WHOOP Podcast - The 4th pillar of recovery: WHOOP adds respiratory rate to recovery algorithm

Episode Date: July 29, 2020

Kristen Holmes and Emily Capodilupo are back to detail a critical update to the WHOOP recovery algorithm. We have long calculated recovery by tracking heart rate variability, resting heart rate, and sleep performance. Now, we are excited to add respiratory rate as the fourth pillar of our recovery score. This powerful update reflects the latest research in health monitoring and provides members with an even deeper understanding of their bodies. Kristen and Emily discuss our COVID-19 research (2:17), the origins of the WHOOP recovery score (3:48), the science behind heart rate variability, resting heart rate, and sleep performance (7:57), adding respiratory rate to recovery (12:05), the factors that can alter respiratory rate (16:59), why an increased respiratory can predict a decrease in performance (21:39), how COVID-19 can change WHOOP data (25:32), tracking your baselines (28:17), and the importance of third-party validation (30:46).Support the showFollow WHOOP: www.whoop.com Trial WHOOP for Free Instagram TikTok YouTube X Facebook LinkedIn Follow Will Ahmed: Instagram X LinkedIn Follow Kristen Holmes: Instagram LinkedIn Follow Emily Capodilupo: LinkedIn

What's up, folks? Welcome to the WOOP podcast. I'm your host, Will Amit, founder and CEO of WOOP, and we build wearable technology to improve you. That's right. We measure strain and recovery and sleep. More recently, we've highlighted respiratory rate and all the research we are doing around COVID-19. You can check that out on previous episodes. We also just announced our partnership with the NFLPA to help NFL players go back to work safely. We're very proud of our relationship with the NFL and its players, and we wish them enormous health and safety as they get back to play. We have a great podcast today, a science-e podcast today. Kristen Holmes and Emily Capitolupa, your two favorite podcasters, are back to detail an update to the Whoop Recovery

algorithm. Now, let's be clear, we don't make changes to the whoop recovery algorithm practically ever, and we think that we have a responsibility when we do to be very public about them. In the past six months, all the research that we've done on respiratory rate and COVID-19 has shown us that respiratory rate is a deeply important metric and a metric that should in fact be included in the Woop Recovery algorithm. So Kristen and Emily discuss that. They talk about our findings with respiratory rate to date, how we calculate recovery and why respiratory rate will help give you an even more accurate score.

Why this mostly affects outlier data points rather than all recovery scores. That's an interesting nuance. And they also get into tips and tricks for getting better recoveries. A reminder as well, you can get 15% off. a WOOP membership using the code Will Ombett. That's WI-L-L-A-H-M-E-D. Without further ado, here are Kristen and Emily. Hello, everybody, Kristen Holmes, VP of Performance here at Woop, and I am joined by Emily Capital Lupo, our vice president of data science and research. Hey, Kristen. Before we launched into today's podcast, wanted to make sure to highlight some of Emily's, most recent research and papers, both extremely relevant to these times. And we've talked about them

previously, but I just wanted to make sure everyone was aware. They're in peer review right now, but the preprints are available on whoop.com. And both papers provide core insight into, I think, two really key concepts that could not be more relevant in today's time. So Woop data during the pandemic and COVID detection. It's just extraordinarily groundbreaking work. And I think just points to really the Herculean work that Emily is doing behind the scenes to keep us at the absolute forefront of innovation. So Emily just wanted to say congratulations on these two papers and just for all the work that you've put into COVID and just making sense of these crazy times. Today's pod, Emily and I are going to talk about the rationale behind a recent update to Woop Recovery and what that change will mean

for you. And we're also going to spend some time talking about core behaviors that are strongly associated with recovery and general health. So, Emily, I thought before we get into the update to recovery, I thought we could just do a quick overview of recovery in general. We have a ton of new members on the platform over the last few months. So I think it won't hurt to be redundant here. So if you just want to do the overview, and then I'll talk a little bit about utility in the field. And then we'll launch into the new update. Sure. So the goal of the recovery score is to measure how ready your body is to adapt to a training stimulus. This is highly correlated with how ready you are to perform. It was originally built to predict athletic performance. But over years of research,

we've discovered that readiness to perform is actually completely independent of the type of performance that you're going to have. So we've seen that the things that predict athletic performance, predict academic performance, they predict performance in the military, predict performance for surgeons, things like medical errors, and also just sharpness. So we've heard anecdotally from CEOs that they feel like they lead better meetings when they have a green recovery score than a red recovery score. And then obviously, you know, dozens and dozens of studies, anecdotes about, you know, all kinds of athletic performance from, you know, weightlifting, Tarendarren sports to team sports.

