HRV: Heart Rate Variability

[Mccoy_Mccoy]

Hi All,

I've read some previous messages on the topic and some users complain about inaccuracies in the measurement of HRV variability in the Inspire3. More specifically, an underestimation of the HRV compared to some benchmarks is reported.

I'd like to join to the complaints. I'm a new user, I've been comparing HRV as measured by the inspire3 to HRV as measured by the PolarH10 chest strap and a specific apps like EliteHRV.

The values measured using the PolarH10, which is sort of a reference in this field, are consistently 80% to 100% higher than the values measured by the Inspire 3. In other words, the Inpire3 consistently and widely appears to underestimate HRV.

I wonder if it is an algorithm issue and if something can be done about that.

[Gr4ndp4]

Hi, @Mccoy_Mccoy, welcome to the community, firstly you may be interested in How do I track heart rate with my Fitbit device? , I apologise if you have already seen this.

The LED/sensor array in your Inspire 3 optically measures the change in the colour of a skin sub-layer caused by blood flow and can't be compared with chest straps which measure the electrical pulses generated by your heart. In addition, sampling rates will be different and as you note each manufacturer has their own algorithm

I wear a Fitbit Inspire 3 and another alternative makers device using optical wrist-measured HR, while they do not match precisely they are not far apart, as I write I have 60 on the Fitbit & 67 on the other device.

I hope this helps.

Cheers

[Mccoy_Mccoy]

Hi Grandpa, thanks for your answer.

I don't know if I construed well your words, but the HR in the Inspire3 is alright if compared to the Polar H10. In other words, the parameter HR seems to be accurate (values very similar to the PolarH10) at rest, on a stationary bike, and even on the treadmill.

It is the parameter HRV (heart rate variability) which is way off the values provided by sensors with a reputation of great accuracy like the PolarH10. That's why I have thought it may be an algorithm issue, even though I'm aware that other parameters are at play in calculating the HRV.

[Baltoscott]

@Mccoy_Mccoy - Can you clarify what you are actually comparing? I downloaded Elite HRV, mentioned in your first post, but I don't see any option for connecting a Fitbit Inspire 3, which I can understand as the Inspire 3 is not advertised to do HRV on demand; instead it is designed to calculate HRV while sleeping. I'm guessing you use your PolarH10 for spot checks and for exercise HRV?

As @Gr4ndp4 mentioned, comparing HRV during sleep across devices can be problematic because of different sampling frequency, the time period measured, and adjustments is wearing the device. Fitbit looks at the longest sleep period over the past 24 hours.

[Mccoy_Mccoy]

Hi Baltoscott, yes, as you have correctly guessed, EliteHRV only connects with the Polar strap and sorry for not having mentioned it before, I've been doing spot measurements in the morning, shortly after I woke up and before engaging in breakfast and other activities. So, while in the case of simple HR the measurements done by the two distinct devices have been carried out during the same time interval, in the case of HRV I compared different time intervals. However, I wouldn't expect the difference between the all-night sleeping HRV and the resting morning HVR to be so wide. Last night it was 23 ms according to the Inspire3 and 49 ms according to the PolarH10 in the morning and such large differences are consistent across days.

On the other side, literature says the long time (24 hours) and short time (<5 minutes) measurements are not interchangeable, so to be rigorous I should wear my chest strap all night and compare directly the results of the two devices. Some difference can be expected but, again, so high?

The trend of course may be a useful parameter regardless of the absolute value.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5624990/

[Triletics]

@Mccoy_Mccoy you are comparing apples to oranges. Different time domain, different sensors, different algorithms (error correction, averaging, etc.). When you compare to EliteHRV you should look ar RMSSD (not at final HRV score, this is a proprietary metric). Still, you will find differences mostly because of time domain. Even if you wear strap at night you won't test the same because you don't know when Fitbit's time domain starts and when it ends. In other words, there is no reference to perform any comparison and claim it's valid. And yes, the difference between 24hrs domain and short 5min domain may be huge. You have lot more data to filter, correct, squash and that may provide vastly different result. In other words, there is no way to test whether Fitbit's night HRV has any accuracy in it because for that you'd need to verify raw data (RR intervals) and know methodology.

[Baltoscott]

As you say,

“The trend of course may be a useful parameter regardless of the absolute value.”

