Using Digital Health Technologies in Cardiovascular Innovation

welcome everyone we'll get started in a few minutes well welcome everybody um I'm am gambari co-chair of the Innovation group at University of Michigan on behalf of the innovation academy and the frankco Cardiovascular Institute I have the great pleasure of introducing my colleague and friend Dr Jessica gobis who is assistant professor of of medicine at the cardiovascular uh Institute and University of Michigan uh school of medicine she received her medical degree from University of Michigan and completed Residency program at University of Pennsylvania she completed her General and advanced heart failure uh training at University of Michigan where we were co-fellows together for a while uh she also received her Masters through the University of Michigan School of Public Health she's actively engaged in clinical practice in areas of heart failure heart transportation and mechanical circulatory support uh her research is really focused on using digital Health technology smart watches to improve delivery of care and optimizing treatments for patients with Advanced heart failure she is a past alumni of the fcbc Innovations program in 2020 and is going to speak to us about Innovations using VIs Health Technologies and cardiovascular care welcome Jesse thank you thank you so much for inviting me here and for the the very kind introduction let me just go ahead and share my slides here okay great hopefully you can you can see that all right um so um I'm going to be talking today about some work that our group has been doing in um using mobile Health Technologies for Behavioral modification in patients with a range of different cardiovascular conditions I really want to focus today on highlighting um a novel intervention design just in time adaptive interventions um and and an associate experimental design micro randomized trials that our group has been leveraging to intervene on patients with cardiovascular disease um in including those enrolled in cardiac Rehabilitation as well as patients with hypertension so I I'll I'll dive deep into a couple of case examples uh I have funding through the nahh NSF and porori um the work that I'm going to be talking about today has been funded um through a number of institutional grants here as well as through an American Heart Association strategically focused research Network Grant and Health Technologies and Innovation so again as I said over the time together I'll I'll start by just level setting make sure everybody understands what mobile Health technology it is um I think we all know that already but what types of data we can collect talk about micro randomized trials highlight how those have been applied in the Valentine and mybp my life studies and then talk a little bit more broadly about challenges and future directions in the field of mobile health okay so what is M Health Technologies I think we all likely know what this is but the World Health Organization defines mobile Health Technologies as the use of Wireless Technologies to support the achievement of Health objectives so this includes things like text messages GPS and use of wearable garments or accessories that can provide physiologic monitoring now importantly mobile health does not include tele medicine which is a separate type of electronic health now wearable devices are a subset of mobile Health Technologies and these are externally applied devices that capture functional and or physiologic data often continuously and are paired with a second device such as a smartphone which can be used to collect and or transmit the data so when we think of smart think of of wearable devices we most often think of smartw watches like our Apple watch or our Fitbit however those we do include a much broader range of devices things like the wearable cardioverter defibrillator or life vs or the aura ring as just a few other examples now this becomes really important later when I think about um and talk about just in time adaptive interventions but it's important that we understand the types of data that wearable devices can collect so we most often think of these devices as passively collecting data on things like heart rate or blood pressure um there can also be blood alcohol level collected through a transdermal alcohol monitor but in addition to this passively collected data wearables can collect data actively and so that's typically through Short mobile questionnaires termed ecological momentary assessments these can capture things such as stress and anxiety pain or fatigue which may be much more difficult to capture passively now these actively in passively collected variables they they may be intrinsically informative but they can also be aggregated as part of risk scores to inform Health prediction so as one example in patients with opioid use disorders a machine learning model was trained to predict heroin in cocaine Cravings using GPS data and participant level demographics so I wanted to turn now and talk about just in time adaptive interventions and micr randomized trials and I'll I'll start first just at a high level with just in time adaptive interventions and start very high with how we pronounce it which is Jedi um but uh Jedi for those of you who are not familiar is a um typically an M Health intervention delivered through the use of a mobile application and push notifications and Jedi were designed to deliver support to the right individual in the moment where they most need it so the idea is that these interventions rely on passively sensed data and data actively collected through Short mobile questionaires to identify the optimal time to intervene so this will be area opportunities in which individuals are at risk say for um uh and those I highlighted with an an um a use disorder perhaps uh an addiction perhaps relapse or opportunities to um uh capture on motivation to improve health uh and and Jedi take advantage of response heterog heterogeneity by individualizing