Totally. And you mentioned just the CEO thing because I think, you know, now that we're, you know, we've moved beyond just a tool that is useful for athletes, but a tool that's useful for really anyone who's interested in elevating performance and just interested in performance levels in general. And we did see in a case study a correlation between recovery and executive function. So that sharpness that you mentioned, we definitely have seen that in some of the research. And we're actually in progress. We're doing a study right now with 100 of the top CEOs in the world. So stay tuned for that research when it comes out. But we're looking at heart and variability, sleep, and executive functioning. So that's going to be a cool one.

Yeah, and I think that sort of this like all-form performance being related to recovery really speaks to the fact that, you know, the recovery score is looking at something holistic. And so what is actually primarily looking at is your autonomic nervous system balance and, you know, at your core, how ready is your body to respond to different stimuli? and that readiness can be applied to the perfect tennis swing. It can be applied to reading someone else's emotions and making an appropriate response. It can be applied to like, you just creativity, all these different things. And so how we then channel that readiness to perform

is very situationally dependent, but that physiological state of being adaptive is sort of like one singular state that can then be applied to a whole bunch of different scenarios. It's been really interesting to see that, like, as we started to branch out, you know, beyond athletes, which was, like, sole focus for the first couple of years at loop and into sort of more, like, performance-oriented individuals, but where, you know, performance had a much wider definition, that we actually didn't change the recovery score at all to work for those

scenarios. It just worked. And so that sort of realization that we were picking up on something, just so much. much more fundamental than something that just purely underlies athletic performance was really cool. Totally. And Emily, you mentioned the word holistic. And I think that's really important. And we've seen this in the research that we did with Corey Stringer Institute that no one of the single metrics alone that make up wood recovery. So hearty variability, resting heart rate and sleep are independently more correlated to performance and outcomes than they are together.

So the fact that, you know, this algorithm that you've built is predictive, a performance is quite incredible. But if you want to just talk for a second about just the combination of hearty variability, resting heart, and sleep and kind of how you think about those working together to kind of give you this holistic picture of next day capacity. Sure. So as you mentioned, the three original inputs are heart rate variability, resting heart rate, and sleep. And when we were originally putting the algorithm together, it was just HRV. That's sort of where we started. This version never, you know, went out the door. But when we were beginning the research, it was just heart rate variability.

Because there's tons and tons of research that we were able to leverage that shows how heart rate variability correlates with next day athletic performance. And at first, we were just looking at heart rate variability. and when we looked at resting heart rate, they sort of were often extremely highly correlated. But then there's a little bit of literature when we started to see in our athletes that occasionally they would trend together instead of their normal pattern of trending opposite.

And when they trend together, there's actually a lot of information in that. So it's somewhat rare, but we actually see there's a phenomenon called parasympathetic saturation where heart rate variability goes down and resting heart rate also goes down. and that can be a sign of a really effective taper in well-trained endurance athletes.

And so we incorporated resting heart rate into the algorithm in order to contextualize some of what's going on with heart rate variability because in a case of parasympathetic saturation, you would misinterpret the meaning of heart rate variability going down if you weren't also looking at resting heart rate. And then we also added in sleep. It gets very little weight in the recovery. score, it doesn't really influence the algorithm all too much, but it can be a really interesting explanatory variable in what's going on. So for example, if you're not getting enough sleep,

