Blog post by Marco Altini
One of the most difficult aspects to understand when getting started with HRV is measurement repeatability, meaning that measuring your HRV twice in a row will most likely result in two slightly different values.
At the end of this post I will give some tips and explanations related to this variability and how HRV4Training interprets your data so that you can still make sense of it. However, for the main part of this post, I simply want to show some data and answer a few questions, hoping to provide some clarity around the topic:
The data shown in this post were collected during a clinical validation study ran by Ben Scott, Daniel Plews and Paul Laursen. We are busy writing up a paper comparing ECG data to a Polar H7 and PPG data, but since we collected data under many different conditions, I will be using the same dataset to explore other relations, for example the one covered here between measurement duration, paced breathing and measurement repeatability.
The subset of the dataset used for this analysis consists of recordings from 26 participants and three sensor modalities (ECG, H7, camera). For each participant we have four measurement conditions, each one 5 minutes long: LB1: lying down with paced breathing, LNB1: lying down without paced breathing, SB1: sitting with paced breathing and SNB1: sitting without paced breathing.
I computed HRV (the usual rMSSD in ms) for each condition. In particular, I used the first 4 minutes of data to analyze differences in minutes 1 and 3 as well as between minutes 1-2 and 3-4, so between each repeated measure I left 1 minute. This is a bit of a random choice but I figured it was a more realistic setup to leave some time between consecutive measurements to analyze.
What differences should I expect when measuring twice in a row?
Here are boxplots of all the data showing differences in rMSSD on consecutive measurements during different conditions. Note that for the analysis that we are doing in this post the sensor modality (ECG, chest strap or camera) is irrelevant, as consecutive measurements differ due to the nature of the ongoing physiological measurements that we are trying to capture via HRV analysis, and these changes are always present. Nevertheless, we will be showing all sensor modalities since we collected the data and we can see if things are consistent or not:
Results are quite similar across conditions. Differences in rMSSD between consecutive measurement average at 15.3 ms (15.0 ms for ECG data, 15.2 ms for camera data, and 15.7 ms for H7 data, nothing noteworthy).
This is pretty much what you should expect when measuring twice in a row. While there are outliers, the boxplots show that much of the data is in the range 5-20ms. If you use Recovery Points in HRV4Training, these differences in rMSSD translate in approximately 0.1 - 0.6 Recovery Points differences, depending on your baseline rMSSD and age.
Is measuring for longer a better idea?
In the following plot I grouped the measurement shown above based on measurement duration, as I computed features over sub-periods of 1 or 2 minutes, to see if the situation would improve. Looking at the conditions above, it doesn't really look like it, except for SNB1. Here is the data by group (1min vs 2 min, regardless of the measurement condition):
Again, pretty much the same. Slightly smaller variations, but we can't really take this as solid evidence that 2 minutes are better than 1. I personally started taking 2 minutes measurements out of curiosity some time ago, but I might go back to 1 minute after seeing these data.
Do I get more consistent results by using guided paced breathing?
Finally, I grouped paced breathing and self-paced breathing regardless of the measurement condition, to see if we could get higher repeatability (theoretically smaller differences between consecutive measurements) using paced breathing:
The plot above is basically summarizing the left vs right side of the first plot. We can see that during paced breathing measurement duration seems to matter less. Results are the exact same across sensor modalities for measurements of 1 or 2 minutes when using paced breathing, while there are some tiny changes for self-paced breathing. Also, self-paced breathing seems to lead to slightly worse measurement repeatability for 1 minute measurements, as differences are slightly higher. You can tell these differences are really tiny.
Looking at these data another question probably comes to your mind: if there is no improvement in measurement repeatability using paced breathing, is paced breathing useless?
Here we need to make an important distinction. One of the reasons to use paced breathing is that you are not measuring one single time as we did in this study. You are measuring for weeks, months or even years. This means that you might change your breathing pattern, or you might notice that with deeper breathing your HRV gets higher, thus subconsciously (or not even) "control" the measurement. For these reasons I believe it is a very good idea to use paced breathing and make sure that over time changes in HRV are due to actual processes that are relevant to you (e.g. training, or other), and not simply to changes in breathing frequency/patterns over weeks or months of measurements.
Finally, all plots above show the same differences across sensor modalities, highlighting once again how the camera measurement is as good as an ECG or a Polar H7.
Adding a plot that shows how the error changes based on a participant HRV. Here I split HRV data in quantiles, meaning that we have clustered all participants into four groups based on baseline HRV values. You can see how differences in repeated measures are slightly higher for higher HRV values, as we would expect (another way to look at this would be to plot the percentage error, which is more similar across groups, highlighting a similar relationship):
As daily measurements taken in succession can vary, and variations of 5 - 20 ms or 0.5 recovery points are perfectly normal, we now have a new way to display the daily advice that also takes in account what your normal values are.