intervention options to the specific and changing needs of individuals so in in the um screenshot you'll see on the the right hand side of this slide this is from a mobile application that Mike DOR designed who I know many of you may know um and this was part of a a clinical trial designed to reduce sodium intake and encourage selection of lower sodium food Alternatives so participants when they were entering say a grocery store or a restaurant would be offered possible um uh low sodium alternatives to foods that they commonly order or purchase so what then are micro randomized trials so micro randomized trials are an experimental design to Aid in constructing empirically based just in time adaptive interventions through the serial randomization of participants to different types of interventions and or different levels of an intervention so the idea around micro randomized trials is that they can be used by researchers to to decide whether to include a Time varying component in a multi-component intervention and in which context delivering the component of an intervention is most effective and I'll show you in the next slide what I mean by that but the idea of a micro randomized trials it's an optimization trial to help with um the design of just in time adaptive interventions so again this micro randomization refers to sequential randomized assignment of an individual to one or several components of an intervention throughout the course of a trial and the unique aspect with which micro randomized trials operate is that people can be randomized very frequently the same individual can be randomized every hour or multiple times a day so uh the wearable device will passively collect data such as current activity level or location to determine whether a participant is Avail available to receive a mobile Health intervention if they're available say they're not driving or not currently Physically Active then they will be randomized with a preset probability to receive or not receive an intervention that time point at which they're randomized that's called the decision point and that intervention is um dictated by tailoring variables so things such as the weather the time of the day based line level of physical activity could be used to make sure that the individual receives the right intervention at the right moment and then the success of that intervention or or the decision to intervene or not intervene can be assessed through proximal outcomes and the idea is that proximal outcomes are mediators of the long-term distal effect so step count for example in a physical activity intervention could be a mediator of changes in in uh functional capacity or weight loss that may be the desired distal outcome and proximal outcomes assess the effect of your intervention or non-intervention so you'll measure for example step step count after a decision point when people are intervened on or when they're not intervened on and it's that within person contrasts that make micro randomized trials so powerful so they're statistically and temporally efficient and allow us to to improve our understanding of causal effects so hopefully that was a a high enough level overview to give you some context now as we dive in deeply into two different case studies here in which we delivered a uh just in time adaptive intervention that incorporated micro randomization so the first example here is the Valentine study so the Valentine study was a prospective randomized control trial designed to evaluate a just in time adaptive intervention to augment and extend the benefits of cardiac Rehabilitation the study integrated a mobile application with physiologic and contextual information from wearable devices so we enrolled low and moderate risk patients and then randomized them to the control or intervention groups of the study and importantly both groups received a um a Smartwatch that was compatible with their smartphone so Android users received a Fitbit Versa 2 and iPhone users received an Apple Watch and then we followed participants for or um the remotely assessed outcomes of six minute walk distance step count um and quality of life so as I mentioned the core of the intervention was a just in time adaptive intervention consist in of contextually appropriate Smartwatch notifications and notifications were one of two types the first type of notification was an activity notification so these types of notifications were designed to promote low-level physical activity and they were meant to be appropriate for the current environment and actionable in real time so these were tailored on factors such as the weather the time of the day day of week and phase of cardiac Rehabilitation the second type of notification were exercise notifications so these notifications encouraged participants to plan their exercise for the next day and suggested new activities to increase their exercise repertoire so in the left you can see an example activity notification that would be appropriate for say a 64 year-old female on a sunny weekend day two months after she started cardia Rehabilitation for a heart attack so I talked before about how in a um uh micro randomized trial you can measure the effects of your outcome of your intervention through proximal outcomes so for the activity messages we looked at step count 60 minutes after a text message after the decision to randomize um as measured by the Smartwatch and then for the exercise planning message our proximal outcome was the subsequent um exercise minutes so in addition to receiving notifications intervention arm participants also had access to a mobile study application which was paired with their Smartwatch this allowed them to self-monitor their activity data things such as step count or exercise minutes and then allowed them to set and complete activity goals which they could then adjust based on their performance between October 2020 and March 2022 940 patients were screened for eligibility of these 422 patients