you would expect to see your resting heart rate should be higher, your heart rate variability should be lower, and when you get more sleep, you'd expect the opposite, right, that those metrics would reflect that you're well-recovered, sort of the way you feel more rested, that feeling is usually real. If you're not getting enough sleep, you're not going to be recovered. If you're getting enough sleep, then that's one of the biggest behaviors you can control that influence recovery. But if you're getting tons and tons of sleep and you're still seeing like poor heart rate variability and resting heart rate, then obviously sleep isn't the problem. It's usually an indicator that something else is going on. Either you're very overtrained. Oftentimes it means

you're sick. And so that can be like a really interesting piece of information to have when you're trying to contextualize the changes in heart rate variability and resting heart rate. Emily and I've talked a bit about just oversleeping too. You know, that can also, you know, if you're spending actually too much time in bed, that can also have a negative impact on kind of heart variability and resting heart rate. And I think that's where WOOP can be super helpful, right? Because it is highly individualistic, right? It's going to depend on, you know, just the person's day.

Like, for example, athletes in preseason, or, you know, if you're in a really hard training block, you know, training for a triathlon, for example, you will need more sleep. So extending that sleep by half an hour in each end is going to be probably really helpful. And that's where Woop can kind of help you dial into Emily's earlier point. If you notice that, you know, you are spending a typical time in bed for you, but yet you're not recovering from, you don't seem to be recovering by looking at heart variability, rest in a positive way, or your Rup recovery is kind of oddly suppressed, that might mean that you need to extend your sleep a bit or reduce your volume and intensity in terms of your training. So it depends on what your intent is physiologically,

but, you know, this is where, you know, you can use this data to really hone in on your optimal set points. And because it is highly, you know, circumstantial and, you know, dependent on what your intent is physiologically and, you know, all the things you have going on in your life, This is where the data can really ground you and help you understand or help you inform, you know, some of the decisions that are happening throughout the day. So you can, you know, be positioned to kind of optimally peak. All right, good. So I think, Emily, this is a great segue to talk about the new input. We are adding respiratory rate.

So would love to hear just the rationale behind adding respiratory rate to whoop recovery. Yeah, so those of you who have been following along with WOOP for the past couple months know that we've been re-examining respiratory rate in all kinds of ways. We've shared research about how in cases of many, many members who have unfortunately contracted COVID-19, that their respiratory rate spiking tends to be an early indicator that they've been infected. And so we've been taking a look at respiratory rate, which was, to be honest, is a metric that we had largely ignored because when we looked at it night after night, it really doesn't do anything. We actually, in a paper that we submitted a couple weeks ago for peer review and is currently in the peer review process, we showed that the average user's 30-day variability in respiratory rate is less than half. of one respiration per minute. So it's really tiny. We were actually the first ones to ever publish on what normal night-to-night respiratory rate in healthy individuals looks like. There is no pre-existing literature, which is another reason why it was sort of something that we hadn't paid

much attention to. It's never been studied in this context as a predictor of next day recovery. And so, but we were looking at it sort of in the context of COVID-19 and sort of discovered this pattern that in healthy people, it varies almost not at all over the course of a month, less than half of a breath per minute. And then all of a sudden we see, you know, when people were getting COVID-19, that it was spiking up sometimes three, four, five breaths per minute, which is a huge increase. So then we were sort of curious, like, well, you know, outside of COVID-19 is respiratory rate predictive of next day performance. And so we went back to, you know, prior to the pandemic, so data that we had collected in the fall. And we looked at what the recovery score was doing and how accurately it was predicting next day performance. But then we can also go and measure next day performance. And the difference between what we measure, the real performance, and what we predicted is the part of performance not explained by the current recovery score.

And we noticed that in cases where that difference was large, respiratory rate was often high. And so what that suggests is that respiratory rate contains information that's useful in predicting next day performance, but that is independent. of the information that we're already capturing in resting heart rate, heart rate variability, and sleep. And so by incorporating it as a fourth metric in the recovery score, we're able to have our predicted performance and actual performance match up more closely. So you can think about that as increasing the accuracy

of the recovery score or the utility of the recovery score. And so we use several million days worth of data across many, many different individuals, different people who are performing into very different ways in order to come up with the modification to the recovery score that includes respiratory rate. So, Emily, I think you make obviously a really important point that we're adding to the accuracy of WOOP recovery,

which is phenomenal. I think it's important to point out that heart rate variability and resting heart rate being kind of nonspecific in that there's a lot of different things that can influence your changes in hearty variability and changes in resting heart rate. So can you just kind of distinguish between respiratory rate and, you know, as being highly influenced by something specific as opposed to heart variability, resting heart rate being kind of these more non-specific factors in how that kind of contributes to this accuracy piece?

