Preprint Smartphone-based monitoring of heart rate variability and resting heart rate predicts variability in symptom exacerbations..., 2024, Aitken+

Abstract:

Background: Complex chronic conditions like Long COVID and Myalgic Encephalomyelitis/Chronic Fatigue Syndrome involve energy limitations and changes in heart rate variability (HRV) and resting heart rate (HR). Mobile health technologies now offer real-time, valid measurements of HRV and HR, advancing symptom monitoring and management. Using a high-density dataset from an observational longitudinal study, we aimed to describe, quantify, and predict within-person co-variations in daily biometric data and subsequent crash, fatigue, and brain fog symptom occurrences.

Methods: Leveraging data collected through a mobile health app (n=4,244), we developed predictive models using mixed-effects linear regression and logistic regression to explore how within-person fluctuations in biometrics (HR, HRV, and respiratory rate) predict dynamic change in symptomology (crash, fatigue, and brain fog). Predictive performance was assessed using 5-fold stratified cross-validation and compared to a 20% holdout set to evaluate model generalizability to new observations and individuals.

Results: Across all symptom domains, within-person changes in HRV and HR consistently emerged as key predictors of symptom change across all models, with higher HR and lower HRV conferring risk for crashes, fatigue, and brain fog. Moreover, 7-day biometric stability (or variable dispersion) was a robust predictor of symptom occurrence and severity. Models trained solely on biometric features achieved moderate predictive performance in the stratified cross-validation set; however, incorporating random effects to capture individual-specific variations and prior-day symptom reports substantially enhanced model accuracy, with AUC values reaching .91.

Discussion and Conclusion: This study is the first to use data-driven models to predict everyday symptom experiences in individuals with complex chronic illnesses based on biometric fluctuations. Findings demonstrate the potential utility of mobile health tools for real-time monitoring of symptoms and highlight the need for further research to refine these predictive models and integrate them into clinical decision-making processes.

Link (ResearchSquare preprint, open access)

 

The major limitation seems to be that these were self-reported diagnoses - those with LC were apparently asked to confirm if they met WHO criteria for LC, but this is a very broad definition, there were no specific criteria for ME/CFS, and the sample contained "individuals with other energy-limiting conditions":

Nonetheless, very interesting to see research using smartphone-derived data.

 

I have been doing HRV monitoring for about a year now, on and off but with months-long periods of sustained readings. I have never really noticed any benefit from a high HRV. Mine is typically in the 55-65 range, which is excellent. I have had many worse days that weren't proper crashes, and it rarely dips, so it's not very sensitive.

It does, however, really pick up acute illnesses. Which I'm not sure is especially useful as the times I saw a significant dip, I definitely knew I was ill and did not need someone telling me. Same with the few days following a vaccine. And drinking alcohol.

So it can probably be of some use, but as a very adjunct thing that is more like having a flag to show if it's windy, but one that is a bit too heavy so that it only tells you when it's so windy outside that you can't miss it because cows are close to flying. Which has very limited uses.

I'm actually surprised at how consistently good my HRV can be, despite being essentially non-functional, even for periods lasting weeks and months. Which tells me that a high HRV means very little, only a low one.

 

My Fitbit also gave me scarily low HRV readings, so I looked on their community forums and saw lots of other people saying theirs were also giving scarily low readings compared to what their other devices were telling them. I think it's just a fitbit 'thing' and probably the key is to look at trends over time rather than the numbers.

 

Note the non random effects model had very poor predictivity - AUC of 0.6-0.61 for crashes or brain fog.

This means in the real world, HRV and HR monitoring is not terribly useful because most people will not have a perfectly fit model, and even if they did, that doesn't tell them much about the specific triggers nor any treatments that prevents the response.

 

I used visible (which is where the data for this study comes from) for about 4 months and I think there’s two problems with a study like this:

1) Expectancy Bias
Visible is literally based around the idea your HRV impacts your PEM threshold. They give you specific scores every day based on your HRV, so someone that got a “bad” HRV score in the morning might subconsciously be likely to report worse symptoms in the end of day questionnaire, for the simple reason they got a bad score.

2) Self-Selection Bias
For people like me, HRV didn’t seem to predict much, so I stopped using the app. The app will likely select for long term users who believe HRV impacts their symptoms.

For a study like this to work, since it relies on PROMs, HRV data would have to be blinded.

 

The pattern is more complicated for me and it's about degree and pattern of change not absolutes or high being good low being bad.