were eligible for steady participation and of whom 223 were ultimately randomized to the intervention or the control group of the study and 220 of these were ultimately enrolled so using this completely virtual Str stry of no in-person contact we were able to successfully enroll 13 participants a month and um I'm happy if if I don't I'm not going to get into our pipeline for um enrolling um patients into these remote strategies today but uh I think that that's something um that could be a talk in and of itself and and I'm happy to talk on the side with anybody who has an interest in in how we approach that so this is our study population so 38% of participants were greater than 65 approximately a third were female we enrolled from both the University of Michigan and from Spectrum Health with Approximately 80% of participants being from Michigan medicine and about two-thirds of participants were enrolled and participated with an iPhone and an Apple Watch and importantly there were no significant differences with between the two groups the intervention and the control group with respect to any of the Baseline characteristics so we had outstanding compliance within our study participants wore their watches for a median of 180 days and 88% of participants wore their watches at least 75% of study days and here demonstrating the power of these micro randomized trials participants were randomized almost 71,000 times to receive or not receive an activity text message and on average this led to about one activity text message per day which is what we were targeting so displayed here are the primary results of the study so this is the primary results of the randomized control trial although I'll show you in the next Slide the micro randomized trial results which I I think are actually the more interesting but just to orient you here to what you're looking at so on the left you'll see our primary outcome which was six minute walk distance and on the panel on the left hand side here that's with respect to Apple watch users Fitbit users are on the right and we have our control group here in this uh orangey color and the intervention group in in teal so in univariate regression analyses a mean change in six-minute walk distance for the intervention group as compared to the control group was not s significantly different um between the two groups in six months we did however conduct and an exploratory analysis in which we adjusted for Baseline characteristics consistent with recommendations from the FDA and while these results were relatively unchanged overall they suggested possible improvements in six-minute walk distance in Apple watch users with a P value that was just over 05 at 3 months importantly in a univariant analysis we did find that mean change in six-minute walk distance was significantly greater for Fitbit users in um uh in the intervention group so suggesting possible near-term effects of our intervention now turning to step count here on the right hand side again Apple watch here and Fitbit on the um on the right and an intervention in in group here in teal and we did not uh have any significant differences in Step counts at any of the time points between the intervention and control group so turning now to the results of the micro randomized text messages which I think are the most interesting results from this study uh so we looked a priority at three different phases of the study which corresponded to phases in cardiac Rehabilitation so we refer to these as the initiation period which was the first 30 days of the study and corresponded with enrollment and cardiac Rehabilitation the maintenance period which was months 2 through four and then the completion period which was the period following graduation from cardiac Rehabilitation and you'll see in the blue here these are Apple watch users and then in the yellow are our Fitbit users so amongst Apple watch users again in blue during the initiation period of the study delivering a text message increased 60-minute post randomization step count by a nonsignificant 10% messages did not significantly impact step count during the maintenance phase of the study although they did significantly increased 60-minute post randomization step count by 6% during the completion phase of the study again the period after graduation from cardiac rehab amongst Fitbit users during the initiation period of the study delivering a text message increased 60-minute post randomization step count by an impressive 177% although this effect diminished over time and we didn't see any significant impact of 60 Minute step count during the maintenance or completion phases of the study so in addition uh to the results that you see here we are doing a uh series of moderator analyses in which we were trying to understand both the person and message level characteristics which are leading to The observed effects so we're looking at traditional factors such as age race and gender but we're also looking at things such as whether the notification was personalized with a participant's name what time of day the notification was sent at the weather and factors similar to that to help us better understand and again construct in the future these optimal just in time interventions we're also trying to understand the effect of our intervention through uh participant interviews and so we're coming at this from multiple different methodologic approaches so this was work that was recently published in jaha and was led by one of our Internal Medicine residents here in nraa lurry and nraa spoke with 177 Valentine participants in the intervention ion group to try and understand their experiences as part of the study and specifically interacting with those notifications so participants in these interviews generally match the larger study population they were on average 65 years of age about a third were female they were evenly split between Apple watch and Fitbit users and were predominantly of high functional capacity now