With heart rate variability, it's responding to sort of how available your audit nervous system is to respond to stimuli. And so it's going to change, you know, whether you overtrained or if you're sick or if you're stressed or all these different things, right? So it's the fact that your heart rate variability has gone up or down in itself doesn't tell you too much about what changed. Like there's this sort of layer of the human needs to kind come in and, you know, add some context with respiratory rate, the prime, so two things that primarily will make your respiratory rate change. And obviously we're measuring respiratory rate when we are talking about it for recovery. We're measuring it during sleep. And so we're talking about

at rest. So it's definitely true that if you're, say, exercising, we've all felt this, right? You start like running and you're running harder. You start breathing harder. So your respiratory rate might increase as your activity level increases because your body's need for oxygen has increased. That's not what we're talking about here. We're talking about respiratory rate at rest. And so it's a roughly fair assumption that night to night, how much oxygen you need during sleep, should be sort of roughly constant. And so given that each breath sort of in a normal situation should be,

be delivering the same amount of oxygen, your respiratory rate could increase because, say, air quality changes. So if you go up at altitude, the oxygen content of the air is lower. And so our respiratory rate will increase because the density of oxygen or the amount of oxygen we're getting per breath. Because really, it's not like our body needs breaths. It needs oxygen. And so if each breath is delivering less oxygen, you'll need more breath to get the same amount of oxygen. And so we see at altitude that like people's respiratory rates increase. That's nothing to be concerned about, but the lower air quality does mean that like you're going to have to work harder to provide oxygen, which is a super basic thing, which means

you're going to have fewer resources available to do things like exercise. So the decrease in recovery at altitude is in fact real, even though you like quote unquote didn't do anything wrong. You're not sick. You didn't over-trained. It's just the reality of your environment. A funny quick anecdote on just the air quality. I was talking to a baseball player yesterday. And in quarantine, he was on a farm and it was really, really dusty. And he noticed these changes, you know, pretty significant changes in respiratory rates. So he thought it might be COVID related. But I think it was more just the air quality, just being exposed to all that dust. Yeah. No. And that, I mean, that would definitely drive up your respiratory rate. Because if, again, like we need to get a certain amount of oxygen. to like power baseline fuel needs. And so, you know, things like thinner air at altitude or dustier air in a dusty farm environment could drive your respiratory rate up.

And again, like that decrease in recovery would be real even though it's caused by something external like the environment instead of something internal, potentially like COVID. So to talk about, I guess, like the other side, so the internal stuff would be that for some reason your lungs are not working as efficiently. So that's actually what's going

on with COVID-19 or with any lower respiratory tract infection that, you know, your lungs are actually damaged. So COVID-19 infects the cells where gas exchange happens between, you know, the air and your lungs and then the blood right outside of your lungs. And so, you know, you can think about your lungs as being like, you know, a drive-thru restaurant. And so it's like the little bloods come up to the window and they receive their little oxygens and then they drive along. And so those windows are damaged.

And so the blood's kind of come by and because those cells are damaged, gas doesn't cross them. So oxygen doesn't come into the blood, CO2 doesn't come, or carbon dioxide doesn't come out of the blood in as many of those little windows. And so because you now have like fewer windows, even though your lungs

are going to bring oxygen into them, you don't get that gas exchange. And so what your body sort of experiences that you know, a smaller breath essentially occurred. And so you're going to need more breaths because each breath is less effective at this gas exchange. So anything that's going to damage your lungs, obviously, is also going to impair recovery. And so we don't know in, you know, these millions of, you know, days of data that we analyzed in order to make this update to the recovery

score. We don't know why respiratory rate was elevated. We didn't separate out, you know, did you have a lower respiratory tract infection? You know, was it dusty? Were you at altitude? Potentially allergies are another thing that can impact the efficiency of gas exchange. We don't know which one it was. We didn't attempt to clean the data that way. But we do know that when respiratory rate was elevated, that next day performance was decreased.

