I believe I said that. “the fact that a more reliable method hasn’t been invented or become the standard yet is at least partly because of systemic racism in the medical industry”. I’m also not sure the pulse-ox problem can be solved through calibration alone- if the melanin in the skin absorbs the light intended to be absorbed by the hemoglobin in blood, there may not be enough light getting through to give an accurate reading even if that difference is taken into account by medical staff. The best solution seems to be a new technology that avoids the difference in melanin entirely, like one under development that uses polarized light.
I think that, in the same way it’s useful to differentiate between systemic racism and racism on a personal level, it’s also useful to emphasize that people can be unintentionally racist by neglecting to think of BIPOC’s needs when designing or advertising products. A lot of people still seem to think they can’t possibly be racist if they’re not throwing around racial slurs and burning crosses on lawns. The term “passive racism”, at least in my opinion, may help them rethink just how deeply ingrained racism really is and force them to confront their unconscious bias and be actively antiracist in their business decisions.
This is false. Most modern hospital-based transmission SpO2 monitors still have problems with darker skin. That’s going to be true for any system that uses a direct conversion ratio of R/IR.
Just fucking calibrate these things on a wide variety of people, instead of just some random white dudes.
The FDA has required pulse oximeter manufacturers to do this for decades. Unfortunately, the requirement applies only to medical devices, not consumer grade ones.
Systemic racism implies it is somehow intentional, or that the “system” is aware of the issue and does not care. That is not the case. This is just physics. Melanated skin blocks more light of all wavelengths. The benefit is that it blocks harmful UV, the downside is that it also blocks the visible and IR wavelengths these sensors use. This is likely something Apple and others can’t get around without unacceptable LED power levels in terms of battery life or heat production.
FWIW, tattoo ink is even worse. If you have a full sleeve in the wrong colors these sensors do not work at all. They can’t even read your heart rate, which requires much less precision than SpO2. Apple did change the wavelengths their sensors use to try and be more compatible with darker/tattooed skin but there is only so much they can do with the technology available.
Unfortunately, that only applies to new approvals. That’s why there are only 3 clinical pulse ox manufacturers in the US - Masimo (which did the work), Nellcor and Ohmeda (which are grandfathered). You’d think that Masimo would dominate the market, but Masimo is a single-product-category company whereas the others are part of huge conglomerates that can bundle their old technology with other product to cling to their market share.
The issue here isn’t the power usage of the CPU but of the sensor. Real medical devices can pump out a lot more light and have more sensitive detectors to compensate for darker skin. They also have larger batteries and form factors that the Apple Watch and other wearables.
Apple and others test these devices on all kinds of skin colors and conditions. They made it work as well as they could given the constraints of the device.
Thanks for the clarification. I didn’t realize they also offer consumer products.
Semi related:
While trying to find info on SET (my google and pubmed skills failed on that one). I accidently came across this whitepaper from a semiconductor company on how their standard system works. I did note a comment about reference signal calibration around skin tones and the ever helpfull “empirical data table lookup”. https://www.nxp.com/docs/en/application-note/AN4327.pdf
I would love to know more about how SET compensates for the SNR loss from skin absorption (do they have alternates table look ups etc…). If you found any public facing info or papers I would be interested to read them. From the watch disco bottom I am curious if the wrist solution is just using more light and sensors to up the DC component and overcome stray light.
Edit to add a rant… - It continually suprises me how many populations get overlooked in medical device/test design. For example with blood tests, we don’t have reference ranges for pregnant women, for quite a few tests that change during pregnancy. And then there is genetic testing (my primary field) where the reference data is very white… I suspect with pulseox ethnic biases proper care may come down to a physician/nurse knowing about the bias and compensating for it in their clinical workup. The major problem with that is you are definitely not guaranteed to get a provider who does that enough to maintain competency (and thus hypoxia is missed way too often). Sadly (as you said above) it comes down to money.
The main difference with SET is that it starts from the assumption that both oxygenated and deoxygenated hemoglobin signals are dynamic instead of assuming the deoxygenated hemoglobin signal is constant. From that, calculating the SpO2 is more complicated but adjusts for variables such as venous flow changes, movement, and skin pigmentation.
No. This is absolutely false. Systemic racism does not require intent. It doesn’t require individual action. It is baked into institutions, systems, and culture. Insidious and subtle but responsible for serious harms.
You’d think that as many times we’ve had these conversations here, people would have learned this by now, but apparently, they still think racism stops at the KKK and neo-nazis… And we wonder how we got Trump in 2016…
Sure, but once again, there are people here just straight up denying that systemic racism even exists, that it’s not a real problem that we need to address, etc. They are gaslighting people right here on this thread, because they are too uncomfortable to face facts. I am sick of hearing this shit that we need to STFU about these forms of racism, because they are less important. Bullshit. It’s the systemic stuff that is harder to stamp out, because it doesn’t seem so direct. And pretty much every person who thinks systemic racism “isn’t real” sets us farther back on fixing our shit.
You don’t want another Jan. 6th, then stop nit-picking and start calling out this shit whenever you see it, even if it seems small and less important than the big stuff. Same with misogyny, homophobia, transphobia… whatever. It will not win you friends, in fact people will despise you for calling this stuff out. But it will actually help the problem, because someone will see it and maybe spend some time thinking about this larger issue that often gets swept under the rug because some people just can’t deal with it.
I’ve been going to this working group at the NIH for a while now, where one of the topics has been how to integrate this kind of data (e.g. noisy, biased, and sparse/incomplete) into mathematical models of disease and physiology. Everyone always gets excited about wearable sensors because they at least solve the data sparsity problem. Personally, I don’t have access to a population level dataset, so I’ve always been a little unsure about how much signal is in all that data. That said, on an individual level, abrupt changes from any baseline ought to indicate something worth looking at. (I say abrupt here because skin tone isn’t likely to change faster than the seasons. Or how ever long it takes you to move the watch to the other arm, I guess.)