nraa identified a number of important themes as she spoke with our Valentine participants and I'm not going to go in in detail into all of these here today but I'll just highlight a few of them and and and provide a few illustrative quotes and then I'll refer you to her article for for kind of the rest of our lessons learned here but the first lesson that nraa identified was around engagement so participants engag with four main features of the M Health intervention so this was activity tracking tailored text messages goal setting and weekly email summaries and overall we learned that at least qualitatively participants preferred the activity messages those that were meant to be actionable in real time and while subset did prefer exercise planning messages most people again were were interested in these activity text messages so here's a quote here the ones telling me to if you have a few minutes to get up and do whatever set exercise from that message those were the ones that I liked the most because they were really quick and it was something that I could do pretty much at any time now uh certainly I would love to say that everybody uniformly loved our messages although this NE wasn't necessarily the case and I will say that people with higher functional capacity generally tended not to like these notifications they found them too simplistic so this one says Hey name hey Jesse why don't you stand up and touch your toes five times I thought those were you know adorable but I didn't really ever take the prompts so this participant felt that it was not appropriately tailored to his Baseline level of functional capacity and this was an important learning point for us now ntha also focused in on personalization so she tried to look into how the M Health intervention fit within participants lives with respect to usability applicability to their goals and contextual appropriateness so even those who had little experience with M Health technology generally felt that the mobile application was easy to navig navigate and integrate into their lives now participants generally wanted more personalization and they felt that individually tailored messages were more likely to motivate action so this was a woman with low functional capacity who said if I set my goal with 10,000 steps and I did 5500 steps I would have preferred a message saying 5500 steps 4500 more to meet your goal I would have liked a more quantifiable message to meet my goal the other thing that we learned is is we explored how participants reacted to notifications that were inappropriately delivered so for example telling somebody to walk outside when it may be raining so something that that was not contextually congruent and for the most part this really didn't bother participants I think they they appreciated what the text messages were designed to do which was encouraged them to find ways to be active in their current environment so here's a quote to support that actually I never looked at them that way I took it as you know hey it's a good reminder not to necessarily go outside and walk around but just to get up and move and finally namratha's work really highlighted a number of important future directions for us that we are we are incorporating into our upcoming studies so participants desired greater tailoring of text messages to their unique en environments as I said and though although participants acknowledge the trade-off with with respect to potential invasion of their privacy they felt it was an AC acceptable sacrifice to improve their health so this was really an important learning point to us we thought that security and concerns about invasion of privacy were going to be a major concern for participants and that we are worried that if we tailor these messages to too great an extent that that would kind of deter people from participating and that really seems not to be the case so this quote highlights this I joined the study because it was a way for me to be more accurate as far as measuring the outcome of what I was actually doing and if I'm doing that and you're watching that part of it I really don't see how I could have an objection to seeing what I'm doing in real time the other learning point for us was that we really didn't have a way to incorporate social support into our intervention so some participants really wanted to see that sense of community community that they felt in cardiac Rehabilitation embedded within the mobile intervention and they desired kind of some friendly competition that they had felt um that they were able to achieve when they were in cardiac Rehabilitation and they wanted to see that competition or gamification within the mobile application so this one says like in my case I had a few people that I actually associated with that rehab if you could connect to a bigger group of people to see how other people were doing I think that would have probably made us a lot better than we were so those were some of the the learnings for us from the Valentine study and and again we're still continuing to to learn and and adapt based on ongoing analyses and specifically work focused on those subgroup and moderator analyses now I'd like to turn and talk briefly um now about um the my BP my life study um and I say only briefly because this study is still underway and the primary results have not been published yet but it does incorporate some novel methodology and is is another way to to illustrate how micro randomized trials are being applied in the field of cardiovascular disease and so I'll just talk here um at a high level about um the the study design so this is a work this is a study that's being led by uh Dr nalu and uh Dr dorch here um but the mybb my life study is a prospective randomized control trial for patients with hypertension and it recruited from two sites it recruited from the University of Michigan Health here in an arbor and from the Hamilton Community Health Network in Flint Michigan and what's really important to