And so by modeling that, again, even having nothing to do with COVID-19, we were able to increase the utility of the recovery score. That is phenomenal. Love that. So what can users expect to see inside the app? Users will likely notice almost no change or really should most users will notice no change in their recovery score itself to reflect the fact that respiratory rate is now, you know, being considered in the recovery score.

we did move the respiratory rate stat from the sleep page where it used to live into the recovery page. So it's now located in a slightly more sensible place. It's the exact same information. It's the same respiratory rate. So it's the median respiratory rate throughout your entire night of sleep. But again, instead of seeing it in the app when you click on hours of sleep and you see it, you know, between efficiency and sleep latency. Now you're going to see it when you click recovery alongside heart rate variability, resting heart rate and sleep.

I'd love to just talk for a second about how folks can interpret respiratory rate. So when they do look at it in isolation, what are the things that they should be looking for, Emily, as it relates specifically to COVID? And then when should they be concerned? And I know we've kind of talked about this on other podcasts and lock, or post, but we might as well just outline it here again for folks who are listening. Yeah. So it's not as simple for COVID-19 in particular as saying like when your respiratory rate is up by, you know, one beat or whatever that that means that you have COVID. That's definitely

not a true statement. What we see with COVID and one thing that's particularly interesting about COVID and why it was so successful at becoming a pandemic is because it has this long incubation period, you know, anywhere from two to 14 days. And during that period, you're infectious. So as soon as you get infected, the virus starts, for lack of a better word, like setting up shop in your lungs. And the reason why that incubation period is asymptomatic is that our lungs are very, very good at compensating. And so we're not experiencing symptoms, but real damage is being done in the lungs. And so as, as that, like damage starts to pick up a little bit, you start to see that, you know, ever so slightly

respiratory rate is increasing during this period. And, you know, at some point it hits like a little bit of a like critical level where our bodies can kind of like no longer hide the damage from us. And that's when symptoms start, you know, when you'd be aware that you're fatigued and having shortness of breath and all that's coming from, you know, you're going to be fatigued because your lungs are working a lot higher. You're going to have shortness of breath because, you know, your gas exchange is less efficient. And obviously all the other symptoms, you know,

as your immune system starts to respond, you can get a fever and all of that. But during that period, respiratory rate is actually often subtly climbing. And so what we're specifically looking at sort of in the context of COVID, and we've submitted a paper for peer review that actually, like completely spells out this algorithm. them. We decided to be completely transparent about it because if we're going to beat COVID, we're going to do it by not, you know, hiding information from each other. So you can go and look at

exactly how we model respiratory rate in order to predict COVID and how well we're able to do that. But you see it's a subtle rise and then right at two days, sometimes one day before symptom onset, you see that subtle rise turns and you get this kind of hockey stick growth. It's subtle, subtle, subtle, and then it spikes up really high. Sometimes the spike takes two days. You'll see like a little spike and then a big spike. It'll jump up one or two beats and then the next day it's up four beats above baseline. And then that actually doesn't last super long in most of the data we've seen in Woot members who have graciously volunteered to share with us their COVID-19 status. Often the respiratory rate comes back down a little bit. And I don't fully understand why that

happens but right at symptom onset you see it like kind of spikes up and it's very high 30 40% above baseline kind of stays there for two three four days and then it comes down a little bit not all the way to baseline but not quite so spiky and so what we're seeing is very different than what we see when somebody goes to altitude where it might jump up but it doesn't have this like slow there's no incubation period of going to altitude so you don't see that like subtle change and like the week before symptom onset, it happens all of a sudden, and then you sort of plateau at this new level because you're in Colorado skiing, good for you. So, you know, you're at like a constant thing. So you don't see that same like subtle increase and then slow decrease pattern.

