How Accurate Are Pulse Oximeters on Dark Skin?

Almost every page written about this sits behind a journal paywall or is addressed to clinicians. The underlying document is a New England Journal of Medicine analysis by Sjoding and colleagues, and its central number is the one above. The device that clips onto a finger in every clinic, emergency department and drugstore in the country reads differently depending on skin pigment, and it errs in the direction that produces false reassurance. The regulator has acknowledged the problem. It has not yet required a fix.

How much do pulse oximeters miss on dark skin?

About three times as often, measured directly against blood. Sjoding and colleagues found in the New England Journal of Medicine that when a pulse oximeter read between 92 and 96 percent, the blood itself was actually below 88 percent in 11.7 percent of Black patients and 3.6 percent of White patients. In a separate 178-hospital cohort, the figures were 17.0 percent and 6.2 percent.

Occult hypoxemia means the blood oxygen is genuinely low while the finger reading says it is fine. The word occult here means hidden, not rare. It is the specific failure mode that matters, because it does not look like an error. A device that reads erratically gets replaced. A device that reads confidently and wrongly gets believed.

Oxygen thresholds decide who gets escalated, who gets supplemental oxygen, and who gets admitted rather than sent home. A number that reads too high does not simply mislead one clinician. It removes a patient from a queue.

Why does skin pigment change the reading?

Because the device is an optical instrument, and melanin absorbs light. A pulse oximeter shines red and infrared light through the finger and infers oxygen saturation from the ratio absorbed by hemoglobin. Anything else in the light path that absorbs unevenly across those wavelengths distorts the ratio the machine is built to read.

Melanin is in that path. So is nail polish, so is poor circulation in cold hands, so is motion. The pigment effect is different from the others in one respect: it is systematic. A cold finger produces noise, which shows up as an unstable trace. Pigment produces bias, which shows up as a steady number that happens to be wrong in a consistent direction.

That matters for how the error persisted. A device validated against a narrow sample of skin tones can look accurate in testing and behave differently in a hospital corridor.

Does this affect every person with darker skin the same way?

No, and the honest answer is that nobody can tell you your personal error margin. The published cohorts compared patients by recorded race, not by measured skin pigment. That is what the data supports and it is where the data stops. The finding is a population-level bias, not a correction factor you can apply to your own reading.

This distinction is the difference between useful information and a new bad rule. Adding points to a reading because of someone’s race would be inventing a race-based adjustment at the exact moment medicine is dismantling them. The correct response is not a mental arithmetic fix. It is knowing that this specific number carries a known blind spot and that other information is available.

The same logic runs through the guide on high blood pressure in Black women. A population average describes a group. It never describes the person holding the cuff, or the finger in the clip.

Has the FDA fixed pulse oximeters?

No. The regulator’s response remains draft guidance, not a final standard. That means the devices sold and used today are not held to a corrected requirement, and no page should tell you otherwise. The problem is documented in the medical literature and acknowledged by the agency. It is not resolved.

The distance between “documented” and “required” is where patients live. A finding published in the New England Journal of Medicine changes what informed clinicians know. It does not, by itself, change what a manufacturer must demonstrate before selling a device, and it does not recall the units already installed on every ward.

So the practical status is this. The bias is real, it is published, and it is still present in ordinary care. Treat a reassuring reading as one input rather than a verdict, particularly when it disagrees with how a person looks and feels.

What can be measured instead of the finger clip?

Blood itself, and the patient. A pulse oximeter estimates saturation through the skin. Arterial blood measured in a laboratory reports it directly, which is why the discrepancy in the New England Journal of Medicine analysis could be quantified at all.

None of this is a test to demand. It is a description of what exists so that a conversation is possible. How each figure here was traced to its primary document is set out in how we research.

A blood test can be thrown off too. The American Diabetes Association advises using plasma glucose criteria rather than A1C when hemoglobin variants are present, the subject of the guide on sickle cell trait and the A1C test and part of the screening picture in type 2 diabetes in Black women.

What has medicine already retracted about race?

Several race-based defaults, formally and recently. That is the honest context for the oximeter finding, because it shows the pattern is not confined to one device. Skepticism about race-adjusted clinical numbers is now the position of the specialty societies themselves.

The kidney function estimate lost its race coefficient on the recommendation of a joint NKF-ASN task force. The American Thoracic Society recommended replacing race-specific spirometry equations with race-neutral ones, stating that race serves as a social construct. The calculator used to estimate the chance of a successful vaginal birth after cesarean dropped race and ethnicity as inputs.

Then there is the human layer underneath the algorithms. Hoffman and colleagues reported in PNAS that half of a sample of white medical students and residents endorsed false beliefs about biological differences between Black and white bodies. Those who endorsed them rated Black patients’ pain lower and made less accurate treatment recommendations. A device that under-reads and a clinician who under-rates pain are the same problem arriving through two doors.

Is a home pulse oximeter worth owning?

For most people, no. Here is the defense of that position in one sentence: the device’s known failure mode is false reassurance, and a false reassurance at home, with no clinician looking at you, is more dangerous than no number at all.

A home blood pressure cuff earns its place because the errors it makes show up as visible scatter. A home oximeter produces a clean, confident three-digit number whose documented bias runs in the direction of telling a Black user that a genuinely low oxygen level is fine. That is the exact scenario the New England Journal of Medicine data describes.

If a clinician has asked you to monitor saturation at home for a specific reason, that is a different situation and their instruction stands. What does not follow from the evidence is buying one for general reassurance. Reassurance is the thing this device is worst at delivering to you.

Common questions about pulse oximeter accuracy

Do pulse oximeters read high or low on dark skin?

High, which is the dangerous direction. The New England Journal of Medicine analysis found the device reported a reassuring saturation while arterial blood was genuinely low in 11.7 percent of Black patients versus 3.6 percent of White patients.

What is occult hypoxemia?

Blood oxygen that is genuinely low while the pulse oximeter displays a normal reading. Occult means hidden. It is not a device malfunction that looks obviously broken, which is precisely why it goes unnoticed.

Should I add a few points to my reading to correct for it?

No. The published data compares groups by recorded race, not by measured pigment, and supports no personal correction factor. Inventing one would recreate the race-based adjustments medicine has spent years removing.

Has the FDA banned or recalled inaccurate pulse oximeters?

No. The FDA’s response is still only draft guidance, not a final standard. Devices in current use are not held to a corrected accuracy requirement, so the bias remains present in ordinary care.

Sources

• Sjoding et al., New England Journal of Medicine, racial bias in pulse oximetry measurement. Published 2020. Checked when this page was last reviewed.
• Hoffman et al., PNAS, racial bias in pain assessment and false beliefs about biological differences. Published 2016. Checked when this page was last reviewed.
• NKF-ASN Task Force on Reassessing the Inclusion of Race in Diagnosing Kidney Disease. 2021. Checked when this page was last reviewed.
• American Thoracic Society statement on race-neutral pulmonary function test interpretation. 2023. Checked when this page was last reviewed.
• Removal of race and ethnicity from the VBAC calculator. 2025. Checked when this page was last reviewed.
• American Diabetes Association, Standards of Care, Recommendation 2.4 on hemoglobin variants and plasma glucose criteria. Current. Checked when this page was last reviewed.