I have noticed my hrv actually shoots up just as I am entering pem before then tanking and gradually working it's way up to mid range as pem slowly subsides. So for me either unusually high or low in range scores are a bad sign. It's run at similar absolute scores with exactly the same pem indicator pattern on two different brand devices now (polar then garmin) over several years.

Two pem episodes are underlined in pink here and the red and orange is the impact of the Covid infection I recently had.

 

I like visible, people seem to really fixate on their daily score and it’s a bit prominent so you can’t have see why.

Then there is all this next step they want to do - predict a crash. I absolutely don’t see the point of pursuing that. Sometimes there’s a trauma. Sometimes there’s no discernible reason. I don’t believe that a crash is as simple as HRV + past performance = cancel your plans for the next 3 days. However, people want to believe it is and they want to be able to make plans, I guess that idea is enchanting.

I don’t think we know enough about ME/CFS or LC to be predicting how it behaves in future. If I’m wrong I’ll eat my words but I think it’s a fools errand. It’s like when people follow the car map thing and drive into a river because it said to, ignoring what their eyes and ears see, in favour of the computer.

 

Garner's understanding of classical conditioning seems to be that the dog's salivating causes the food to arrive.

 

I agree, I'd much rather people learn to listen to their body than overly focus on heart rate and HRV.

 

I think it would still be a subjective thing - you’re asking people with cognitive dysfunction to self report their cognitive dysfunction and link it to their heart rate variability? But I’m not a scientist.

Based on absolutely no evidence and just my opinion (!) I don’t think it’s a straightforward quid pro quo I think it’s more complex. I feel like you’d be trying to work out the way a supermarket logistics picking system works by following the journey of a pallet of oranges.

The giant sticking point in any female study - my heart rate has a crazy boost about a week before my period, then drops back down over a few days, so my HRV will change due to that. I will also be entering the “the back of my head has opened and my brain has rolled out” phase of PMT.

I absolutely love Visible for many things, but I really can’t be bothered to let it dominate my life, I tried the Facebook group but it was awful, more like when I’ve done weight watchers, people obsessed about points, what they tried to do not to use points, how they accidentally burned their points….I generally just live my life and check it at the end of the day tbh, use it to show other people what sitting up does to my heart, do my FUNCAP and daily symptoms.

 

Visible is not available in NZ. I monitor HRV with the app Welltory (or through the Apple Health App) for a couple of years. Initially it was to make sense of my OI and OH and being stuck in PEM for the longest period so far but in the end it was more useful for stress management when needed as is tied in with the breathing app on my Apple Watch. Welltory, annoyingly keeps telling me to breath, meditate and exercise when a nice rest works better so the later makes it not suitable for pwME, imho.

Welltory has moved recently to a Wellness score which compares your HRV to previous readings and to population norms (but that could be the population using the app). I am comfortably in the population norm. It did predict a nasty URTI I got but while I was in PEM following that it did not show that. I have a few brief periods of PEM since and it is debatable if it has predicted it. Maybe Visible has different metrics.

I would definitely not use an HRV monitor to predict a crash or to advise on whether to exercise in the next heart rate threshold, I have apps specially for that, just for academic interest, and the scores rarely reflect the reality of my ME and PEM. To a degree, heart rate is more reflective of my body's state than HRV. I do still use Welltory as one of may biggest problems is sleep and I find the graphics useful which includes a body battery.

But health apps, although interesting, I do not think should replace learning to recognise one's own symptoms and how they relate to PEM, basically like other's have said listening to your body.

I have both OI and cognitive problems and my average HRV is 20ms, it has gone down from about 30 a few years ago. I have had a few readings of 0 which didn't reflect my assessment of my body's state at all. A poll would be interesting for academic interest @SNT Gatchaman, maybe linked to severity with a FUNCAP score comparing different dimensions of symptoms.

 

Been wearing a Garmin with HRV monitoring for a couple of years. My HRV reliably drops when I've had a stressful day or consume alcohol. Haven't noticed any correlation to crashes. N=1 though.

 

To be fair it’s my data they used in this study, I’ve just realised.

 

HRV is typically shown to be lower during stress, so the idea that people have been using is that low HRV is predictive of feeling worse.

We are looking to do some similar work here but a little different than single morning HRV measures. It's encouraging to see that it was accurate. The recall numbers I couldn't see but I think they aren't as impressive.

Recall is most important though, that is telling you how often you could predict a crash that actually happened. For ME/CFS you really want 100% recall, you want to predict a crash everytime they have a crash. That way you could give warning a crash was coming. A simple way to do this would be to tell someone evey day that they will have a crash, but obviously that's also not helpful. You want to get every crash day right with the most amount of predicted non-crash days.