So you might reach a higher level, but just the pattern of it relative to baseline looks really different. And I think that, you know, not to get too off topic, but that's one of the things that I really like or I think was kind of clever about the approach we took to monitoring respiratory rate as a predictor of COVID-19 in that paper because we were able to leverage the fact that WOOP is measuring this every night. And so to leverage all that data from baseline and understand respiratory rate in the context of an individual user's baseline. So we're definitely not saying something that like, oh, when your respiratory rate

equals 28 respirations per minute, that means COVID, because that would be a totally nonsense statement. That's kind of what hospitals are doing because they don't have your baseline. So they sort of being tachypneic, which just means an elevated respiratory rate, they define as being, you know, above 20 or in some cases above 30 respirations per minute, and they just draw this hard line. But what we're seeing in our data is actually it's much more complicated than that. It has to do much more with like how it's changing relative to a personalized base.

line. Right. Yeah, I can see the emergency department setting it being not actually the best metric. I think that's where really all of our data, I think, becomes so relevant is that we've got these, you have a robust baseline that you can compare, you know, what's happening today against. And that gives you the context that you need to understand how these changes might actually be influencing your performance on a daily basis. Yeah, and I should add to that just because you brought it that this change that we made to the recovery score is not at all meant to predict COVID. The recovery score change actually specifically was only developed using data from the fall, so long before COVID infected any of our members.

And so don't confuse, you know, a low recovery score with whoop saying that you have COVID-19. That's absolutely not the intent and not what's going on. Yeah, I think that's a really important point. Woop recovery, as we've mentioned, you know, it's those four now, the four kind of inputs together that are really most important and most relevant. But as it really specific to COVID, people are going to want to pull out and look at respiratory rate in isolation. Yeah, just keeping in mind, right, that red recovery scores existed long before COVID

and, you know, we'll continue to exist long after COVID. So, well, we are seeing that, you know, obviously people who are unfortunately experiencing COVID-19, like, their recovery score will go down. But, you know, all the other things that used to make you red will still make you red. And so, you know, that's don't jump immediately to a COVID conclusion. And, you know, it's important to try and understand all the factors. Well, I think, you know, when we consider just our mission here at Woop, it's to help you understand. your body better. In order to arrive at that, accuracy is absolutely critical. So I just think the innovation that's happening behind the scenes and the fact that we just keep pushing to make our

product, you know, more and more accurate, I think says a lot about just how we think is a company. And the experience, most importantly, we're trying to provide our members, you know, which is one where, you know, if they are interested in calibrating their lives to a degree around these data points, we want to make sure those data points are as accurate as possible. So, you know, that's the whole kind of intent behind any update that we push, frankly, is the goal of making your experience better. The validation efforts kind of outside our four walls at Woop have been, I think, really important and powerful and have absolutely separate us from, you know, other wearables on the market, frankly. And one of those things is just how well we do sleep.

So I think it's worth noting, if members aren't aware of this, that we did a robust third-party validation. University of Arizona was the institution who led this validation effort. It was a sleep validation, but within that, we also were able to validate respiratory rate, and it was within one breath from end of the gold standard. And we are literally the only wearable on the market who has respiratory rate validated by a third party. So I think, again, just this mission that we really have at WOOP to ensure that we are providing our members with the most accurate, most robust data is something that we take very seriously. And, you know, we're going to keep pushing the envelope and continue to ensure that, you know, all of our metrics are validated outside of our four walls. It's really important to us.

But, yeah, I think that that is an important point as you're kind of searching for a wearable or instant. that you know you can really count on the accuracy of. Emily, thank you so much providing all of that insight. I think that this update is going to just provide another layer of accuracy, so folks can just be even more dialed in on their training and how they need to build their behaviors to optimize recovery. So thanks so much for all the insight and the good work kind of putting this new update together. update together.

It was my pleasure. Folks, have an awesome rest of the week, and we'll see you again soon. Thank you to Kristen and Emily, as always deeply insightful. I hope you all watch your recovery scores carefully during this time and your respiratory rates carefully during this time. A reminder, you can get 15% off a WOOP membership if you use the code, Will Ahmed. That's W-I-L-L-A-H-M-E-D. You can follow us on social at Woop at Will Ahmed.

We'd love to hear from you. Stay healthy, stay green, and keep that respiratory rate flat. Thank you.