The average error is small โ the medical thresholds sit right where it lands. On darker skin the reading runs about one or two points high 1. That barely matters until you notice where the important cutoffs are: 88% is the floor for safe oxygen, 92% triggers oxygen on a ward, 94% gated COVID treatment. A two-point shift across those numbers is the line between treated and sent home.
In real hospitals the small error became missed emergencies. Black patients whose clip read a comfortable 92โ96% were nearly three times as likely as white patients to actually be dangerously low on oxygen 2. Undetected low oxygen carried a 41% higher death rate 3. In one COVID cohort, 451 patients, mostly Black or Hispanic, never qualified for drugs by the device's number despite meeting the criterion by their actual blood 1. The bias also touches Hispanic, Asian, and South Asian readers, smaller but real 1.
The device is one input, not the verdict. Read it like a clinician who knows the bias.
The $20 pharmacy clip is a "wellness" product the FDA never checks for accuracy 5. Fine for curiosity โ a check after a poor night, oxygen at altitude. Not the device you want when an illness might be turning serious. For that, the over-the-counter medical oximeters cleared in 2023โ2024 (Masimo MightySat OTC, Nonin TruO2) were tested across the full range of skin tones. They run about $200โ300, roughly a single ER copay.
The fine print โ when to skip it, and what people get wrong
Oximeters don't read the same on everyone, and haven't for 35 years 6. The real fix isn't a race correction in software: race is a proxy for skin colour, so coding it in just moves the problem. It's validating hardware against real skin tones, which the FDA now requires of new devices 7.
Where it bites: a ward patient reading 92% is judged stable and gets less oxygen than a white patient at the same true level 8. Separate trap: oximeters read normal even at lethal carbon-monoxide levels, on any skin.
- 1Fawzy A, Wu TD, Wang K, Robinson ML, Farha J, Bradke A, Golden SH, Xu Y, Garibaldi BT (2022). Racial and Ethnic Discrepancy in Pulse Oximetry and Delayed Identification of Treatment Eligibility Among Patients With COVID-19. JAMA Internal Medicine. link
- 2Sjoding MW, Dickson RP, Iwashyna TJ, Gay SE, Valley TS (2020). Racial Bias in Pulse Oximetry Measurement. New England Journal of Medicine. link
- 3Wong AI, Charpignon M, Kim H, et al. (2021). Analysis of Discrepancies Between Pulse Oximetry and Arterial Oxygen Saturation Measurements by Race and Ethnicity and Association With Organ Dysfunction and Mortality. JAMA Network Open. link
- 4Bickler PE, Feiner JR, Severinghaus JW (2005). Effects of skin pigmentation on pulse oximeter accuracy at low saturation. Anesthesiology. link
- 5FDA (2021). Pulse Oximeter Accuracy and Limitations: FDA Safety Communication. link
- 6Jubran A, Tobin MJ (1990). Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent patients. Chest. link
- 7FDA (2025). Pulse Oximeters for Medical Purposes โ Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations; Draft Guidance. link
- 8Gottlieb ER, Ziegler J, Morley K, Rush B, Celi LA (2022). Assessment of Racial and Ethnic Differences in Oxygen Supplementation Among Patients in the Intensive Care Unit. JAMA Internal Medicine. link
Related in the handbook (6)
- โ Monitoring oxygen at home is part of living with COPD โ know that the finger clip reads high on darker skin, so treat borderline as low.
- โ A CO alarm matters because a pulse oximeter won't catch poisoning โ it reads normal even at deadly carbon-monoxide levels.
- โ Like the race-based kidney formula, the pulse oximeter carries a built-in bias by skin tone โ a sibling story in how tools fail quietly.
- โ Pulse oximeters have the same flaw โ they read falsely high on darker skin. Another number that needs a caveat before you trust it.
- โ Pulse oximetry is the bedside version of this problem โ a 'normal' SpO2 can hide a real low, especially on darker skin.
- โ Home sleep tests count overnight oxygen dips with the same finger sensor โ the pigment bias can quietly make your apnea look milder than it is.