know is that for those of you who aren't aware the Hamilton Community Health Network is a federally qualified Health Center in FL Flint Michigan and serves a predominantly black population participants in the mybp my life study were randomized either to the control group or to the intervention group and the intervention was a just in time adaptive intervention promoting increased physical activity similar to the Valentine study but also the selection of lower sodium food choices and I'll show you in the coming slides what that looks like now we completed enrollment of 602 participants in July of 2023 and participants are followed for 6 months for the primary outcome of change in blood pressure and so we had followup from our last participant the last week of January and so we are really excited to see the results of this study in the coming weeks so one of um the core aspects of the intervention was again these delivery of these push notifications promoting physical activity and lower sodium food choices so physical activity interventions similar to Valentine were designed to disrupt sedentary behavior and to encourage low levels of physical activity on the order of bouts of 250 to 500 steps so these notifications were tailored on a in a much greater number of factors than Valentine and included things such as the time of day and day of week the weather but also the community and mobility in contrast dietary notifications were designed to encourage participants to select lower sodium food choices and included suggestion for lower sodium food choices to their commonly consumed or purchased Foods relying on the nutritionix database notifications were tailored to the time of day day of the week community and confidence in following uh lower sodium food choices now importantly both physical activity and diet notifications were tailored based on participants communities and participants received an ad mixture of community and expert generated notifications so one of the really unique aspects of this trial that the mybp my life study will be able to answer is whether having Community congruent text messages improves their efficacy so this was really great work that was led by lesie scolis who was here at the time and now is at Northwestern um but what uh Dr Scaris did is is she worked with each of these communities to generate these messages and then to identify structurally differences between the messages um so so to give you a flavor of what that looks like at the first row here a researcher expert generated message this was somebody from our study team likely that wrote this it's going to be a beautiful day started off with a walk check the app to see how many steps you got this was the second in the second column you'll see a community generated notification and this one was from the Flint community so this says ask the Lord to not only wake you up but get you up and walking in the morning morning and then finally on the right the community generated notification you're going to be and this was from the the University of Michigan Community you're going to be late hurry up you have to walk you have to take your walk today so you can see that there are differences um between each of these three notifications and one thing that Dr Scar's team really identified was the theme of religiosity that arose in the um um Hamilton Community Health Network notifications and so will'll be able to see the impact of uh Community congruent notifications on their efficacy as we start to dig into these results so here are a few screenshots from our mobile application so participants interacted with a central visualization which you can see on the left or panel a and on the left of that heart you can see how many low sodium food choices they met relative to their self- selected goal and how their step C compared to a self- selected goal they also had the ability in the subsequent panels as you can see to track their low sodium food choices relative to goals and step counts review recent messages and see how their blood pressure compared to Aha recommended target blood pressures and so um again I I'm not going to go into the results here but just to give you a flavor of what these interventions can look like um and and hopefully um in the coming months you all will be hearing about the results of this trial for myself and others on the team so turning now to the final section I wanted to talk about some of the challenges that I see in um mobile Health um Technologies and mobile health research um and these are a little bit more General than just Jedi and micro randomized trials so so I've kind of broken these into three categories um design and Del delivery interpretation and implementation so starting now um with um uh design and delivery we really lack sufficiently granular behavioral M Health theories and I think that that can become really important when you're thinking about designing and delivering Jedi and micro randomized trials the way that people interact with the intervention the time scale the frequency are clearly very different than with more traditional in-person studies and so I think having more granular behavioral mhealth theories to guide the design of these interventions is something that's going to be really important for the field moving forward we also need greater methods um greater involvement of diverse stakeholders and so certainly we need traditional stakeholders in the form of clinicians and patients providers uh caregivers um but uh you know certainly other stakeholders such as health information technology become really relevant here and finally we need more fasile methods of evaluation so micro randomized trials are one example of an efficient means of evaluating our mobile Health interventions although there's been work in other experimental designs as well as using real world evidence and Dr Matic trials to evaluate our interventions and this is really important because unlike traditional drug trials mobile Health