project from the beginning has been that it is a partnership with the participants in this study throughout this work and in to do so it was very important to us that we were had an open channel of communication with anybody who was interested in the study from the beginning the final thing I'll note on the Recruitment and screening methods for this study that are unique is in comparison to a lot of other prospective observational Registries that may have in-person contact given all the limitations with recruiting individuals and participants early in the pandemic particularly around exposures research staff safety we built a recruitment model for this at many sites in a virtual platform where essentially we were able to make virtual contact with patients often through something as simple as a text message early to express interest in a study and then go through their entire Recruitment and enrollment process in a virtual manner without ever having the in-person contact that many of our study states were used to in a pre-covered era next slide so what did this look like often a participant would access a link to Hugo Health which is the digital Health platform we used via a website that could be a computer smartphone tablet screening questions would be answered to confirm eligibility and that would trigger a process on the participant side as well as on Research staff side to identify whether or not and confirm whether an FDA approved screening test was used for covid-19 once that was done participants would link their digital Health portals to the Inspire study by giving permission for the research team to use their data for research purposes and then begin to receive surveys via the um digital Health platform Hugo Health itself or via emails or they could be received text links and people would sort of you know ask for the getting the content in a way that was best suited to them surveys are conducted every three months from Baseline up to 18 months and intermittently we're also able to pull digital personal health record data largest electronic health records into the study data set as well next time this is a quick picture of what this looks like you can click again for the animation you know amongst 5982 that had been enrolled and were actively participating in the study you can see that about 50 3076 have an active Active Health System portal there are some other incremental percentage that also have a pharmacy portal that is linked to the study but the important thing I think that's worthy of recognizing within this is that just as there is uh different challenges and follow-up with any type of prospective observational study and outcome ascertainment whether you're calling patients or you're following up with them via surveys or sending them emails the same is true for the connectivity of patient portals and electronic health records so despite a lot of the improvements that came from the cures act that reducing data blocking there are still many practical barriers that can make it challenging and difficult for individuals to link their their own Hospital health system portal to a digital platform and more importantly to keep it connected over over time often there will be password changes different sort of version changes that happen within a platform within an app within the actual electronic health record system used by a Hospital health system and each of those creates a different friction point that makes sort of the sustaining of a portal connection and the sustaining of that real world data increasingly challenging next slide and click to the next one what kind of data do we see when we are able to make that portal connection when we're able to see that EHR data what we often find is that um we get a lot of diagnosis data immunization data test result data particularly uh is probably the most uh prevalent of all the data types that we see but even amongst that what's interesting is that it's not always that there's not always a coveted data point present uh we also do see a fair amount of medication data through these portals The Challenge and the reason I think I wanted to present this slide for everybody to understand is that in general it's very difficult to get a full fingerprint or a full picture of someone's electronic health record data because we receive care in such disparate and fragmented settings and from disparate and fragmented providers what we often see is that even in a primary portal linkage where somebody has access to in these cases often enterprise-wide implementations electronic health records it's really half to two-thirds of the data that we're able to capture for many given individuals next slide another data challenge that's presented is trying to triangulate and understand the differences between real world electronic health record data and something such as in our case an RA verification of gold standard one of the analysis we did early on was to understand the electronic health record data around covid testing in comparison to the project study step we have in which Ras manually verify a lab test result so amongst the 5983 that had been enrolled 3 000 participants had had an EHR data portal connection of which 1658 had in that EHR covid-19 test result available at the time of their enrollment that aligned with the time that the ra verification was preserved what we said is we said let's look if plus or minus three days from when the ra verified a test result is positive negative to see if we can find the same result in the electronic health record and what we found is in some ways very fortunate 99 of the time if the ra said it was negative we find a negative result in the EHR very low amounts of time where the ni only one case for the ra said it was 9 negative and the EHR had a positive test what's challenging here amongst all of these is the cross where the ra um in 285 instances or 25 percent of cases identified a person as being positive for covet even though the EHR had that as negative that reflects the complicated nature of covet infection we know there are false positive and false negative results to initial tests and that certainly is a component of this we know that the likelihood of those test results changes with each day of illness so you can imagine many scenarios where somebody may have a negative result on a pacr performed at a hospital take a rapid home test the next day that comes back positive and ultimately be included in the Inspire registry as a positive for covid-19 when in fact the EHR result shows something different and so these are the sort of data considerations that we've been wrestling with as we start constructing studies sub studies within our registry next slide this is my last one so how do you deal with this I think from a real world perspective when we're trying to do research with these types of data we have to start understanding that no data is perfect that there are a variety of data that we will get uh with different levels of signals and noise and we start thinking about how we can do something Smarter with that data and do research in a different way and so one of the things we have done within the Inspire study for example is when trying to understand the relationship between vaccination and long covet symptoms is to use both not limit a study to only those that may have electronic health record data available or limited only to people who've responded to the surveys knowing that follow-up rates vary but rather glean what we can from both where we have 20 percent of participants that may have say EHR data available but limited survey data on the other hand we have 