Oxygen Saturation measurement seems useless - am I missing something?

[mmcc73]

I got a FitBit Sense recently in no small part due to its SpO2 measuring capability. I'm finding that function to be pretty disappointing....

• It only works while sleeping

• It only provides max/min/avg without any way of looking at more detailed data

• Most importantly, it seems completely inaccurate. I had a sleep study done while also wearing the Sense - the real oximeter showed readings regularly in the low 80s and high 70s (sleep apnea is a bummer). The Sense said the overnight minimum was 92.

I'm wearing the device fairly firmly a good inch above my wrist joint.

I was hoping this thing would be an easy way to see how well sleep apnea treatment was working, but it doesn't seem like it will be useful for that.

Is anyone else having better luck?

[LokeAa]

Fitbit clearly states "The Fitbit SpO2 feature is intended for general wellness purposes only and should not be used or relied on for any medical purposes."

Too see how well your sleep apnea treatment are working, uou should listen to your doctor, not your smartwatch.

[mmcc73]

FitBit makes that disclaimer for every measurement their devices provide... if we are to disregard all of the information they provide provide it kind of defeats the purpose of having the device in the first place.

I don't expect it to be as accurate as a calibrated instrument in a hospital, but I do it expect it to be directionally correct to the extent that I know if things are getting better or worse. Right now that does not seem to be the case.

[Rich_Laue]

@mmcc73 The variable SpO2 measurement hby fitbit, is not a direct measurement of the SpO2 level. 

It measures the amount of variation. 

[mmcc73]

I beg to differ…

It also shows variability in the app, but what gets reported out on the watch is max/min/avg of SpO2.

Here’s the variability chart - note the lack of units and scale defined on the y axis. My middle school science teacher would break out the red pen for such an offense. It sure would be nice to know what “High” and “Low” mean.

 

[LokeAa]

This is what Dr. Conor Heneghan, Fitbit’s Lead Sleep Research Scientist says about Fitbits SPO2 feature.

[mmcc73]

Thanks - that is an interesting article. However, it is a bit out of date. The article says...

Fitbit's detection is a little different. Rather than spitting out a single percentage score it examines the difference in the highs and lows of your blood oxygen, and looks to highlight potential underlying conditions.

Whereas the devices do spit out percentage scores. The variability curve is interesting, but along with it I'd sure like to see the actual measured values. I suspect the inaccuracy (particularly of the min value in my case) is that they are overly smoothing the data. The min is more like the min-1-minute-average instead of the actual minimum reading. If there are any FitBit people floating around, it would be nice to get confirmation on that.

[Rich_Laue]

Fitbit has not designed their graphs to see actual SpO2 levels. 

From Fitbit"a help page

• Estimated oxygen variation (EOV) is an estimate of the changes in your blood oxygen saturation levels. For instructions, see How do I track my estimated oxygen variation in the Fitbit app? The EOV feature is designed to see specific events overnight when your blood oxygen saturation levels varied, which could indicate breathing disturbances.

  Fitbit SpO2 estimates your average blood oxygen saturation levels and range while you sleep. It’s designed to see potential changes to your fitness and wellness.

[mmcc73]

My only complaint with the EOV charts is that the units and values aren't defined for the y axis.

My main complaint is that the daily SpO2 max/min/avg values seem grossly inaccurate. I'd like to be able to see the actual second to second data to understand what is going on better.

[mmcc73]

I've been playing around with the Web API which does provide some intraday SpO2 data. There seems to be a problem with authorization of access to the data, but I was able to get some to look at through the API explorer (https://dev.fitbit.com/build/reference/web-api/explore/)

This is progress, but not much.

The intraday API (https://dev.fitbit.com/build/reference/web-api/intraday/get-spo2-intraday-by-date/) provides one value per minute, and "Spo2 values are calculated on a 5-minute exponentially-moving average".

When I'm experiencing sleep apnea, my blood oxygen oscillates with a 30 second period. Providing 5 minute exponentially-moving average data obscures most of the interesting stuff. 

PLEASE provide the per second raw data. Via the API is fine. I'll do smoothing on the data if necessary. Once I finally get a CPAP machine, I want to be able to use the data to determine how much impact it is having and to help figure out what settings work best. My fitbit is much more comfortable to wear while sleeping than a oximeter that clamps on my finger.

[CraigHansen]

After the recent software upgrade, the overnight SpO2 variability graph is totally broken. It's been like this for several days, with no fix in sight.