Your normal values are simply derived from your historical data, in particular from the past 2 months of data, and should make it easier to interpret the daily score and understand that there is always a normal variation even between successive measures, as your physiology is never in the exact same state.
Check this article for more information on how your normal values are computed and used to provide meaningful advice while taking into account the fact that there will always be some variation in your measurements. When your data is within the non-transparent area that covers the central part of the color-coded advice bar, it means that nothing really changed, and the fluctuations you see in there, are simply normal physiological variations. For the daily advice or in general when analyzing acute HRV changes (see this post for a write up on acute HRV changes), you'd expect that a day following intense training your HRV is reduced due to higher physiological stress, so if your rMSSD is normally 60 ms, getting a 40 or a 50 is still a reduction, regardless of small variations in consecutive measurements, and the app will be able to interpret this reduction.
Here is another example with my own data in HRV4Training Coach, where you can see for example the effect of being sick on my physiological data. While there is still the usual day to day variability, values are quite lower than my normal, as they are lower than the yellow band indicating my normal values:
Finally, when you have enough data to look at Insights / HRV Trends, you will see that these differences play a smaller role, as you look at moving averages and baseline changes that should be indicative of the big picture, instead of just acute daily changes.
In this post I analyzed a dataset consisting of recordings from 26 participants and three sensor modalities (ECG, H7, camera). For each participant we had four measurement conditions, including lying down and sitting with or without paced breathing.
We found that on average we have about 15 ms differences in consecutive HRV measurements under these conditions, and that paced breathing as well as slightly longer measurements (2 minutes instead of 1), did not change much the outcomes. Nonetheless, for the reasons explained above, we would still recommend using paced breathing to make sure your breathing pattern does not change over time, as app usage is continuous, while this study was a single measurement per person.
I also showed how we deal with this variability in HRV4Training, by always analyzing your data with respect to what is your normal variability based on your historical data, and therefore making meaningful interpretations. Hopefully, this gives some clarity on a topic that is not much discussed but clearly puzzles many, as these questions are among the most common we receive.
Register to the mailing list
and try the HRV4Training app!
This blog is curated by
Marco Altini, founder of HRV4Training
The Ultimate Guide to HRV
1: Measurement setup
2: Interpreting your data
3: Case studies and practical examples
1. Intro to HRV
2. How to use HRV, the basics
3. HRV guided training
4. HRV and training load
5. HRV, strength & power
6. Overview in HRV4Training Pro
7. HRV in team sports
1. Context & Time of the Day
3. Paced breathing
4. Orthostatic Test
5. Slides HRV overview
6. Normal values and historical data
1a. Acute Changes in HRV
1b. Acute Changes in HRV (population level)
1c. Acute Changes in HRV & measurement consistency
1d. Acute Changes in HRV in endurance and power sports
2a. Interpreting HRV Trends
2b. HRV Baseline Trends & CV
3. Tags & Correlations
4. Ectopic beats & motion artifacts
5. HRV4Training Insights
6. HRV4Training & Sports Science
7. HRV & fitness / training load
8. HRV & performance
9. VO2max models
10. Repeated HRV measurements
11. VO2max and performance
12. HR, HRV and performance
13. Training intensity & performance
14. Publication: VO2max & running performance
15. Estimating running performance
16. Coefficient of Variation
17. More on CV and the big picture
18. Case study marathon training
19. Case study injury and lifestyle stress
20. HRV and menstrual cycle
21. Cardiac decoupling
22. FTP, lactate threshold, half and full marathon time estimates
23. Training Monotony
Camera & Sensors
1. ECG vs Polar & Mio Alpha
2a. Camera vs Polar
2b. Camera vs Polar iOS10
2c. iPhone 7+ vs Polar
2d. Comparison of PPG sensors
3. Camera measurement guidelines
4. Validation paper
5. Android camera vs Chest strap
6. Scosche Rhythm24
7. Apple Watch
9. Samsung Galaxy
1. Features and Recovery Points
2. Daily advice
3. HRV4Training insights
4. Sleep tracking
5. Training load analysis
6a. Integration with Strava
6b. Integration with TrainingPeaks
6c. Integration with SportTracks
6d. Integration with Genetrainer
6e. Integration with Apple Health
6f. Integration with Todays Plan
7. Acute HRV changes by sport
8. Remote tags in HRV4T Coach
9. VO2max Estimation
10. Acute stressors analysis
11. Training Polarization
12. Lactate Threshold Estimation
13. Functional Threshold Power(FTP) Estimation for cyclists
14. Aerobic Endurance analysis
15. Intervals Analysis
16. Training Planning
17. Integration with Oura
18. Aerobic efficiency and cardiac decoupling
1. HRV normal values
2. HRV normalization by HR
3. HRV 101