Technologies evolve so quickly and so relying on our more traditional means of evaluation would leave the technology Antiquated at the time that the trial was complete there are also challenges with respect to interpretation so we lack longitudinal data collection in diverse populations many of you are familiar with the Mya study that conducted here that included that enrolled participants 7,000 Michigan medicine participants and followed them for three plus years and and this is one of the important gaps that my pact is trying to address um but but certainly many of our mobile health studies have been conducted over relatively short periods of time and so understanding these longitudinal changes in Mobile Health parameters ERS and informing prognostication is going to be very important we also need device validation in diverse patient populations um I am a uh heart failure cardiologist and so one of the things um that is very challenging is that our wearable devices perform poorly in heart failure patients they tend to perform poorly in patients with chronic conditions who have Lower Gate speeds um and so um again under understanding the performance of um of these devices Within These different populations is important and finally we need correlation of digital endpoints with gold standards to facilitate interpretation and I just wanted to highlight one um study that we recently published that is is trying to address this Gap although certainly much more work is needed so this was a um a retrospective analysis that we did of the chief HF study um and we actually were privileged to be a a a top recruiting site for this study but what Chief HF did for those of you who may not be familiar is it enrolled patients with heart failure patients got randomized to receive a drug canagan or Placebo and then people were followed for um nine months but during the first 12 weeks of the study everybody wore a Fitbit actually for the whole study everybody wore a Fitbit but first 12 weeks of the study everybody wore a Fitbit and answered um kccq Kansas City cardiomyopathy questionnaires on their phones nearly weekly um for those first 12 weeks of the study and for those of you who are less familiar with the kccq it is the gold standard for patient reported outcomes in the field of heart failure and kccq scores have been shown to correlate to associate with a range of important heart failure markers um and outcom including um mortality and heart failure hospitalizations amongst others so the kccq is an incredibly well validated um end point in the field of heart failure so again in the chief HF study they wore a Fitbit um for for 12 weeks and answered nearly weekly kccq scores on their phone and we looked at the association between changes in Step count from the Fitbit as well as floors climbed and changes in kccq scores the idea was we really wanted to understand what is a significant change Ina in Step count and what is actionable so we found that changes in Daily step count were significantly associated with changes in kccq total symptom and physical limitation scores over 12 weeks and importantly we found a strong Association for increases in Step count and increases in kccq scores but no association for decreases in Step count and decreases in kccq scores and while this was a nonlinear relationship we were able to quantify this so for an example step a step count increase of 2,000 steps per day was associated with a 5.2 Point increase in kccq total symptom scores and the kccq score ranges from um 0o to 100 and a clinically meaningful change in kccq score is thought to be anywhere from 3 to Five Points so 2,000 step change was associated with a clinically meaningful change in kccq scores and again this certainly we need a lot more work both in heart failure and in other cardiovascular conditions but I just wanted to provide this example to highlight work that our group was doing to help provide um a contextual framework for understanding this mobile device data so finally um there are challenges in the field of implementation so this includes to access to and coverage of mobile Health interventions there has been increasing um awareness of um mobile Health Technologies is a super social determinant of Health um and ensuring that um there is access to this technology and and understanding who will cover um this technology going forward um is an important question there are also issues of workflow and data stream integration much of this work is being conducted in parallel systems um and so to to develop sustainable interventions I think will require that these be integrated into our existing um electronic records and I think it's really important that these workflow issues are addressed upfront I think this is really important for avoiding clinician burnout but also ensuring that patients understand who is reviewing their data and at what Cadence I think some patients expect that their wearable device data is perhaps reviewed daily or weekly and and and that may not be the case and so I think level setting upfront with both clinicians and patients as to the expectations is going to be important for developing a durable model and finally there are issues of enhanced privacy protections for users um although that is certainly a talk um in and of itself so turning now to just some closing points um mobile Health technology is broadly defined and used and has the potential to collect vast amounts of data both passively but also actively using ecological momentary assessments micro randomized trials or mrts are a novel experimental design and these are used to guide the delivery of just in time adaptive interventions Jedi and mrts levered with inperson contrast through sequential randomization and are thus a temporarily and statistically efficient way of determining causal effects and there are multiple real world challenges to mobile device use in research and clinical practice also turned into a recent paper a scientific statement out of