While most of the time, O2sat is measured at the finger, there are times when it is necessary to measure it elsewhere, such as ear lobe, Achilles, or toe.
2 Likes
It may not be just skin tone, a very dark pigment might block too much signal when trying to transmit
through the finger (there’s bone in there too).
Maybe earlobe is a better place to try and transmit light through, if it’s thinner than a finger.
Also might have more blood flow. It’s where they originally put the sensor; before the
pulse tracking technique to cancel out background, they used to pinch your earlobe until
it turned white, to get the background absorption level without blood in the tissue.
2 Likes
The headline had it right. It’s bias. As in literal instrumental bias. Just need to have a calibration of increased melanin absorption.
2 Likes
The existence of confounding results?
From 1998
Effect of Skin Pigmentation on Pulse Oximetry Accuracy in the Emergency Department
Conclusions:Although several prior studies suggest the contrary, this study found that skin pigmentation does not affect the bias or precision of pulse oximetry. Furthermore, skin pigmentation has no clinically significant effect on PO signal quality.
So far from it being a problem that was blithely ignored for decades, it’s been studied for twenty odd years, with inconclusive results.
Why did the authors cite only two papers?
3 Likes
I appreciate your spirit. I have tried simuating flesh colours from spectral data to try and get consistent flesh tones for motion pictures, and it is not very promising. You getting less light, so any measure is going to get harder. Melanin (there are two types of melanin) lies above the blood cells. Melanin scatters light as well as absorbing it. And the oil level on the surface of the skin can have a lot of influence on how light interacts with the upper levels of the skin, but this matters much less when all the upper levels are pretty transparent.
I have a pulse oximeter. They are handy for telling whether you are getting worse or getting better. But some people with poor circulation can get low readings even when their breathing is OK.
We have very few numbers to tell how people are. We have temperature, blood pressure, heartbeat, and now oxygen level. It is good to have a new one, but it isn’t magic. People are different; we should keep our wits about us, and not rely on one measurement for everyone.
4 Likes
Earlier this year my wife was in hospital in the UK with a sudden epileptic seizure. I was in the AMU while she was being treated. The oximiter she had kept failing - you could see it on the readout. The attending nurse clicked her tongue and changed the clip from one hand to the other and back again. “They’re not very reliable,” she told me. “Don’t worry, we’re keeping an eye on her.” Gods bless the NHS.
7 Likes
I’d guess these devices were primarily created, and tested, with white skin in mind…
It’s also possible that the testing phase skewed disproportionally towards south east Asians – rather than western whiteys.
I am not saying this to absolve my white ass in any way, just wanted to point out that just removing the possibility of Asian engineers from the equation isn’t all that woke either.
4 Likes
In a very brief search, I found a dozen or more that were pretty uniform in showing a problem with Black adults, with far less to no issues in children and infants, dating back over a quarter century to shortly after the introduction of the techno.ogy to 2017, the most recent I came across. I do not think a single study can make me discount all the others. I agree about confounding variables, though, as youth seems to factor in strongly. Maybe skin thickness? As I mentioned earlier, I have never seen this, but given what I do, and the results I found, that actually makes sense.
6 Likes
Thank you for this detailed reply. Technical detail and anecdote - perfect! Comments like this are why I love BB.
3 Likes
Now 50 percent off at the Boing Boing shop.
10 Likes
They tend to wrap it around baby’s feet, too!
1 Like
The fact that it hasn’t made it into medical training or practice is the part that means it’s been ignored. Having studies in a journal doesn’t mean shit if it isn’t put into practice.
And from a market perspective, there were two approaches in pulse oximetry: very reliable and accurate, but expensive; and cheap, unreliable, and less accurate.
Cheap won. So that’s what’s available on the market.
3 Likes
There are all sorts of medical “facts” about Black people (e.g. involving skin thickness, lung capacity, pain resistance) that are completely untrue, that we know are untrue, yet are still taught in texts and medical schools. Things are more fucked-up than you might imagine, in terms of motivation to fix known problems, when it comes to providing care for Black people.
