SpO2 Across Every Skin Tone: What Our Data Shows

Of everything our validation study measured, this is the result we were most anxious about going in, and most relieved by coming out. Oxygen saturation across skin tones is precisely where the medical-device field has a known, documented failure. We had built and recruited specifically to test whether our screen avoided it. The data says it did.

The problem we were testing against

Standard pulse oximeters have been shown to overestimate oxygen levels on darker skin. The mechanism is straightforward: the measurement reads light through the skin, and skin pigment affects the light. When devices are calibrated mostly on lighter-skinned people, they inherit a bias that shows up as falsely reassuring readings for darker-skinned patients. A 2022 regulatory communication flagged this directly. For a population that is mostly medium-to-dark-skinned, an oximeter that quietly performs worse on darker skin is not a neutral tool. It is a hazard.

So the test was not just "is the screen accurate." It was "is the screen accurate for the people most often failed by this exact measurement."

What we found

In the validation study, SpO2 from the screen was compared against a clinical-grade reference oximeter in our cohort. The overall mean bias was −0.31% — close to the reference, and biased very slightly low rather than high, which is the safer direction to err. Most readings fell within tight bounds of the reference value.

The result that mattered most: when we broke the data down by skin tone across Monk categories 4 to 10 — including the darkest tones, 8 to 10 — the accuracy held in every subgroup. There was no group where the screen quietly fell apart. The agreement that looked good on average also held up when you pulled the average apart and checked each part of it.

Why we report it this way

A single averaged accuracy figure is exactly the thing that lets skin-tone bias hide. A device can look fine "overall" while failing a subgroup, because the subgroup gets diluted into the average. The only honest way to claim accuracy across skin tones is to publish the breakdown, including the darkest tones, and let it be checked.

That is why we recruited deliberately across the full range, and why our reporting gives the per-skin-tone picture rather than a comforting headline number. If a device is going to claim it works for everyone, the proof has to be that it works for the people the field usually leaves out of the testing. For SpO2, our data supports that claim because we set the study up to be able to fail it, and it didn't.

The takeaway

Equity in measurement is not a slogan you can add to a brochure. It is a property you either built and tested for, or you didn't. We built for it, tested for it specifically, and the oxygen-saturation accuracy held across every skin tone we serve. For the women and families of North 24 Parganas, that is the difference between a reading they can trust and one that might quietly mislead them.