Jack that Dr gambari just authored um that talked um about many of these challenges in more detail so um these are just a number of people um of which there are many more um that have participated in this work um uh and importantly a thank you to our study participants in the Valentine and the mybp my life stud studies uh and with that I am happy to to take any questions maybe I can start um Jesse that was fantastic um I have a question about implementation if you could maybe um f us a little bit through um implementation and some challenges as far as payments are concerned and if you could maybe comment on um new uh testing strategies like arpa H and funding mechanisms that may be funding some of this work um and and can you tell us if you're planning to kind of you know test um some of your interventions in the setting of an RP type of funding um that's a great question um so I think from implementation um you know I think when I think about this work um you know I'm not sure that um you know anything we've done has quite been ready to be implemented um uh you know I think that there's been a lot of Fairly robust data that has shown that smart watches in general work for say increasing physical activity levels I I think that that's out there um but whether kind of our interventions are incrementally increasing physical activity I think it's still like an unanswered question and so um I I think we're we're a little bit Upstream at this point um in terms of of kind of thinking about implementation um I I for one I'm kind of trying to dive deeper into understanding kind of the behavioral health that guide some of this and using that to to kind of work with micro randomized trials to to guide our interventions um so and I and then to answer I I have not dug deep into that funding mechanism so I'm not not sure that I can I can answer that question yeah I was just thinking of your blood pressure treatment um interventions that would be a good fit for that yeah and the follow and the follow question I have is can you comment a little bit on Intervention fatigue and kind of approaches that you take to try to optimize engagement and maybe comment on what is ideal engagement for an advention like you Institute yeah that's a that's a really great question um so um I think engagement um can be defined in a number of different ways um H and I think that there have been different definitions of Engagement is it just participating and wearing the um um you know wearing the Smartwatch probably not um it's probably more than that and engaging with the intervention one thing that the ibp my life study is quantifying is actually interactions with the mobile application how many clicks how much time on each page things like that um in terms of thinking about engagement I think the more interesting question is how do you improve engagement and I think most of these Jedi have shown kind of early engagement um that you see typically in the first month very similar to kind of what we saw that that often decrements over time I think interestingly we saw kind of a bodal response which is a little bit um different and perhaps promising but but this kind of early uh engagement and then decrement is pretty typical of what you see in many M Health interventions in Jedis are not unique um peda clasna who's here in our school of information um and Susan Murphy who used to be here and now is at Harvard is is as you know doing a lot of pioneering work in this field of Engagement through reinforcement learning um and so we're looking forward to to collaborating with them on some of our future studies and incorporating it into our future interventions so the idea behind reinforcement learning is not that just you're looking kind of retrospectively to understand what types of notifications people respond to but actually in real time for that individual person you're you're understanding which types of notifications they're most likely to respond to um and in which context and you're delivering those in a higher dose um and you're varying the overall dose of your notifications based on um their response um so perhaps if you're finding that response is decrementing over time that maybe your algorithm is built in a way so that actually kind of gives them a break from those notifications and then reintroduces them at a lower frequency um so um Dr Dr clasna has um recently gotten a number of Grants including um an ro1 that is actually focused on this idea of reinforcement learning and understanding that and it's actually going to be in a cardiac Rehabilitation population similar to Valentine uh so um I think that that will that will be really illustrative no more questions um is Step count is activity really the best um what digital biomarker for the types of interventions that you're instituting or are there like other more interesting or more complex digital bi markers that you could use um yeah I think I think probably um I think it depends on on what your what Your outcome is and um what your intervention is designed to do you know there's a really interesting study that just came out in um circulation um last month um the watchful study I don't know if you saw that but it was kind of similar to to to us and it was designed to um increase physical activity and their overall goal um was to actually improve functional capacity and so their intervention actually increased step count without increasing functional capacity um which was their ultimate kind of goal um uh which they assessed by a six- minute walk distance and so kind of their hypothesis was that like step cow was actually you know it was increasing their physical activity but not at a high enough level to you know lead to changes in functional capacity which is what we really care about um so um I think step count may or may not be the right um outcome depending on your population and what you would like your intervention to do I think there are also potentially you know more sophisticated markers that