50 percent of respondents with good survey data about vaccination status but no EHR portal linkage you put those two together and we get roughly 65 or 70 percent of the entire registry where we can start to construct a variable around who's had vaccination prior to their a coveted infection to understand what that role of a coveted infection a post-vaccination infection is means for long-term symptoms and so these are the ways we've sort of been dealing with some of these data I'm going to turn it over to Erica now so she can describe some of the patient reported outcomes for this study and then describe some of the next steps of where we're going thanks so much Arjun next slide please I'm just going to take a little detour to talk about our patient reported outcome collection and discuss some of the surveys that we're using as well as some of the challenges in collecting prioros and interpreting pros and what we anticipate to be some of the challenges that come up when we are matching with clinical data um one of the main challenges for long coveted research is that there's no definition of long covid so we don't have a survey that can diagnose long covid um and long covet is very heterogeneous and can look very different and so one of the um main questions that's asked of our study is is how many people have long covid or what is the severity of long covidm amongst people who have ongoing symptoms and there's really no good answer to that and so we're pretty careful about not diagnosing conditions or labeling people based on symptoms or surveys that were not created for a long covid at the same time we recognize that many people are experiencing symptoms and impacts on their quality of life and we want to capture the range of these uh many different effects and that's important to really reflect the patient experience and to make sure that we're being comprehensive but it also poses a challenge for analyzes right we don't have a primary outcome and so if we're looking for factors that are associated with long covid we need to kind of create composite variables that potentially describe long covid recognizing that we could get into a pickle of multiple analyzes and what that means for understanding the factors that are associated with these outcomes and of course in any longitudinal study as Arjun was mentioning the Inspire study is 18 months there's Dynamic coveted positivity status so the people that were enrolled in our study that were initial negative may be exposed later on and become coveted positive and so tracking the dynamic covet status of patients and interpreting that in light of their Pros is another challenge that really requires good data collection and like Arjun was saying integration of both patient reports as well as clinical data from the EHR next slide you could go to the next slide great um so what are some of the pros or other surveys that we're using in our Inspire study we are mostly looking at a pretty large range of Baseline characteristics that go in uh that to understand the acute covet illness how severe it is whether they were hospitalized again as origin was saying we have very few that are hospitalized since we were mostly targeting outpatient covet testing centers we'll also get a baseline check on their symptoms from covid their Baseline comorbidities and Baseline social um and lifestyle factors like their employment status and social determinants of Health um at every three months we are doing a pretty deep dive into symptoms as well as patient reported outcomes and we are using the um uh persons under investigation CDC symptom checklist to understand whether symptoms have uh never occurred resolved persisted or emerged um we are also using a CDC short symptom screener for myalgic encephalomyelitis chronic fatigue syndrome that is used by the CDC to understand some of the more systemic symptoms um and then for outcomes we're using the promise 29 which as many of you may know has seven dimensions including physical function anxiety depression fatigue social participation sleep disturbance and pain as well as the promise 8A which assesses cognitive function and then some more Global Health measures like return to work exercise Vital Signs PTSD and Global health status um next slide this is the CDC short symptom screener for mecfs and I show it to you because the questions are really good they delve in deeply into the things that we hear most from from patients that we see clinically who are struggling with symptoms fatigue and exhaustion muscle aches pain in the joints unrefreshing sleep forgetfulness and memory issues difficulty thinking or concentrating dizziness or fainting and these questions are posed to patients and they're asked how long have you had the symptom was it present before you had your covid test next slide and then it goes deeper into what's the frequency of these symptoms the severity does rest make it better what's its impact on occupational educational social or personal activities is it worse with physical or mental exhaustion but still the challenges is that this is a screener and that these symptoms for mecifs but as you can see if we take a real look at what's in this survey many of these symptoms can be present with other post-infectious syndromes um that people may be presenting for as well as they can accompany chronic disease they can accompany acute illness so distinguishing what is long coveted from what may be another viral sequelae or other medical illness is a challenge next slide we do a good job of trying to look at our data across uh across the different sites with quality checks and we have a few papers that are in the works around the first 1000 patients and weekly Communications around enrollment the ratio of covet positive to covet negative the survey completion rates and the connection to EHR data next slide this is just an example of like the weekly emails that we give to try to make sure that we are being as complete in our survey data collection at three months and six months so we'll get a weekly email about how many people are um up for their three-month survey how many completed it and this helps our team really rapidly pivot to engage people that are not filling out their surveys to ensure that our data are the most represented next slide so I'm just kind of gonna close because I know we're out of time um there's a lot of trade-offs with different outcome assessments Pros are critical to understanding low long covet but leave a lot of data gaps there's a lot of concerns that we're not measuring clinical entities like pots which require Vital Signs and lab tests and are uncertain about what that how much that is to characterize long covid and PRS may not distinguish long covet from other post-infectious syndromes or other disease entities that cause prolonged symptoms fatigue and other systemic disorders but I think that they help to capture still the patient experience and help us differentiate people that are really struggling in their recovery versus those who aren't thank you so much for um your attention and I look forward to the Q a session thank you Dr spatz now we'll hear from the discussion Andrew weitz uh thank you and actually a really big thanks to our speakers um those were some amazing state-of-the-art demonstrations about how mobile apps wearable devices and other electronic Healthcare data can help inform the status of an individual as well as a population of individuals um there's a lot of content packed in those presentations so I'd like to take maybe just a minute or two to review some of the key points and then we can move into the question
2022-11-07