9 Likes
Comparing spectrums has already been taken into account. The biggest frequency restriction is Hb vs Hb02 spectrum. ~660-680 nm seems to maximize signal to noise ratio. Going lower will make the influence of melanin greater, but going higher doesn’t improve the effect of melanin, but does greatly decrease the O2 measurement sensitivity. So pulse oximeters use a 660 nm red LED.
But they need a calibration anyways. This is done at 940 nm. Between the two, a ratio of Hb to HbO2 can be calculated.
The absorbion spectra of melanin is relatively flat in this range and much lower than at short frequencies (melanin is great stuff! Great at blocking UV). Not a perfectly flat slope, but even the research papers don’t agree on the magnitude or slope. And type of melanin has an influence.
So the effect of melanin is a signal to noise ratio problem. Increasing the intensity won’t fix that. Knowing the amount of melanin might not be effective either (i.e. all it could tell you is what noise ratio to expect, not compensate for it.) So having a switch for skin tone would be ineffective.
What might be marginally effective is to add other frequencies. Good news is that this could help with signal to noise by adding another signal. That would improve accuracy for measurements, regardless of skin color. Bad news is that it very likely won’t bring errors for patients with more melanin in line with those with less.
This isn’t a story about bad engineering. This is about the limitations of engineering.
8 Likes
Great post! Unfortunately, adding additional wavelengths would also add cost. That’s a non-starter unless there is a regulatory push to correct the problem. In the US, FDA could force the issue by putting a sunset date on older designs that fail more often and streamline approval for new systems that are more accurate across many skin tones.
1 Like
Melanin absorbs IR and visible red light equally and the measurement is done as a ratio. Melanin levels don’t matter for the simplest forms of pulse oximetry – it was worked on very carefully at the beginning of pulse oximetry development. Newer models may be doing something else creating this bias, but the basic concept isn’t affected by melanin.
1 Like
I remember a story from a while ago that this skin tone confounding sensors is also a problem that exists in the sensors in hand driers found in bathrooms as well.
I’m not sure LED price is reason. The price of LED’s and sensors is ridiculously low compared to a few decades ago when these first came out. So is processing power. And the improvement is going to be very small at best. The real cost is going to be regulatory.
Getting new clinical claims requires clinical trials. That’s a minimum of $3,000,000. But even with new claims, hospitals won’t pay more for the devices. Reimbursement costs are fixed. An improved version will get the exact same reimbursement code.
So manufactures are left with the decision of making the same product, or making a more expensive, but only tiny bit better product that sells for the same price. It’s a gamble few companies are going to be willing to take. Especially if hospitals aren’t asking for it.
As for sunset dates, that’s a bad idea. It’s not like pulse oximeters are bad. They just aren’t as good as we’d like. Sunset dates make manufactures choose between redesigning a product or just exiting the business. If they have to develop a new product, do clinical trials, and make less money, why not make change to making some other product with better margins? You’ll just see manufactures drop out of the pulse ox biz.
But if someone can show the clinical benefit (in terms of dollars) of better sensor tech, then hospitals will demand better sensors and pay more even though reimbursement to them is constant. Manufactures will jump at that.
BTW, this is true of any health care system. Single payer or private care, economics don’t go away.
2 Likes
Two possible problems. First, newer pulse oximeters are not necessarily better than the older models-- they may simply be cheaper to manufacture (or they rely on patents that have recently expired). Second, it is expensive (and potentially unethical) to test oximeters on hypoxic subjects-- so all of those testing protocols may miss the bigger picture.
The nejm letter says
The pulse oximeter type (transmittance-mode versus reflectance-mode) used in SpO2 measurements was not recorded for either cohort.
so, detailed analyses of “older design” vs “newer design” appear to be precluded by this dataset.
1 Like