Exactly. 

[Triletics]

@Mccoy_Mccoy see this post showing comparison between two devices of sleep HRV. It shows why it's trend that matters not absolute values.

[Mccoy_Mccoy]

---

@Triletics wrote:

@Mccoy_Mccoy you are comparing apples to oranges. Different time domain, different sensors, different algorithms (error correction, averaging, etc.). When you compare to EliteHRV you should look ar RMSSD (not at final HRV score, this is a proprietary metric). .....

---

Yes, now we are reasoning. I discovered that HRV in ms is not a standardized unit. My mistake was assuming that this field is comparable to other engineering fields, where a parameter is described everywhere by the same unit, or different units, which can be converted easily among them. 

As you are saying, I discovered that EliteHRV uses a score, derived from a scale enlargement of the ln(RMSSD) quantity, which is proprietary. So, as you say, fitbit & HeliteHRV= apples & oranges. One millisecond measured with a method is different from one millisecond with the other method.

Everything else is apparently different . The ECG-comparable sensors, like the H10 Polarstrap, measure the R-R interval and apply suitable algorithms (which are probably proprietary themselves). The sensors based on green light signal modulation use another parameter (another user said the P-P interval) which requires a different numerical treatment.

[Mccoy_Mccoy]

---

@Triletics wrote:

@Mccoy_Mccoy see this post showing comparison between two devices of sleep HRV. It shows why it's trend that matters not absolute values.

---

Thanks trilectics, I read the whole huge thread and the considerations from a user who is apparently working in the specific field, or a related one. 

Please also read my previous answer to your previous post of wednesday.

Now, pragmatically speaking, how can we utilize at best the data provided by Fitbit. It was easy to ascertain that the Inspire3 measures HR accurately enough, even in non-resting states like jogging on a treadmill hands free (up to 140 bpm in my case).

So, the night HRV parameter is not a garbage-in, garbage-out situation. It has its meaning even if it cannot be directly related to other HRV metrics.

I would suggest technical assistance to provide statistical data on Fitbit HRV values for different age brackets in different ages, as other labs are doing, so that we can locate ourselves in terms of percentiles and decide if we look statistically low or high in this specific HRV score. This would avoid the confusion and concern caused by different apps providing non-comparable scores in the same time-unit (milliseconds). A very tricky situations.

Another useful concept is that the trend may govern and not just the score itself. The other night I was sick. The Inspire3 HRV score correctly picked this anomaly by a dip in nightly HRV. I don't know if this is due to the sleep interruption of about 1.5 hours or to the HRV fluctuation itself, the dip was there though, as it was a higher-than-usual nightly average HR.

This is a snapshot of the plot, sorry the Italian language, the image speaks for itself though.

[Mccoy_Mccoy]

.............

[Mccoy_Mccoy]

OK, I tried to dive deeper into the difference between the HRV values indicated by Fitbit and those read on EliteHRV. It turns out that the latter as already discussed provides a proprietary score, but can also provide other parameters, like a pure RMSSD number in ms. As far as I can understand, Fitbit provides a RMSSD measurement, at least they write in their in-app info (my translation from Italian): "Generally, the formula used is the common one denominated RMSSD...".

Today I lied on the sofa and measured again my HVR using the polarH10 and elite HRV. This time I consulted the declared RMSSD value, which was 31 ms.

Tonight reading of my HRV by the Inspire3 was 32 ms, almost identical to the one measured by the Polar strap+EliteHRV.

So, while I agree that the HRV averaged throughout the night is not exactly comparable to the one measured during a one-minute interval, the values seem to agree very well, this time being evidently expressed in the same identical unit.

Bottom line, the HRV Fitbit algorithm seems to be pretty accurate, and all the fuss with the low values was probably due to the EliteHRV score which yields amplified numbers.

I'm going to post further measurements, again, I believe the relative value = the trend with respect to the individual baseline is meaningful rather than the absolute value. Some statistical values are also available, the following is a snapshot of my nightly HRV trend, with the latest value=32 ms.

[Mccoy_Mccoy]

This is an interesting plot to compare the median HRVs of males and females of all ages. Source the Oura Ring website. It is not specified what the color fields are, probably variability in percentiles (5th to 95th? 10th to 90th? 0th to 100th?)