Blood oxygen readings wildly inaccurate!

[JulianL]

The technology of pulse oximetry was based on resolving the blood flow pulse in a finger tip (or other extremity) with two colors of Infra-red light. They use two led light emitters and two light detectors. They chose the finger because that's a great place to measure arterial blood oxygen as blood pumps thru the capillary bed of the fingertip. Monitoring the pulse waveform and measuring at the peak of the pulse gives the most accurate measurement. Most pulse oximeters display the waveform during use so that the operator can visually see if the blood flow is good, by the shape of the displayed pulse, and if compromised in some way they can re-position the arm etc.

The Apple watch blood oxygen measurement, on the back of the wrist, is not at all optimal to measure SPO2. I'm pretty sure that is why they don't show the waveform as I suspect it may not look like a classic crisp pulse ox waveform from a finger. They take a 15 sec measurement and probably have an algorithm to compensate for the measurement location. All that said, decent job in my opinion, good enough for a personal trend measurement. SPO2 is a "boring" measurement as we are all normally 95% and higher. When it's lower than this, one would want to use a certified device to monitor before administering any form of treatment.

Interesting extra info. Thanks. I'm afraid though that for me personally, at least on the automatic background readings, the s6 doesn't do a decent job in my opinion. Perhaps I'm simply unlucky, or genuinely quite ill, but in the 3 weeks that I've had my watch I've seen background readings as low as 85% a few times and they're often in the upper 80s or the low 90s. If I take manual readings I'm pretty much always in the 98-100% range. On the times when the automatic readings have shown <90% numbers I have never been feeling in the slightest bit unwell, dizzy or tired at the time which I assume (maybe wrongly) would be likely symptoms of <90% levels.

My conclusion is that accuracy is very heavily influenced by how still and well aligned (watch facing upwards) the wrist is when a reading is taken.

I wonder whether a future watch (s7 I hope) might be able to introduce a version 2 of the sensor that might increase accuracy maybe by having stronger lights meaning it was more tolerant of bad sensor positioning or maybe by having other sensors so that it can know if there is a slight gap between the sensors & the skin's surface and/or if the sensors might be at a slight angle to the wrist (for instance if someone wore their watch quite looks and had their arm hanging down) so that the algorithms could compensate. Or maybe just more fine tuning on the algorithms themselves might improve accuracy at least somewhat and that wouldn't need to wait until s7 unless the improvements made things too computationally intensive for the watch to handle (but then it could always be passed off to the iPhone for processing I suppose).

Hopefully Apple can improve things at some point in the next year or two if not sooner because its a nice feature to have and I wish that I could trust my automatically taken background readings more than I do.

 

[RemE]

Interesting extra info. Thanks. I'm afraid though that for me personally, at least on the automatic background readings, the s6 doesn't do a decent job in my opinion. Perhaps I'm simply unlucky, or genuinely quite ill, but in the 3 weeks that I've had my watch I've seen background readings as low as 85% a few times and they're often in the upper 80s or the low 90s. If I take manual readings I'm pretty much always in the 98-100% range. On the times when the automatic readings have shown
My conclusion is that accuracy is very heavily influenced by how still and well aligned (watch facing upwards) the wrist is when a reading is taken.

I wonder whether a future watch (s7 I hope) might be able to introduce a version 2 of the sensor that might increase accuracy maybe by having stronger lights meaning it was more tolerant of bad sensor positioning or maybe by having other sensors so that it can know if there is a slight gap between the sensors & the skin's surface and/or if the sensors might be at a slight angle to the wrist (for instance if someone wore their watch quite looks and had their arm hanging down) so that the algorithms could compensate. Or maybe just more fine tuning on the algorithms themselves might improve accuracy at least somewhat and that wouldn't need to wait until s7 unless the improvements made things too computationally intensive for the watch to handle (but then it could always be passed off to the iPhone for processing I suppose).

Hopefully Apple can improve things at some point in the next year or two if not sooner because its a nice feature to have and I wish that I could trust my automatically taken background readings more than I do.

One can hope. The back of the wrist is a tough spot to measure from. Fingertips yield consistent results because it's far easier to position the sensor every time. Plus the light passes thru the nail and is received under the finger tip, so stronger signal.

This is why we are seeing new bands, seeking tight contact with skin. Many people, my wife included, prefer a loose fit watch. This messurent is doomed for them. I prefer a tight fit and have used 3rd party stainless link bands, carefully fitted, for years now. So for me this measurement has been pretty consistent.

Algorithms can only do so much, the physical connection must be adequate and consistent or it's garbage in, garbage out.

 

[Hyloba]

My question is is it is good enough to detect altitude sickness? Last year I was hiking in the himalayas and I would have appreciated an oxymeter.

 

[cdcastillo]

We also see different results on different fingers. The one you are using is widely seen as the most accurate but others may give results more in line with the watch. But yes, there is a high chance the watch is not right and guessing.

Yes, according to Basaranoglu Et al. the right medium and thumb fingers tend to show the highest saturation consistently of the 10 hand fingers.

The technology of pulse oximetry was based on resolving the blood flow pulse in a finger tip (or other extremity) with two colors of Infra-red light...

Usually it's with green (537 nm), red (660 nm) AND infrared (880 nm) light. Oxygenated haemoglobin has a lower extinction coefficient with red light than de-oxygenated haemoglobin; the opposite is true with infrared light. The green is for control purposes and to help with detection of the pulse wave.

 

[chanerz]

Apple watch blood O2 measurement will not be accurate. It is just a gimmick. Medical grade measures on the finger, ear lobes, and toes.

 

[Rogifan]

The Health app gave me a blood oxygen reading of 82% it said was taken at 3:28 in the morning. I’m glad I don’t need to care about these readings because they‘re laughably inaccurate.

 

[dlopan]

My automatic readings are really close to my norm, the manual ones are complete crap.

 

[EugW]

I am surprised that the reviews talk up the O2 sensor so much, because from a medical perspective, it is completely useless. It would be very, very foolish for anyone with a medical problem requiring a pulse oximeter to rely upon this. For example, one popular reviewer suggested this product would be good for someone like his son with asthma. No, no, no.

I guess Apple thinks so too, considering it’s calling it just a “wellness” feature. But even from that perspective it is not really useful. For it to be useful, it would need to be accurate.

 

[fhopper]

My Garmin shows my overnight O2 to often drop into the 80's and average 90-91 while my daily Oximeter reads 97-98. I suspect the Fitbit people have the same situation. And has been stated the Trend is what matters...and a watch will never have a consistent position to read from.

 

[sclawis300]

My Garmin shows my overnight O2 to often drop into the 80's and average 90-91 while my daily Oximeter reads 97-98. I suspect the Fitbit people have the same situation. And has been stated the Trend is what matters...and a watch will never have a consistent position to read from.

Are you sure you don't have sleep apnea?

 

[EugW]

Are you sure you don't have sleep apnea?

S/he should try an oximeter at night. If the pulse oximeter is reading 97 and the watch is reading 80, the watch isn't working properly...

...which makes sense because watches simply aren't designed to read this properly. Thus personally I think the O2 feature is pointless in these watches.

 

[fhopper]

Are you sure you don't have sleep apnea?

Your concern is kind. I have had sleep apnea for over thirty years and have a machine that syncs with my phone to report how my "nights" are. I do well.