[CraigHansen]

Which is a huge problem. When I bought this Sense watch, I was promised an SpO2 feature. All I get is a vague Estimated SpO2 variability, only after I've sleeping for several hours, and only if my frequency of PVCs while I sleep is low enough that the stupid device thinks my heart rate isn't too variable. Otherwise, all I gets is a chart showing when it thinks I'm asleep and awake, and all the other data is suppressed.

I can buy a fingertip SpO2 meter and get real-time data for about the price of a replacement Sense charger cable (which I have to purchase every few months because they keep failing). But the fingertip monitor doesn't record the numbers so I can review them - that what I bought the Sense for.

When is the SpO2 feature going to arrive? For realz? I've had this watch for almost two full years, and it's still highly problematic.

[CraigHansen]

A 5-minute exponentially moving average filter can be reversed with a mathematically inverse filter, which would get you the raw data at one value per minute. That doesn't get you 30 second data, but even with a fingertip meter, how often is the SpO2 display really updated? I can make the number move with some deep breathing, but the update rate seems pretty slow with a fingertip meter.

Is there an easy way to get past the authentication hurdles needed to see the API data? Before I'd invest in writing code, being able to type web APIs into a browser would be helpful. Thanks!

[mmcc73]

I have this oximeter: https://a.co/d/535RjPJ

It logs readings once per second and you can pull them into an app on your phone in the morning. I don't use it every night as it isn't terribly comfortable.

There are other models that have a wrist thing and a finger thing that might be more convenient for sleeping in.

https://a.co/d/84vlKUr

[mmcc73]

Regarding trying out the API, try the Web API Explorer described here:

https://dev.fitbit.com/build/reference/web-api/troubleshooting-guide/debugging-tools/

My guess is that the per minute values are averages of 5 minute per second readings, but that is just a guess.

The EMAY oximeter provides per second readings, max/min/avg, total time in various ranges, number of ODI 3% and ODI 4% events (when blood oxygen drops 3% or 4% from baseline)... my sleep doc seemed pretty impressed with it.

There's no reason the Fitbit couldn't provide this information... or at least per second data so this information could be calculated.

[mmcc73]

So I wore my EMAY finger oximeter last night for a few hours in order to compare it against the FitBit. Something is definitely off.

For what it is worth, I had my daughter - who currently has Covid mind you - put on the EMAY finger oximeter for 5 minutes yesterday and her readings were between 99 and 97. 

The chart below shows SpO2 data from the Fitbit - obtained through the Fitbit API - the raw data from the EMAY device, and the EMAY data with a 5 minute exponential moving average applied.

The Fitbit was on my left arm. The oximeter was on the index finger of my left hand. Yes, my finger oximeter readings get real low - sleep apena sucks and I'm waiting on a CPAP machine.

I tend to believe the data from the finger oximeter as in the past it has largely agreed with data collected during sleep studies, and I've tried it on other family members whose data looked much more normal.

[Rich_Laue]

Since fitbit says they give a relitive reading, let's look at the graphs. Fitbit does seem to follow the trend. 

There must be something we are missing, the FDA approved, it was designed with the help of the medicke community, and in a few cases a users doctor understood what they where looking at. 

[mmcc73]

They say the give a relative reading for SpO2 variability, but the intraday SpO2 gives every indication of being intended to be absolute value. 

I draw your attention to the API documentation: https://dev.fitbit.com/build/reference/web-api/spo2/

There is no mention of it being a relative value. The only references I’ve seen to relative values have been as it pertains to SpO2 variability, not the max/min/avg values that are reported out on the watch or the time series data available from the API.

The fact that it is FDA approved means they jumped through the necessary regulatory hurdles to get approval. This unfortunately does not necessarily mean it is accurate. 

Please note: I’m pretty confident that lack of oxygen causes general grumpinesses which I’m definitely experiencing.

[CraigHansen]

What's even the point of "FDA Approval" of a device that so emphatically claims not to be a medical device? Even the well-under-10-dollar fingertip SpO2 device that provides instant SpO2 readings claims "FDA Device Listing #D428836."
Coming from the other side, the FDA says: "Registration of a device establishment, assignment of a registration number, or listing of a medical device does not in any way denote approval of the establishment or its products by FDA."
So, "FDA Approval" doesn't even mean that the FDA approves of the product.
You keep using that word; I don't think you know what it means. 

[mmcc73]

Inconceivable!