you can use um and that we thought about so for example for us um in in trying to get these higher levels of physical activity and cardiac Rehabilitation and Roes is at one point we tried to do time in their target heart rate zone right you're getting heart rate data from your your Smartwatch and and you could probably look at that um you know it ended up being more complicated than that and and and I think physical activity through step count ended up being what was actionable but I I I I suspect that there probably are um uh more interesting um and and probably better suited digital biomarkers yeah I ask you because oftentimes in clinical care I um what what I hear from patients is that what I was able to do I can do less vigorously or not I'm not as quick doing things that I want to do so I want to me it seems like somehow you know if you like the speed at which you were performing your daily activities that you were previously doing seems like an interesting biomarker do you do you get um raw 3D accelerometer data in your studies or is just mostly aggregate step count to you're getting yeah so we get um so we do get you know for for fit we get like the intraday file so so it's more than just kind of the daily aggregate we get kind of bout level physical activity minute to- minute heart rate or every you know five minutes whatever the frequency is so we do get the daily data we we don't quite get the you know the raw accelerometer data and so I think one thing that we have been thinking about a lot in our group is whether we should use a kind of third device for example like an actograph and give the same device right to everybody right the first week of the study the last week of your study both your control and your intervention groups um both get um you know the actograph in addition to whatever Smartwatch they're getting as part of the study um and and I think that that's something that we probably do need to do in the future right so you know one thing I I didn't talk about and probably should have should have talked about was right we analyzed our Apple watch and our Fitbit separately for the micro randomized trial and that's because there are measurement differences between the devices step count as measured by an Apple Watch is not the same as step count measured by a Fitbit it's probably small the differences but you're comparing apples to oranges or you know maybe different like types of apples or something and so we chose to to measure them separately um and and there are limitations to that and so you know we didn't do it at the time in retrospect I probably wish we had done it but they give an actor accelerometer and then we could have analyzed potentially at least for those distal outcomes at at Baseline and in six months um th those group using the same scale yeah it's super interesting you can imagine looking at periods of rest within you know two people can have 2,000 steps that they've taken but if they're taking breaks every five minutes or every 500 steps for example or or the response of hard rate response to the same level of activity yeah really interesting work that can be done there it's super interesting yeah and I I I think you know to your point you could think about things like chronotropic competence for example um in and how assessing assessing you know using kind of these more robust digital biomarkers than just putting somebody on a treadmill and and assessing that but um we haven't gotten there yet very interesting lots of interesting and novel things that can be done there um it looks like we have a question in the Q&A uh from Katherine barnier um and she she asked um given this that some participants mentioned that they would have valued a semi competitive Community I'm curious to know whether any of the data you collected might allow you to learn anything about the potential influence of existing social communities on compliance for example our participants who typically Walk With Friends rather than walking Al alone more likely to follow suggestions from prompts alternatively in the upcoming study you mentioned in the cardiovascular Rehabilitation setting are there any plans to have a kind of control group that doesn't have the same social support to see if it influences compliance um yeah that's a great great question um so I I'll get at that kind of for for from two different ways um so so unfortunately kind of the studies that we have going on are really allowing us to assess that question um it's more kind of an area for for future research um we do have one study that is randomizing people to iners versus tella Health cardiac Rehabilitation um and inperson and teleah health cardiac Rehabilitation are actually very similar um with kind of the um exception being social support excuse me um and so one of our hypothesis is that if these two are not equivalent that then maybe social support is what factors into that I do also have a k23 grant um that is looking at using those ecological momentary assessments short questionnaires understand when with whom and how patients with heart failure are physically active and so we'll be asking them repeatedly over the course of multiple days multiple weeks um who um um they are being Physically Active with to understand that social support and we're going to be correlating their answers on those ecological momentary assessments with objectively measured physical activity so some of our studies will start to kind of get at that question of social support but unfortunately the data we have right now is is is probably not going to answer that question well um thank you so much Jesse for a fantastic talk if you guys have any questions um you can either send it to us or email uh Jesse directly and we look forward to hosting you again and hearing about the follow up on all your studies thanks again great yeah thank you so much for having me and and and thank you for being here today I appreciate it thank you so much bye

2024-02-13 00:35

Other news